c1d9ec5bbb993790f9e131a51908810ce6e723db
295 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 builtbykev
|
c1d9ec5bbb |
Path coverage: every branch from acquisition to the first grade callback
Audited the corridor rather than trusting the recorder. Nine upstream operations
sit between acquisition success and gradeAndCacheSlate — game binding, per-date
schedule fetch, roster index, attachEventIdentity, book-price capture, team-stat
refresh, hits-factor context, matchup keys — each with its own catch. NONE of
them early-returns, so the corridor always reaches the grader; but seven of them
were SILENT, and two could be misattributed.
COVERAGE WAS INCOMPLETE. Closed:
* BINDING — bound / unresolved / already-had, and its throw
* SCHEDULE — dates requested, game count, and its throw. A schedule outage
previously surfaced as an EVENT IDENTITY error because it lands in that
catch; it now records SCHEDULE_STAGE_ERROR and rethrows unchanged.
* ROSTER — indexed players/teams/failed and evidence-date-validity, and its
throw. Its catch only console.warn'd, so this stage was entirely invisible.
* DEDUPE per-reason accounting off the filter's OWN branches: invalid fields /
non-model book / duplicate identity / capped / not examined, plus
model-book eligibility. Counts reconcile to the input exactly.
* GRADE BOUNDARY — `reached` is derived from a candidate count and proves
nothing. GRADE_LOOP_ENTERED, FIRST_GRADEBESTSIDE_STARTED and
FIRST_ONGRADED_OBSERVED are now separate control-flow facts. No model
output is recorded; a test greps for p_win/grade/confidence/edge/side.
Terminal states now name the stage: SCHEDULE_STAGE_ERROR, ROSTER_STAGE_ERROR,
BINDING_STAGE_ERROR, EVENT_IDENTITY_STAGE_ERROR, ADMISSION_STAGE_ERROR,
DEDUPE_STAGE_ERROR, IDENTITY_ALL_UNRESOLVED, ALL_REJECTED,
DEDUPE_ALL_NON_MODEL_BOOK, DEDUPE_ALL_INVALID_FIELDS, DEDUPE_EMPTY_OTHER,
READY_FOR_GRADING, GRADE_LOOP_STARTED, FIRST_GRADE_CALLBACK_OBSERVED.
TRACE COMPLETENESS INVARIANT. `reconcilePregrade` — acquisition NONZERO +
CONTINUED with no correlated downstream state is an OBSERVABILITY_GAP, never a
pipeline verdict. This programme has twice read an absence as a conclusion
("MLB exited at acquisition", "all props rejected at admission"); both were
wrong. Now it is a typed state with tests.
dedupeProps takes an OPTIONAL stats object and increments on the branches it
already takes, in the same order — reused, never reimplemented. Without the
object it is byte-identical; a test asserts that.
A PRODUCTION-BREAKING BUG CAUGHT BY THE FULL SUITE: the frozen no-op recorder
did not implement gradeStarted/firstOnGraded, so any caller without a recorder
threw inside the grade loop — and gradeAndCacheSlate's catch turned that into
{written:false,count:0}. Every slate would have graded NOTHING, silently. Fixed,
NO_PREGRADE now covers the full recorder surface, and a test asserts it does.
Exception semantics unchanged throughout: every added catch records and RETHROWS
the identical error. Admission rules, dedupe predicates, MODEL_BOOKS, event
identity, gameBinder, gradeBestSide and its arguments: 0 changed lines.
eventIdentity, oddsService, retentionService, gameBinder, bookRoles,
analyzeViaEngine1, probabilityEstimator, snapshotScheduler and ledgerService:
UNCHANGED. Zero new external calls — the only diff hit is the existing
getScheduleWithPitchers line re-indented into its own try.
Twelve teeth, injections verified present, against a green baseline of 89:
upstream catch silent (1) · missing trace as failure (1) · admission exception
as ALL_REJECTED (2) · non-model-book as duplicate (3) · dedupe-empty as
rejection (1) · falsely says grading started (6) · onGraded unrecorded (1) ·
sport overwrite (2) · intraday overwrite (1) · different attempt id (3) ·
observer adds an external call (1) · store failure changes outcome (2).
Restored byte-identically; teeth 3/4/6 re-run after the NO_PREGRADE fix.
The first teeth pass ran against a baseline the finer states had invalidated;
six superseded assertions were updated first and the run repeated.
388 suites / 5,269 tests pass. web tsc exit 0. Lineage stays OFF.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CQJeAG8vcDoL5zkiaJyVb8
|
||
|
builtbykev
|
8d741e3932 |
Pre-grading stage trace: record which branch loses the cohort
The acquisition recorder proved the previous diagnosis wrong: the 03:00 MLB
attempt acquired 6,365 props from a propline cache hit and CONTINUED. Zero
reached the first grading callback, and nothing durable says why.
WHY REPLAY WAS IMPOSSIBLE. The 03:00 input was the cache object written
02:20:24.889Z. ODDS_CACHE_TTL defaults to 3600s, so it expired ~03:20; it is
now gone. No historical copy exists: closing_captures holds no MLB rows past
01:40:52, model_snapshots none, and `bookprices:mlb` carries no timestamp the
book-comparison route exposes. Calling /api/odds/mlb was REFUSED — a cold cache
would fetch and fire recordDownstream -> gradeAndCacheSlate, writing product
state and manufacturing a false recovery. So the input is NOT RECOVERABLE and
no replay was attempted.
A SECOND CLAIM IS WITHDRAWN. The previous tranche concluded "all 6,365 props
were rejected at event admission", reasoning that dedupe cannot empty a
non-empty list. That is FALSE: `dedupeProps` also FILTERS — it drops props with
no player/stat_type/line and any prop whose book is not a MODEL book. A fully
admitted cohort can still dedupe to zero. Demonstrated in test. So the failing
branch was never established, only assumed — which is exactly what the order
forbade, and why DEDUPE_EMPTY is a first-class outcome here.
THE REGION IS SMALL AND EVERY BRANCH LOOKS IDENTICAL FROM OUTSIDE:
identity annotation (runSnapshot, mlb only, own catch)
-> admitForGrading (pure, can throw)
-> dedupeProps (pure, can throw, ALSO filters)
-> mapLimit(gradeBestSide) <- first onGraded-capable call
gradeAndCacheSlate swallows every throw in that region and returns the same
{written:false,count:0} it returns for an honest zero.
The recorder distinguishes them: READY_FOR_GRADING · ALL_REJECTED ·
IDENTITY_STAGE_ERROR · ADMISSION_STAGE_ERROR · DEDUPE_STAGE_ERROR ·
DEDUPE_EMPTY · NO_INPUT_PROPS · OTHER_PREGRADING_ERROR. An exception is never
folded into ALL_REJECTED, and with no admission evidence the classifier refuses
to classify at all — a test pins that.
EXCEPTION SEMANTICS UNCHANGED. Each stage is wrapped to record and then RETHROW
the identical error, so the enclosing best-effort catch still handles it exactly
as before: nothing caught that was not caught, nothing swallowed that was not
swallowed. Admission and dedupe rules, the impossible-binding invariant, event
identity, model books and the first-row-wins cap are untouched — the only
behavioural lines in the diff are `const gate/unique` becoming `let`.
Correlated to the SAME snapshot_attempt_id the acquisition recorder minted — not
a new run id — and stored under its own key `ops:pregrade:{sport}` so it can
never displace the acquisition record. Same atomic LPUSH/LTRIM pattern, bounded
per sport, scheduled-only, best-effort at the call site, auto-disabled under
test.
CAUGHT DURING BUILD: `savePgTrace` was defined and NEVER CALLED — the recorder
would have persisted nothing, the same "built, correct, never invoked" failure
this programme has hit before. A test now drives the real runSnapshot and
asserts a trace is persisted carrying the acquisition's attempt id.
Ten teeth, injections verified present, against a green baseline:
empty collector read as ALL_REJECTED (2) · admission throw reported as
admitted=0 (1) · dedupe throw reported as output=0 (1) · dedupe-to-zero
mislabeled as rejection (1) · identity exception hidden (1) · observer reorders
the candidate array (1) · trace failure changes product outcome (1) · intraday
overwrites scheduled (1) · different attempt id (2) · secret leak (1).
Restored byte-identically.
THREE OF THOSE LANDED AND PASSED FIRST TIME — coverage holes, not safe defects:
the dedupe-throw and identity-throw tests only exercised the recorder directly,
never the real path, and nothing asserted the CANDIDATE array is not reordered.
All three closed with real-path tests, then re-run failing.
Two stale source assertions updated with the reason recorded: both pinned the
exact `const gate = …` / opts-key order that the recording wrapper changed.
387 suites / 5,237 tests pass. web tsc exit 0. Lineage stays OFF.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CQJeAG8vcDoL5zkiaJyVb8
|
||
|
builtbykev
|
238f0f67cf |
Scheduled acquisition trace: record the fork instead of inferring it
MLB stopped producing anything at the 22:00 and 01:00 slots on 2026-08-27 while NBA/WNBA ran normally. The differential narrowed it to exactly two runSnapshot exits — getOdds THREW, or getOdds RETURNED ZERO PROPS — and production retained nothing able to tell them apart. A failed acquisition has no snapshot_id, writes no ledger row and updates no slate, so it left no durable evidence at all. The alert channel could not fill the gap either: quota/test-alert reports sent:true while the ntfy topic replays 0 messages, so absence of alerts is not evidence. This records the decisions the existing control flow already makes. It changes no acquisition behaviour: provider order, the single existing retry, the quota threshold, the fallback and cache policy are all untouched. The only behavioural line in the diff is getOdds gaining an optional recorder argument, defaulted to a frozen no-op so every existing caller is byte-identical. IT MAKES NO PROVIDER CALL. Measured: zero added fetchAllOdds/getProps/gateway calls, and the trace module contains no HTTP of any kind. Tests assert both. WHY REDIS, NOT MEMORY. server.js arms the scheduler in EVERY process and there is no lock or leader election, and a rolling deploy demonstrably serves two containers at once — so process-local evidence could be written by a container nobody later probes. Storage is LPUSH + LTRIM, which is atomic: two schedulers racing the same slot both survive instead of one silently overwriting the other, and each attempt carries a process_generation so they stay distinguishable. WHY A BOUNDED HISTORY, NOT "LAST ACQUISITION". The scheduled MLB attempt failed at the hour and an intraday attempt SUCCEEDED ~20 minutes later. A single last-value would have erased the failure with the success — precisely the evidence needed. Only SCHEDULED attempts are retained (persist refuses any other trigger), each sport keeps its own key, and intraday structurally cannot write one because intradayRefreshService never calls runSnapshot. WHAT IS CAPTURED, per attempt: cache decision; PropLine outcome as NONZERO/ZERO/ERROR/NOT_ATTEMPTED with count; the odds-api fallback with allowed_at_invocation, blocked_reason and the quota AS OBSERVED AT THAT INVOCATION — reading provider quota hours later and calling it historical evidence is the exact mistake this exists to stop; then the final result and the runSnapshot terminal outcome. The EXISTING retry appears as a second attempt; no retry was added. Sanitized: keys, tokens, URLs and long opaque strings are redacted, and no prop payload is retained — counts only. Tests assert a dirty provider error and a real prop array both come out clean. BEST-EFFORT AT THE CALL SITE, not just in the default dep — a teeth proof showed an injected store could still throw into a healthy snapshot. Now any implementation is safe. persist also auto-disables under NODE_ENV=test unless a client is injected (the opsNotify precedent); without that the default path opened a real ioredis connection inside every suite driving runSnapshot. Read-only GET /api/internal/acquisition/:sport behind the existing internal auth. It runs no pipeline and makes no fetch. Nine teeth against a green baseline, injections verified present: intraday overwrites scheduled (1) · one global slot (2) · thrown getOdds with no terminal trace (2) · zero mislabeled as error (1) · quota not captured at invocation (1) · observer makes a provider call (1) · credential leak (2) · scheduled/intraday share an identity (1) · telemetry failure breaks the snapshot (1). Restored byte-identically. TWO OF THOSE LANDED AND PASSED FIRST TIME — coverage holes, not safe defects: the provider-call scan did not forbid getOdds, and nothing exercised a non-scheduled trigger through runSnapshot. Both closed, then re-run failing. Model, retention, event identity, admission, dedupe, ledger, lineage, cadence, quota tracker and the PropLine adapter are all UNCHANGED. Lineage stays OFF. 386 suites / 5,210 tests pass. web tsc exit 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CQJeAG8vcDoL5zkiaJyVb8 |
||
|
builtbykev
|
048e4eaa3f |
Event-aware retention identity: two games, two receipts
One player prop in Game 1 and the same-looking prop in Game 2 are two different
historical claims. The retention conflict identity did not know that.
SEMANTIC IDENTITY FIRST. Two outbound rows are the same retention proposition
within one cycle when they share the cycle, the EVENT, the participant, the
stat, the line and the side. Book is deliberately absent — collapsing books is
dedupeProps's actual job and the price anchor is chosen later. The database
index is enforcement of that answer, never the definition of it.
THE EVENT COMPONENT NEVER FABRICATES. canonical_event_id where a sport has a
resolver — MLB's admission gate rejects unresolved/ambiguous/contradicted props
BEFORE grading, so every row that can reach retention has one — and game_id
otherwise, which is NOT NULL in the schema and is the only event label sports
without a resolver possess. Both are in the identity, so the weaker label still
discriminates where the stronger is absent.
NULLS NOT DISTINCT IS LOAD-BEARING, NOT STYLISTIC. canonical_event_id is NULL
for every non-MLB row. Measured on a disposable PG17: under PostgreSQL's default
semantics the same NBA proposition inserted twice produced TWO rows — every
retry duplicating for ever. With NULLS NOT DISTINCT the same test yields one.
That measurement is what rejected the plain composite option.
MIXED-FLEET BRIDGE. A rollout serves both builds at once (measured 11/12 new,
1 old). Old and new writers need different indexes and NO schema state satisfies
both: with the legacy index present a new writer fails 23505 on a doubleheader;
with it gone an old writer fails 42P10. A bare ON CONFLICT DO NOTHING would have
bridged this, and PostgREST does not emit one — `ignoreDuplicates` WITHOUT
`onConflict` was measured raising a real duplicate-key error, so that bridge does
not exist through this client.
So the writer bridges it. It targets the event-aware identity and, on exactly
the two errors meaning "the schema is not in the state I expect" (42P10, or
23505 NAMING the legacy index), retries the SAME chunk on the legacy target. A
failed chunk rolls back atomically — measured 0 rows — so the retry cannot
double-write. Correct in every schema state: legacy-only and both-present
degrade to legacy semantics with no outage; new-only keeps both games.
The bridge is deliberately narrow. A supersedes conflict is ALSO a 23505, and
swallowing it would destroy the forked-history guard, so the legacy index must
be named. All three model_snapshots writers (persist, commitPublication,
recoverFromFork) go through it; no hardcoded legacy target survives.
MEASURED, through the real supabase-js -> PostgREST -> Postgres path on
production-shaped PG17:
* 1,000 REAL propositions from the verified 2026-08-17 STL@CIN doubleheader
(1,738 retained rows under ONE game_id), replayed across both real gamePks:
OLD index materialized 1,000 of 2,000 — 1,000 LOST. NEW index materialized
2,000 of 2,000 — 0 lost.
* retry idempotency, over/under, line, stat, player, non-MLB same-game and
non-MLB different-game all behave correctly under the new index.
* ORDINARY-SLATE PARITY over ALL 434 real cohorts / 328,262 retained rows:
old identities 328,262, new identities 328,262, delta 0, cohorts changed 0.
The index is therefore guaranteed creatable and nothing historical splits.
CONFLICT_IDENTITY is now DERIVED from RETENTION_CONFLICT rather than restated —
a test caught them silently disagreeing, which is exactly how the materialization
check could have expected an identity the database no longer enforced.
EXPAND/CONTRACT are separate files on purpose. 048 is additive and retires
nothing; 049 drops the legacy index and must not be applied until fleet
convergence is proven by sampling, never assumed from a fast rollout.
NO BACKFILL. Legacy rows keep NULL canonical_event_id and remain LEGACY
EVENT-AGNOSTIC RETENTION, which is what that NULL truthfully says.
The materialization defence is untouched and now reports the bridge honestly:
while the legacy index still collapses a doubleheader, expected 4 vs actual 2
yields MATERIALIZATION_MISSING and the cohort is refused.
Nine teeth, injections verified present, against a green baseline of 97:
1 event distinction removed (10) · 2 phases collapsed (2) · 3 bridge swallows
everything (6) · 4 NULLS NOT DISTINCT removed (1) · 5 old-container error as
success (3) · 6 semantic/DB identity disagree (8) · 7 collision detector removed
(2) · 8 partial transport usable (3) · 9 collision unannounced (1).
Restored byte-identically.
Model and product untouched: gradeSlateService (event-aware dedupe), event
identity, ledger, calibration, chain, lineage config and the status route all
UNCHANGED. Zero cacheSet changes, zero web paths, schema contract unchanged (no
new columns). Lineage stays OFF.
385 suites / 5,178 tests pass. web tsc exit 0.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CQJeAG8vcDoL5zkiaJyVb8
|
||
|
builtbykev
|
11277a1b99 |
Materialization truth: a completed write is not a complete cohort
Transport truth says every intended write succeeded. Materialization truth says
every identity that should exist actually exists. The retention writer could
only report the first, and the gap is not theoretical.
THE CONFLICT IDENTITY, traced to the real index:
model_snapshots_cycle_prop_uniq UNIQUE (snapshot_id, player_key, stat, line, side)
written by `upsert(..., { onConflict: same five columns, ignoreDuplicates: true })`.
Measured in production: no key column is ever NULL (0 of 328,262 rows), so
NULLS DISTINCT never applies and the identity is plain column equality. `line`
is an unconstrained numeric, so identity normalises it — 0.5 in and "0.50" out
must not read as two identities for one stored row.
PRIOR-CYCLE COLLISION IS IMPOSSIBLE. `snapshot_id` is in the identity and is a
fresh UUID per cycle, so no row can be suppressed by an earlier cycle.
Append-only chronology across cycles is safe, and `captured_at` is not in the
identity, so a cohort cannot be split by timing.
INTRA-CYCLE COLLISION IS REAL, AND WE CAUSED IT. `canonical_event_id` is NOT in
the identity. A doubleheader — same hitter, same stat, same line, two genuinely
different games — is ONE identity. Demonstrated through the real collector: 4
outbound rows, 2 distinct identities, 2 rows discarded by ignoreDuplicates with
no error, `written` counting all 4 and the terminal status reading COMPLETE.
Before event-aware dedupe the second game was dropped before grading, so the
collision could not arise; that fix moved the loss downstream into retention.
The conflict identity is NOT changed here — that is a separate decision with its
own before/after. This makes the loss visible instead of silent.
EXPECTED vs ACTUAL. `expectedMaterialization(rows)` derives the identity set
from the FINAL outbound payload using the exact database identity — never from
`attempted`, which counts rows sent, not identities that can exist.
`reconcileMaterialization` compares SETS, not counts: two sets of equal size can
still differ, and a cohort that swapped one identity for another passes every
count test ever written. A collision passes set equality by construction (the
discarded row was never in the expected set) while real rows were lost, so
collision_count > 0 fails the cohort on its own.
A cohort is evidence-complete only when transport is COMPLETE, missing = 0,
extra = 0, and collisions = 0.
OBSERVABILITY stayed minimal. `last_retention` was already PER SPORT (a Map
keyed by sport), so no fix was needed there and the route is UNCHANGED — the new
fields ride the existing entry: outbound_rows, expected_materialized_count,
outbound_collision_count, expected_identity_digest. Counts and a digest only,
never the identities, which carry player names. The expected set is the one
materialization fact unrecoverable from the database afterwards, which is why it
is the only thing recorded at runtime.
A collision leaves transport COMPLETE, so the existing failure alert could never
see it. It now has its own high-severity alert naming the counts, the cycle and
the build, and says the cohort is not evidence-complete.
Seven teeth, each injection verified present, against a GREEN baseline of 63:
1 attempted===written as evidence completeness -> 1 fail
2 COUNT(*) equality instead of set equality -> 1 fail
3 snapshot_id dropped from expected identity -> 4 fail
4 unexpected collision allowed to qualify -> 1 fail
5 single global last_retention slot -> 2 fail
6 partial chunk failure treated as usable -> 3 fail
7 collision loses its announcement -> 1 fail
Restored byte-identically (retention 742f116473d97f49, snapshot 81129facbabeb280).
Three brittle assertions repaired, with the reason recorded: two windowed on a
byte count that a neighbouring block outgrew — a test failing because of its
neighbour, not its subject — now windowed to syntactic landmarks; and one
counted TERMINAL.COMPLETE occurrences, which a legitimate comparison
incremented. It now asserts one DECISION and one READ.
persist() and createCollector are BYTE-IDENTICAL. onConflict and
ignoreDuplicates appear in the diff only as prose. Model, event, ledger,
calibration, chain, lineage config, and the status route: UNCHANGED. Zero
lineage/publication files, zero cacheSet changes, zero web paths. Lineage OFF.
Schema contract unchanged: release 64, prod 67, prod-only 3 (debt, not
authorized), missing in prod 0.
384 suites / 5,144 tests pass. web tsc exit 0.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CQJeAG8vcDoL5zkiaJyVb8
|
||
|
builtbykev
|
9809626c99 |
Retention completion: a cohort is complete only when the writer says N of N
The previous bug made the recorder write nothing. The dangerous successor is a
recorder that writes half and looks healthy: persist() writes in chunks of 250
and STOPS AT THE FIRST FAILED CHUNK, so chunks committed before the failure are
already durable. Rows exist under the snapshot_id, captured_at is uniform, Redis
kept working — and the cohort is short.
So row presence was never completion evidence, and neither was a matching
timestamp. Completeness is now proven by the writer or not at all.
TERMINAL RETENTION STATES (retentionService.classifyPersist):
NOTHING_TO_PERSIST attempted 0 — a refusal-only slate is still a cycle
SKIPPED_NO_DATABASE no database configured; not a failure
COMPLETE attempted > 0, written === attempted, no error
FAILED_ZERO_WRITE written === 0 — first chunk failed
FAILED_PARTIAL 0 < written < attempted — a later chunk failed
FAILED_UNRESOLVED_ERROR counts look complete but an error is unresolved;
unreachable through today's loop, and kept because
the alternative is reporting COMPLETE holding an error
The invariant: any written < attempted with attempted > 0 is a FAILED cycle. A
partial cohort is never degraded success.
classifyPersist reads the EXACT persist() result and refuses anything else — it
never recomputes attempted or written, because a second calculation could
disagree with the writer and then the status would describe a cycle that did not
happen. persist() itself is byte-identical to
|
||
|
builtbykev
|
35da190f2c |
Retention hotfix: drop published_side, derive the schema contract, break the silence
`createCollector.onPublished` set `published_side` beside `published`.
`published_side` is not a model_snapshots column. supabase-js declares the
UNION of row keys in the `columns=` parameter, so one invalid key made
PostgREST reject the ENTIRE batch with a 400 — every sport, every cycle.
Retention is best-effort, so nothing surfaced. Confirmed in edge logs.
The field was redundant as well as invalid: `side` is already on the row.
Deleted rather than added to the schema — a column would preserve an
accidental artifact.
Three things missed it, and each is now closed:
1. WRONG SHAPE INSPECTED. The manual check sampled the collector after
onGraded only and never called onPublished, so the offending key was
not yet on the row. It read a pre-publication shape and reported the
final outbound shape as clean. The new test captures the array actually
handed to .upsert(), after the full production call order.
2. NO CONTRACT. Every retention test injects a permissive fake client that
accepts any column set, so 381 suites proved the logic and never once
compared a row against the database. The contract is now DERIVED — the
migration chain applied to a disposable postgres, read out of
information_schema (scripts/generate-schema-contract.js). A
hand-maintained list would be a second opinion about the schema, and a
second opinion is what let this through. scripts/verify-schema-contract.js
checks the contract still describes a live database.
3. SILENT FAILURE. A failed batch reached one console.log. It now emits a
high-severity structured event carrying sport, snapshot id, stage,
error, code_sha and timestamp. Best-effort semantics are unchanged —
the product continues and says so — but the failure is observable.
`skipped` (no database configured) is not a failure and does not alert.
Teeth, each with the injection verified present before the run:
- published_side back into the final payload -> 4 tests fail; restored
byte-identically (sha 6a0ced7c52134135 both sides)
- settled_at (a REAL contract column) -> accepted, so the guard
discriminates by contract membership, not by novelty
- alert block deleted -> 3 tests fail; restored byte-identically
Model and product behaviour untouched: analyzeViaEngine1,
probabilityEstimator, gradeSlateService, lineageCanaryConfig all unchanged.
Lineage stays OFF. Net source change is one behavioural line plus the alert.
382 suites / 5,094 tests pass. web tsc exit 0 (zero web paths touched).
Measurement blackout recorded, NOT backfilled: last good retention write
2026-08-27T19:08:32Z;
|
||
|
builtbykev
|
ceaa896f77 |
Runtime observability: report the build and canary state the system acts on
The rollout stalled at RUNTIME_UNVERIFIED because two facts were answerable only
as a side effect of a scheduled snapshot writing a row: which build is running,
and whether MLB lineage is effectively enabled. Every state transition therefore
waited on cron rather than on asking the service.
- src/services/lineageCanaryConfig.js — THE canary resolver. Parsed once at
module load (unchanged semantics), normalised sorted/deduped/trimmed, frozen.
snapshotService's write gate now delegates to it, and the status probe reads
the SAME state. A route that parsed the environment itself would be a second
version of the truth, free to drift from the gate it claims to report.
- GET /api/internal/snapshot/status gains runtime.code_sha (the production
codeSha resolver — never git, never gitea/main; null when unavailable),
runtime.started_at (computed ONCE at module load, so it marks a boundary
rather than reading as now; deliberately not called deployed_at), and
lineage_canary {enabled, sports, configuration_source}.
- No raw environment value is returned; sports is the normalised set and
configuration_source says only ENVIRONMENT vs DEFAULT. Router-wide
requireInternalAuth is unchanged: 200 with key, 401 without.
- Effective lineage config is fixed for the process lifetime, so
runtime.started_at is a defensible lower bound for how long that state held.
Strictly observational — the handler still only reads Redis.
Model and decision code byte-identical to
|
||
|
builtbykev
|
8c6aef1e12 |
Event admission gate: an unresolved or contradicted game does not earn a Read
A fractured identity keeps two bad records from merging. It does not make an unknown game true. Until now a prop with an unresolved or verified-impossible event still continued into grading under that synthetic key; it no longer does. - event_binding_status contract: RESOLVED / UNRESOLVED / AMBIGUOUS / CONTRADICTED / UNSUPPORTED. CONTRADICTED and UNRESOLVED stay distinct — one means we know the association is wrong, the other that we do not know. - ONE admission gate (gradeSlateService.admitForGrading), before dedupe and before grading. Admission requires RESOLVED *and* a canonical_event_id; a legacy derived game_id can never satisfy it. Rejected props are returned, not discarded, so retention keeps them as evidence. - Roster evidence is now DATE-SCOPED. statsapi honours ?date= and it changes the answer (Joe Mack is on the 2026-08-26 Marlins roster, absent on 2026-04-15). Evidence that does not describe the slate's date can only yield UNRESOLVED, never CONTRADICTED — uncertainty must not become an accusation. - A mis-nested market is REFUSED, never re-bound. Knowing Joe Mack is a Marlin does not license moving a provider record into the Marlins game; that would invent provenance. Measured on the real 2026-08-26 slate: 2,878 props -> 2,874 admitted, 4 rejected (EVENT_PLAYER_TEAM_CONTRADICTION), each player's correct game still resolving. Doubleheader 824514/824478 both remain independently RESOLVED and admitted. No change to probability, projection, side, grade, confidence, ranking, normalization or calibration. Lineage remains disabled. Suite 380/5,061/0 from the release worktree; web tsc exit 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CQJeAG8vcDoL5zkiaJyVb8 |
||
|
builtbykev
|
352016790a |
MLB canonical event identity, impossible-binding refusal, event-aware dedupe, publication commit
Release-isolated slice built from
|
||
|
builtbykev
|
6c34af3414 |
checkpoint: chain shadow, WNBA possession feed, baseball chain
Backup commit of uncommitted working-tree state found during Legion recon (Tony resurrection, STEP 0). This work existed only on the laptop disk. - chain shadow accrual + probe script (038_chain_shadow.sql) - WNBA possession feed: ESPN adapter, usage service, verify script (039_wnba_player_game.sql) - baseball chain - retention/snapshot service updates, tableKeys, matchupKeys - specs: chain-v1, wnba-possession-feed, wnba-source-survey - unit tests for the above Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QnvJAkC3h5QGmb6dipoiWn |
||
|
builtbykev
|
0657b71d18 |
Drop lock_lines and retire its writer (dead table, dead writer)
lock_lines was built in Session 64 for a staleness audit - join lock-time
per-book lines to closing_captures and ask whether our locked line was
stale-high vs consensus. That audit was never written. ruler-comparison.sql
records why: only 43 settled rows ever joined it with >=2 two-sided books.
Measured before removal: 367,595 rows, 104 MB, zero readers in src/, scripts/
or web/src/ - the only from('lock_lines') was an upsert, every other mention a
comment. Zero dependents: no FK, no view, no trigger. The newest pg_dump held
all 367,595 rows, pg_restore-verified before the drop.
The write is off too, because a dead table that keeps refilling is only half
solved: it was accruing 36,440 rows/day, 10.3 MB/day, 23% of all database
growth, for a question nobody was asking. buildLockRows is kept and still
tested - the logic was never what was wrong, and re-arming is one flag plus
re-creating the table.
DB 510 MB -> 406 MB: 81% of the 500 MB cap, +94 MB headroom, under it for the
first time in months.
Moat and grade untouched.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
||
|
builtbykev
|
fb0010222d |
Stop persisting missed_window refusals (halt the bleed at source)
missed_window means 'this capture pass ran after first pitch'. It is a fact about our cron cadence, not the market: once a game starts the same prop emits a fresh refusal every ~20 minutes for the rest of the night, per book, per side. Measured over 7 days of production that is 332,608 rows/day - 83.1% of all closing_captures writes, ~66 MB/day - and since B1 filtered both readers, nothing reads them. The filter lives in persist(), not buildCaptureRows(), and that is the whole trick: the caller computes the capture-rate alarm from the full in-memory array, so filtering at build time would have blinded the ops alarm to the exact condition it exists to catch. captureRateAlarm is pure; a test asserts the caller still passes the full array, and that a 10-priced/90-late pass still fires at 0.10. Narrow by design: one_sided_price still persists (liquidity signal), priced captures unchanged, and the rare fault refusals still persist because each names a pipeline fault worth seeing. A missed_window row carrying a price is kept. Growth drops 400,469 -> 67,862 rows/day. The 4.2M historical rows are now static, so the cleanup is a calm decision rather than a race. Nothing deleted. Grade untouched. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
builtbykev
|
2271f46ab2 |
dCLV close leg: read priced captures only (the unfixed twin)
computeDirectionalForRow selected the latest closing_captures row with no missed_reason filter, while its sibling attachClosingProb has had one since the CLV instrument repair. closing_captures records a refusal for every prop x book x side on every cycle after first pitch, so a refusal always carries a later captured_at than the last real price - latest-first returned a refusal on 26,448 of 26,448 identity groups, and computeDirectionalClv refuses on a missedReason. That is why dclv_state has been 'unknown' on 100% of rows since Session 64. Measured on 600 real settled rows: 100% unknown becomes flat 45.8%, negative 23.0%, positive 22.8%, unknown 8.3%. ClvBadge will render MOVED TOWARD US 114 and MOVED AWAY 117 per 600 - near-symmetric, which is the honest shape. Existing rows do not recompute (first-computation-wins). The re-stamp is described in BUILD-STATE, not run. The grade is untouched. No rows deleted. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
builtbykev
|
f61ec6b391 |
Read integrity, as-of context, and the shadow matchup resolve (A1-A7)
Seven orders of measurement-first repair. The served grade does not move. A0/A1 — the unordered page walk returned the right COUNT and the wrong ROWS: 410-617 of 2,490 duplicated with an equal number never returned, while rows.length matched the server exactly. safePaginate orders on a real unique key, verifies the tuple at runtime, and THROWS on a query error instead of treating it as end-of-data. Both hits PROVES are withdrawn: they were drawn through that reader, and defense_by_direction's distinct-n was likely below the gate floor all along. A2/A2b — rolled across every reader: 11 FAIL -> 0. Composite keys pulled from pg_index (the context tables are dated-composite and had no single unique column). The unordered helper is deleted, not parked. A3 — ledgerService and retentionService defaulted the SAME env var to DIFFERENT versions, so no ledger row ever carried the marker eligibility requires. One source now. model_snapshots settlement moved onto the cron: 15,484 -> 28,894 settled, repaired-champion 0 -> 7,556. A4 — hitsFactorContext takes an as-of cutoff. Refusal over reconstruction: no row at-or-before the date means the factor does not apply, never the nearest row. Live path unchanged, proven 400/400 on real rows. A5 — factor_inputs freezes what the factor READ, never the multiplier, so an audit can recompute and check. It also recorded the finding: the three hits factors have NEVER fired. prop.opponent and prop.opposing_pitcher are read by the resolver and written by nothing. A6/A7 — matchupKeys resolves those keys from the posted lineup plus the schedule's probable pitchers, and fires the factors into a SHADOW freeze: 248 fires on 308 props, 245 of which would move the grade. The served forecast is untouched. specs/a8-shadow-factor-gate.md pre-registers the test that decides whether they ever go live. Nothing is turned on. CALIBRATION_DEPLOYED stays []. Both verdicts stay withdrawn. 4,772 tests / 371 suites green, web build exit 0, read-integrity harness 34/34. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
builtbykev
|
71d3b7b786 |
E10 Report issue template + E12 /report archive, to spec
PHASE 0 — the spec, read not recalled. E10: "Hybrid: dark billboard header that survives every client, light paper body Gmail can't wreck. 600px, stacked, no webfont dependence." Content law: "One email per slate day. Top read, what changed, the record. Nothing else." E12: "EVERY ISSUE SHOWS ITS OWN DAY RECORD -- THE ARCHIVE IS A LEDGER TOO." COMPOSED, NOT FORKED. The audit had E10 as PARTIAL, not absent: newsletterService already builds the daily report's CONTENT and lints its voice. What was missing is the designed hybrid SHELL, so reportTemplate.js is a template over that builder rather than a second report -- the same call made for the movement strip, and for the same reason. PHASE 1 — the hybrid shell is an ENGINEERING constraint, not a look, and the tests say so: Gmail strips style blocks, Outlook ignores flexbox, and a dark body renders as a black rectangle in several clients. Hence tables, inline styles, 600px fixed, system fonts, no image required to read, and the green SHIFTS from #00D4A0 to #00A57D on paper because the dark-mode green is unreadable there. FACT-CONTRACTED: a section whose data is absent is OMITTED and NAMED in `omitted`, never filled. There is no code path producing a placeholder figure. The honesty block carries the real numbers -- graded count, cleared-ceiling count, the realized rate against baseline, and that we do not issue A grades. E1'S LAW TRAVELS EVEN THOUGH ITS RENDERING CANNOT. An SVG strip is not reliable in email, so movementText carries the RULE: green only when the move favours the read, and a flat market says FLAT · [N]D rather than showing nothing. NO DESIGNER SAMPLE DATA. Nabers 1,120.5, No 128, DAY RECORD 9-4 are a spec for what a live issue renders; pasting them in would be fabrication carrying a designer's authority and would look entirely correct. Tested. PHASE 2 — /report is now the real archive, REPLACING the S41 redirect to /blog. That redirect existed because the surface did not; E12 built it, so the placeholder is correctly gone and the S41 test is updated rather than worked around. Every row carries its own day record, and an unknown record says UNSETTLED -- never a dash that reads as zero. Empty archive is an honest state. Backend: public read-only /api/report over Redis issues, plus the Next proxy. Both surfaces registered under the reachability guard. A test bug I made twice now: my check for forbidden sample values matched the template's own doc block, which NAMES those values as things never to paste. Documentation worth keeping, so both suites strip comments before matching -- a guard that reads its own warning is not reading the code. WAVE-2 STATUS: E1, F9-F11, E10, E12 done. Still gated -- F5 article media and E16/F8 on the card-system reconciliation; the in-season hub IA on the social chat's formula; E9/E15 on model; E2/E6 on licensing. Read-only throughout; serving fingerprint unchanged including newsletterService; accrual clock unchanged at 0 eligible dates. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
49e76068da |
Doctrine + E1 movement strip + F9-F11 offseason hub shell, to spec
PHASE 0 — specs/ARCHETYPE-TAXONOMY-DOCTRINE.md records the ruling as
shared law: 83 designed glyphs are the full four-sport taxonomy; a glyph
renders ONLY where its archetype is modeled and proven. 39 of 83 map to a
real archetype and are wired; the 44 unmapped are DORMANT SLOTS for
WNBA/NBA/Soccer, not a wiring gap. Wiring them would mean inventing 44
archetypes to consume artwork -- decoration presented as classification,
which is forbidden. DUAL THREAT and PAINT BOSS are modeled archetypes with
no mark: the mirror gap, flagged to the design side. When a sport's
archetypes ship, activation is a MANIFEST lookup, not new art.
PHASE 1 — E1 movement strip. The spec's own line is "the movement strip is
defined once here and reused everywhere a line has a past", so it is a
primitive, not a fourth chart.
RECONCILED RATHER THAN FORKED: lib/gradeShift.js ALREADY implements E1's
colour law -- toward/against/flat, including the direction flip that makes
an UNDER's favourable move the opposite sign of an OVER's. MovementStrip
CONSUMES buildGradeTimeline instead of reimplementing it, and a test
asserts it never redefines isUnder. GradeShift stays the grade-history
view; this is the reusable strip. That is the card-fork lesson applied
before it could happen again.
Spec laws honoured: STEPS NOT CURVES (H then V, no smoothing -- a curve
invents prices that never traded, and a test rejects any C/S/Q/T command);
green only when the move FAVOURS the read; FLAT renders as a hairline plus
FLAT · [N]D because a flat market is a finding; and too little history
says NO MOVEMENT HISTORY rather than rendering blank.
PHASE 2 — F9-F11 offseason hub shell, built from Vyndr Offseason.dc.html.
The spec's load-bearing words are used verbatim: "OUTLOOKS REPRICE ON NEWS
· NOT GAME ODDS" (an offseason number is not a game line), the QUIET WIRE
empty state ("No outlook-moving news since X. We don't manufacture
movement."), WHAT CHANGED TODAY as the hero with the countdown ambient and
top-right, the tag-colour-is-meaning row anatomy, the open -> NOW -> FAIR
triplet with the movement strip embedded, and the OUTLOOK ONLY block where
every row carries NOT GRADED.
THE DESIGN FILE'S SAMPLE DATA IS NOT IN THE COMPONENT. Wembanyama +420 ->
+330, Nabers cleared 11:42 AM, the Summer League names -- all of it is a
SPEC for what a live feed renders, and copying it in would be fabrication
carrying a designer's authority. A test asserts none of those strings
appear.
The IN-SEASON information architecture is NOT invented here. The spec
covers an offseason hub; nothing specifies how content, articles, wire and
the live slate share year-round navigation. That remains the open design
gap, and the route notes it.
Two test bugs caught and fixed: my first assertions matched my own doc
comments -- the ordering check found "WHAT CHANGED TODAY" in the header
block and the no-curves check caught the word "curve" in the sentence
explaining why curves are wrong. A guard that reads its own explanation is
not reading the render; both now strip comments first.
PHASE 3 — both surfaces registered under the reachability guard. Read-only
throughout, serving fingerprint unchanged, accrual clock unchanged at 0
eligible dates.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9
|
||
|
builtbykev
|
60469422af |
Wave D1 primitives — and the audit says E1/E9 were not Wave 1
PHASE 0 — the order proposed E1 + E9 as Wave-1 and told me to follow the
audit if it disagreed. It disagrees: E9 calibration curve is WAVE D3,
gated on MODEL work ("resolve n>=20 vs N30, accrue buckets"), and E1
movement strip is WAVE D6, a large surface build. E9's gate is live right
now -- calibration is WITHDRAWN at 0 eligible dates, so the curve could
only render its empty state today. Building it would ship a component
whose entire purpose is unavailable.
PHASE 1 — three of the five real Wave D1 items were ALREADY DONE, and the
2026-07-31 audit has aged:
D1 glyph library audit: 38/83 wired (46%)
now: COMPLETE for everything wireable -- 39 of 83
designed glyphs map to a real archetype, all 39
are wired, colours match the registry exactly
(0 disagreements).
A1 card token audit: BUILT-BUT-DRIFTED, "in only 1 file"
now: BUILT-TO-SPEC -- it IS the --bg-1 token,
consumed by 32 files. The audit counted literal
hex, which is what a correctly tokenised value
looks like.
B1 boundary blue audit: PARTIAL, hex in 2 files
now: BUILT -- --priced-out/#8fb2de is a token with a
documented colour law, 4 consumers.
The 44 unwired glyphs are NOT a wiring gap: they have no backend
archetype, so wiring them means inventing 44 archetypes to consume
artwork -- the fabrication this programme refuses. That is the 41-vs-74
scope question and it is Kev's call. Separately, 2 registry archetypes
have NO designed glyph (DUAL THREAT, PAINT BOSS) -- a design gap.
PHASE 2 — what was genuinely absent is now built. web/src/lib/motion.js:
nudge() capped at 180ms so it reads as acknowledgement rather than
latency; bootStagger capped at 240ms because uncapped, row 40 waits 1.1s
and the stagger BECOMES the latency it exists to disguise; rowHover
returns handlers not CSS so touch cannot stick a hover state; and
revealOnIntersect returns an unobserve in every path and reveals
IMMEDIATELY when there is no IntersectionObserver or motion is reduced --
content is never hidden behind a capability check.
Reduced motion is honoured, not softened. The sharpest of the 10 tests:
bootStagger under reduced motion returns opacity 1, not merely delay 0 --
if the CSS animation supplies the opacity, skipping it leaves the row
invisible forever.
PHASE 3 — the reachability guard gains a PRIMITIVES section: a module
built to be embedded must declare its exports AND name its intended
consumers, because a primitive imported by nothing is the same
built-but-unread class as an unmounted component.
WAVE-2 UNGATED: F9-F11 offseason hub, F5 article media, E10/E12 Report,
E1 movement strip. GATED: E9 + E15 on model, E16/F8 on the resolution tail
and the card-system reconciliation, E2/E6 on licensing, E13 on another
order.
Read-only throughout; serving fingerprint unchanged; accrual clock
unchanged at 0 eligible dates.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9
|
||
|
builtbykev
|
c575a708c7 |
Content studio API + preview page; widen the reachability guard; correct
two inventory errors INVENTORY CORRECTION, and it was mine. Phase 2's two "orphans" are NOT orphans -- my board grepped only web/src/app and missed component-level mounting. The transitive check says both are already mounted: BookComparisonPanel -> GradeResultCard -> app/scan/page.tsx NewsWire -> ExploreHub -> app/explore/page.tsx So book comparison is DONE (wired to /api/books, rendering on the grade card) and THE WIRE is DONE-BY-DESIGN, mounted in ExploreHub. Its header names an "Offseason Hub" as its home, and that hub genuinely does not exist -- but that is board item #8, not a mounting bug, and inventing a surface to satisfy a comment would be the wrong fix. The lesson is the same one this session keeps teaching: I checked one directory and reported a conclusion the check could not support. ALSO CAUGHT: I overwrote src/routes/content.js, which was the Session-29 content-templates route, by picking a filename without looking. Restored from git with no work lost; the new surface lives at /api/content-studio and both now coexist. PHASE 0/1 — /api/content-studio serves finished posts (copy, branded card, card_svg, the fact_contract each was REQUIRED to have, and the facts that actually backed it) plus a POST for editorial status in Redis. Private via internal key; the Next proxy holds the key server-side so the browser never does. /studio renders it as a thin client -- copy and card side by side with the fact contract visible, because reviewing copy by reading it is exactly how a wrong number ships. Never-blank: a night with nothing generated says so. API-FIRST is the point: the endpoint an autonomous poster will call is the one the page already renders, so the agent handoff is a pointer change, not a rebuild. Contract documented at docs/CONTENT-STUDIO-API.md. EXPRESS 5 BROKE 23 SUITES at first: `router.get('/:date?')` throws at mount time in Express 5, taking down everything that imports app.js. Two explicit routes instead. PHASE 3 — the reachability guard is widened from grade-fields-only to a general built-but-unread check. Book comparison, THE WIRE and the content studio are now registered surfaces; a page counts as its own entry point (Next mounts it by convention) while everything else must trace to one. 22 checks green; a registered-but-unimported surface still goes red. FULLY ISOLATED: read-only on model/slate/ledger, serving fingerprint verified unchanged, accrual clock unchanged at 0 eligible dates. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
74aa75945e |
Content engine: posts that structurally cannot lie
PHASE 0 — contentEngine makes Truth Law structural, not careful. Copy is
token-substituted and an unbacked {token} REFUSES to render -- there is no
code path that produces a plausible default. The fact contract is asserted
before any string is built. Card and copy render from ONE fact object, so
a caption and a card cannot disagree. No live model writes factual claims:
the voice is in the template, the facts are pulled, and the voice-polish
port is deliberately unwired, because an LLM that can rewrite a sentence
can rewrite a number.
18 tests carry the proof. The one that matters most: ZERO IS PRESENT.
"0 cleared B+" is our most honest possible post, and treating 0 as missing
would be the Number(null)===0 breach wearing its opposite coat -- it would
silently delete exactly the post the brand is built on.
PHASE 1 — three templates, generating real posts from tonight's data:
hot hitters off the repaired full-season log, the honesty flex off the
real servedGrade distribution (2,140 graded / 70 cleared B+ / 42% not
separable / A unissuable), and streaks verified from settled outcomes only.
THE ENGINE CAUGHT A BUG IN ITSELF, and it is the sharpest lesson here. The
first run published "No hitter is meaningfully hot tonight -- we could
dress up a middling week as a streak. We don't." That was FALSE: the
box-score cache spans only the settled window, every player had under 20
games, and the pool was empty. A broken pull was publishing as considered
editorial judgement -- the fourth appearance of this class tonight and the
first where our OWN HONESTY COPY was the disguise.
Fixed structurally rather than by patching the number: an absent() variant
may now DECLINE to speak, and the template separates "no candidates at
all" (SKIP with a reason) from "candidates judged, none hot" (honest
absence). Both locked by test. Source corrected to mlbStatsAdapter.fullLog,
the same log the repaired champion reads.
PHASE 2 — cardRenderer emits SVG rather than canvas: it is text, so it
diffs in review and its numbers are greppable, which matters when the
whole claim is that the numbers are real. VYND white + R green, slashed-Y,
scanlines, mono. The card never formats its own facts -- every string
arrives pre-rendered and gate-checked.
PHASE 3 — scripts/generate-content.js writes copy + card per template to
.content-out/<date>/. Template N+1 is a registry entry: requires, pull,
copy, card, absent. Queued as stubs, not built: hot takes, daily reads,
"grades we DIDN'T give", cross-sport streak variants (the streak template
is already sport-agnostic -- settled outcomes and a noun).
FULLY ISOLATED: read-only on every source, zero writes to serving, model
or ledger tables. Serving fingerprint verified unchanged. The accrual clock
is untouched at 0 eligible dates.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9
|
||
|
builtbykev
|
55b210cb95 |
Fix the dormant basketball window-bug before it ships; guard the class
PHASE 0 — audit. espnStatsAdapter's slice(0,20) was already fixed at |
||
|
builtbykev
|
981a05cbd6 |
Render-reachability guard: make built-but-unread a CI failure
Three consecutive orders shipped a backend-correct field that never
reached a screen, all on a green suite: gradeBands (required by no
serving code), served_grade (dropped at the adapter boundary),
GradeScaleLegend (imported by nothing). Each was caught by luck on a later
re-check, and in two of the three I had already reported the wiring done.
WHY GREEN TESTS COULD NOT SEE IT: backend tests stop at the API payload.
They prove a field is PRODUCED and say nothing about whether it is
CONSUMED. Invisible by construction, not an oversight in any one test.
THE TRAP, NAMED: the difficulty pools in the backend, so by the time a
field exists on the payload it feels finished. What remains is a
three-line adapter change nobody considers worth verifying, so it gets
claimed rather than traced. The last inch is the one with no friction,
which is exactly why it gets skipped. "I added the field" and "a user can
see it" are different claims and only the first is fun.
THE GUARD traces each promised field the whole way: payload -> adapter
consumes -> component renders -> component is MOUNTED. Mounted is
transitive to a Next entry point (page/layout/template), the only thing
that puts a pixel on screen, depth-limited so an import cycle cannot hang
the suite.
Container rows are exempted EXPLICITLY, not silently: served_grade carries
container:true plus a rendersVia list, and a separate assertion checks
every named part actually renders. The exemption is auditable and cannot
hide an unrendered field.
The guard tests itself -- an orphan component must report unmounted, and
the contract must be non-empty, since an empty contract passing vacuously
is how this would most plausibly rot.
RETRO-PROOF: run unchanged against
|
||
|
builtbykev
|
3591c7626e |
Total grade cutover + the ceiling stated as a position
PHASE 0 caught my own repeat of the failure I diagnosed one order ago.
|
||
|
builtbykev
|
91927a4a8a |
Serve an honest grade: the letter was carrying 1/6 the information of the
number beside it PHASE 0 corrects the order's premise. A grade letter has been served all along -- engine1.gradeProp builds it from an additive factor index, computed INDEPENDENTLY of p_win. gradeBands is orphaned for a different reason than assumed: it defines what a letter MEANS from realized outcomes, and every band collapses to base-rate at current resolution. The measurement that changed this order, on 3,417 settled props: grade n realized mean p_win A 8 0.500 0.647 <- the TOP grade did WORST B 985 0.640 0.700 C 1,695 0.602 0.676 D 303 0.558 0.604 F 426 0.535 0.588 letter resolution 0.00116 (0.48% of variance) p_win resolution 0.00715 (2.98%) -> the letter carried 0.16x the information of the number beside it Concretely, from the hand-verify: Christian Encarnacion's 0.95 over graded C and his 0.05 under ALSO graded C -- same hitter, opposite forecasts, same letter. The gap was never that grades don't ship; it is that the weaker of two available signals shipped as the headline. PHASE 1 — model/servedGrade.js derives the letter from p_win with bands anchored on MEASURED realized rates (B+ 0.663 / B 0.646 / C+ 0.615 / C 0.589 / C- 0.548 / D 0.512 / F 0.447, base 0.6005). NO MANUFACTURED A, structurally: A+/A/A- are UNISSUABLE, not rare. The realized rate plateaus at 0.65-0.68 above p_win 0.70, so no band has earned a top letter; a test sweeps every p_win 0..1 and asserts none produces one. Even 0.99 tops out at B+ with its realized 0.663 attached. Raising that ceiling later is a deliberate, visible act. Bands that cannot separate SAY so -- C+/C/C- carry separates_from_base_rate false and copy naming it, which is the honest description of a forecast explaining 3% of variance. Every grade states its basis (forecast_only vs forecast_plus_matchup_factors, naming which factors fired) and calibrated:false. engine1.grade is preserved as engine_grade so nothing downstream breaks. PHASE 2 — refusals render real states: insufficient_data -> "not enough history to call this one"; juiced_no_edge -> "the book has priced the vig past any edge on this side". 1,870 refused snapshots carry exactly those two reasons and both now surface. PHASE 3 — hand-verified on 12 real served props. Freeman/Rice/Encarnacion 0.95 overs now B+ (was B, C, B); the 0.05 unders now F (was C). Refused doubles render NO READ with their reason. never-blank PASS, no-manufactured-A PASS. Serving change; nine frozen model modules unchanged including engine1; p_win never mutated; no calibrated number leaks (deployed set empty); no Bonferroni slot. STILL TRUE: the forecast explains ~3% of outcome variance. This order did not make the model better. It made the letter stop overstating it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
494c83cf76 |
Hunt the window-bug class: three more paths, and the forward re-audit rule
in code PHASE 0 — getStatRows is the single base-rate path, so every branch is audited, plus the feature builders since l20_avg is the season reference projectionFor reads: getStatRows MLB -> estimator base fullLog CORRECT ( |
||
|
builtbykev
|
929fd81940 |
Repair the champion: it was reading ten games, not a season
PHASE 0 — the defect is real past the peek. Against a FAIR point-in-time baseline (each player's rate over games strictly before that date, >=10 prior games, box scores back to 05-01), the served champion LOSES on all four stats, three of four CIs excluding zero: hits 0.00251 vs 0.00774 CI [-0.0074,-0.0011] TB 0.00393 vs 0.00619 CI [-0.0055,-0.0003] rbi 0.02481 vs 0.03133 CI [-0.0153,-0.0005] runs 0.00181 vs 0.00683 CI [-0.0114,+0.0008] PHASE 1 — the cause is the WINDOW, not the weights. estimateProbability builds its base rate as the frequency over every row it is handed, and featureCache.getStatRows handed it res.last10. So the "season rate" was a TEN-GAME rate, and 0.4 of the forecast was the last five OF THOSE TEN. The 0.40 recency weight costs resolution on all four stats (-0.00086, -0.00107, -0.00562, -0.00365). Nudges are mixed and small -- harmful on hits and rbi, marginally helpful on TB and runs -- so they are left alone. PHASE 2 — two lines, no new data, no extra API call, because fullLog was already fetched by the same adapter call that produced last10: getStatRows now reads fullLog, and RECENCY_WEIGHT goes 0.40 -> 0.20. hits 0.00251 -> 0.00817 (tripled; now above the fair baseline) TB 0.00393 -> 0.00734 (above baseline; vs old CI [0.0020,0.0067]) rbi 0.02481 -> 0.02727 (still below baseline, CI includes zero) runs 0.00181 -> 0.00436 (still below baseline, CI includes zero) Gate stated exactly: hits and TB now exceed the fair baseline on the point estimate; rbi and runs remain below but EVERY CI now includes zero, so no stat reliably loses to a frequency table. That is a tie on rbi/runs, not a win, and it is reported as one. Only TB's improvement over the old champion is CI-confirmed; the rest are directional. STALE-FIT GATE: CALIBRATION_DEPLOYED is now EMPTY. The low-param maps were fitted on the retired forecast and fromLedger cannot rescue them -- settled ledger rows still carry OLD p_win, so refitting today would refit the retired forecast. Nothing is served calibrated until dates settle under the repaired champion, and the favourite-longshot bias must be re-measured rather than assumed to survive. The shadow duel is void. PHASE 3 — the hits factor lift is NOT re-measured, and cannot be yet: it needs settled rows produced BY the repaired champion, which ships in this commit. Replaying would score the factors against a reconstruction rather than the served forecast. Deferred, explicitly. The factors remain wired and transmitting; only their lift is unquantified on the new baseline. PHASE 4 — standing flag, and it is large: EVERY factor verdict in this programme, every null and every THEATER, was measured against a champion worse than a frequency table. Signal added to noise reads as noise. Prior verdicts may deserve re-audit. Logged, not re-run. Re-queued not built: rbi lineup-slot / RISP opportunity through the two-part gate, now landing on a repaired champion. Serving-path change by design; the byte-identical invariant inverted and all four stats move. Nine frozen model modules verified unchanged. No Bonferroni slot -- resolution accounting on the champion's own knobs. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
43f65d30cb |
Wire the three proven hits factors pre-grade: transmission proven, gain
inconclusive THE BUG THIS NEARLY SHIPPED AS A FINDING. The first audit reported 0 factors fired on all 1,140 rows. Not a result -- my paging helper ordered by `id`, and batter_spray, team_defense, platoon_splits and statcast_aggregates have composite primary keys with NO id column. The query errored, the loop broke on error, and four fully-populated tables read as empty. hitsFactorContext.js -- the PRODUCTION loader -- had the identical defect, so live wiring would have loaded nothing and served unadjusted while logging success. Third occurrence of this class in one session. Both loaders now order by a real column and THROW rather than degrade. The Phase 2 gate is what caught it: no resolution number was quoted until transmission was proved. PHASE 1 — pipeline is now base -> FACTORS -> CALIBRATE -> GRADE. Context built in snapshotService BEFORE gradeAndCacheSlate (was line 640+, grade at 454), threaded per prop, applied to p_over before p_win is set with p_win_prefactor and a full trace retained. Hits only. Coverage 859/1140 rows (75%): 474 with all three factors, 256 two, 129 one, 281 none. PHASE 2 — TRANSMISSION PROVEN, 12/12 sign-correct, 4/4 per factor, each applied IN ISOLATION. My first table compared each factor's expected sign against the COMPOSITE change and showed 3 false failures -- with three factors firing the net can oppose any single member; that was a flaw in the test, not the wiring. Two under-side rows confirm the flip is handled: a factor raising p(over) correctly lowers p_win. Switch hitters (Bailey, Bell, Rocchio) took no spray adjustment while their other factors fired normally -- the refusal is selective, not a blanket skip. PHASE 3/4 — both maps refit on the factor-adjusted forecast; the shadow-duel baseline is VOID and restarts, since it accumulated against a different forecast. Point-in-time, 765 held-out rows: reliability 0.00795 -> 0.00828 RESOLUTION 0.00229 -> 0.00345 (variance explained 0.93% -> 1.39%) Brier 0.25398 -> 0.25305 delta -0.00093 CI [-0.00225,+0.00002] Resolution rose 51% relative. The CI TOUCHES ZERO on 4 eval dates, so the composition does NOT earn a proven keep -- three isolated passes did not grant a composed pass. INCONCLUSIVE, reported as such. The gain is far below the sum of the isolated effects, which is expected: all three run through the same pitcher-batter confrontation and share signal. PHASE 5 — 1.39% of variance is still far below what band separation needs. The pivot was correct and incomplete: the plumbing defect was real and is fixed, three proven factors reach the served number for the first time, and transmission alone did not buy grade separation. Next arc is factor STRENGTH and BREADTH, not more plumbing. PHASE 6 — rbi anomaly logged, not chased: 14.51% variance explained vs hits 1.03%, on the stat we do not serve corrected and which has no proven factors. Either the biggest lever on the board or a mirage; it deserves its own order. The byte-identical invariant INVERTED for hits by design. All 13 frozen non-hits modules verified unchanged, probabilityEstimator included -- the factors ride outside it. No new Bonferroni slot; the composed OOS claim is reported with its CI and not claimed as a pass. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
e872eff4ce |
Instrument the calibration duel forward; diagnose the resolution ceiling
— the proven factors were never wired in
PHASE 0 — two truths recorded. The swap is a BET, not an OOS win:
isotonic beat low-param on identical held-out rows (hits +0.0028, rbi
+0.0042, TB tied) and we serve low-param anyway on an untestable prior
about shared daily structure. At 19 dates nothing here can test it. And
the MIN_SLOPE catch is preserved as standing rationale: a near-zero or
negative slope collapses toward base-rate-for-everything, which LOWERS
Brier while destroying all resolution -- a metric win that guts the
product.
PHASE 1 — the duel is now falsifiable. Both corrections computed on every
hits/TB prop; p_win_lowparam served, p_win_isotonic_shadow logged in its
own try so it can never break serving. calibrationDuel.adjudicate encodes
the rule IN CODE before any forward date exists: >=10 forward dates and
isotonic winning with a date-block CI excluding zero => REFUTED, revert;
otherwise UPHELD; under 10 dates PENDING regardless of the numbers. A
date counts as forward only if NEITHER map was fitted on it -- otherwise
we would be scoring which map memorised better. Nothing swaps now.
PHASE 2 — the ceiling, quantified via Murphy decomposition:
stat reliability RESOLUTION uncertainty variance explained
hits 0.01353 0.00252 0.24532 1.03%
TB 0.01419 0.00442 0.24329 1.82%
rbi 0.00654 0.03268 0.22531 14.51%
runs 0.00788 0.00130 0.23182 0.56%
Calibration did exactly what theory says and nothing more: hits
reliability 0.01353 -> 0.00233 (-0.0112, 83% of the error removed) while
resolution moved -0.0002. Unexpected: rbi has 13x the resolution of hits
and is the one stat we do NOT serve corrected -- it needs calibration
least and discriminates most.
PHASE 2 DIAGNOSIS — NOT-TRANSMITTED, and not weak, ABSENT. Traced in code:
sprayDefense.js and platoonSeverity.js are required by NOTHING in src/,
only by analysis scripts and their own tests. The served p_win
(intelligence/probabilityEstimator.js:54) reads exactly four inputs --
game-log frequency, opp_rank_stat +/-0.03, home_away +/-0.015, and a cv
pull -- with zero occurrences of spray, platoon, hard-hit or
contact-profile. And snapshotService grades at line 454 while computing
challenger/context at 640+, so everything proven is computed DOWNSTREAM of
the grade it would inform. The three proven hits factors have never once
moved a served number.
That reframes the recent nulls: "calibrated p_win does not separate within
archetype" was never a statement about factors. The factors were not in
the forecast.
PHASE 3 — bands rebuilt on SERVED values (hits/TB low-param, rbi/runs
raw): 28 archetype slots across four stats, ZERO show lift. No longer an
open shrug -- it is the arithmetic of resolution 0.0013-0.0327 against
uncertainty ~0.23. A forecast explaining 1% of variance cannot produce
separating bands, and no correction to its numbers will change that.
HEADLINE: calibration is complete, delivered honest numbers on two stats
and zero grade separation, because the counter has no resolution -- and
the proven factors are not wired into the forecast at all. The second is
the reason for the first, and it is plumbing rather than a modelling wall.
Per-archetype grades need proven factors that actually reach p_win. Last
calibration order.
Serving unchanged from
|
||
|
builtbykev
|
74cf1ce974 |
Robust bias established; low-parameter correction replaces isotonic
PHASE 0 — sample-limit truth on record: on 19 dates BOTH stability
instruments are underpowered. LODO power 0.014-0.093 (best 0.337 across
every k tried); deploy CIs rest on 2-4 date clusters, where a
cluster-robust interval has ~1 df. This is the SAMPLE, not a fixable
instrument, and the gate-refinement loop stops here. Runs corrected: its
DATE-DRIVEN label was an artefact of the coin-flip ruler (2 reversals in
3 drops never cleared cutoff 2) -- it is an ordinary no-fittable-map
refusal.
PHASE 1 — the bias is ROBUST, tested model-free and map-free with a
date-block bootstrap. Pooled over-prediction rises monotonically -0.0076
/ +0.0428 / +0.0963 / +0.1589 / +0.2451 across deciles from 0.5 to 1.0,
sign stability 0.9946 over 17 date blocks, and 4 of 4 stats replicate
(bar was 3). Also visible: realized rate PLATEAUS at 0.65-0.68 from p=0.7
upward -- the 0.9+ bucket (0.6624) does no better than the 0.8-0.9 bucket
(0.6841). The model has no high-confidence reads, only high-confidence
numbers.
PHASE 3 — Platt, two parameters over the whole curve, shrunk toward
identity by fit-date count. Validated as a NEW estimator vs RAW with
date-block CIs:
hits a=0.406 shrink 0.565 0.2626 -> 0.2540 CI [-0.0112,-0.0069] DEPLOY
total_bases a=0.472 shrink 0.333 0.2490 -> 0.2429 CI [-0.0062,-0.0059] DEPLOY
rbi a=0.775 shrink 0.231 0.2011 -> 0.2007 CI [-0.0007, 0] REFUSE
runs a=-0.032 REFUSE
A GUARD THE FIRST RUN NEEDED: runs fitted a = -0.032. A non-positive
slope inverts the forecast rather than flattening it, and near zero the
curve collapses to a constant predicting the base rate for everything --
which LOWERS Brier while destroying all resolution. It would have scored
as a win while making the product worthless. MIN_SLOPE now refuses it by
name, with a test.
STATED PLAINLY: on the identical held-out rows isotonic BEAT the
low-param on hits (+0.0028) and rbi (+0.0042) and tied on TB. The swap is
a CAPACITY JUDGEMENT, not a measurement -- the window spans 2-4 date
blocks and that is exactly what a flexible map produces when it captures
structure shared by fit and eval. Labelled as a judgement.
PHASE 4 — hits and total_bases serve the correction, basis
direction_robust_magnitude_provisional (direction bootstrap-robust,
magnitude thin-sample and shrunk). rbi is WITHDRAWN to raw -- it was
deployed on isotonic at
|
||
|
builtbykev
|
ced40421ed |
Audit the LODO instrument: it cannot evaluate any stat, and both prior
FAILs were false
PHASE 0 — the gate at
|
||
|
builtbykev
|
1f40014256 |
Power-derive the LODO threshold: hits restored through the gate, rbi/runs
routed as date-driven PHASE 0 — threshold derived BLIND, before any stat was re-read. A reversal is informative only if that date's Brier delta is distinguishable from zero at its row count. Per-row Brier difference d_i = (pc-y)^2 - (p-y)^2, so SE(n) = SD(d)/sqrt(n) and n* = (SD(d)/|effect|)^2. Pooled across all four stats so no single stat's verdict could shape the threshold deciding it: pooled rows 3,417 | SD(d) 0.09816 | |effect| 0.01175 n* = (0.09816/0.01175)^2 = 69.8 -> 70 The hand-chosen 20 sat at 0.54 SE -- a coin flip. That is the defect this removes, and why the previous verdict moved with the number. Committed as calibrationRegistry.LODO_MIN_HELD_ROWS = 70 with LODO_THRESHOLD_BASIS; a test recomputes (SD/effect)^2 and asserts it equals the constant, so it cannot drift from its own justification. The derivation script prints no stat verdict, no date and no reversal. PHASE 1 — LODO at n*, applied cold: hits 5 informative drops, 0 reversals PASS total_bases 4 informative drops, 0 reversals PASS rbi reverses 2026-08-01 (n=99) FAIL runs reverses 08-01 (n=86), 08-05 (244) FAIL hits held-out deltas -0.0041/-0.0080/-0.0192/-0.0140/-0.0139 across 123-272 row dates, favourite sign holding on every testable drop. THIS IS THE INSTRUMENT FINALLY POWERED, NOT VINDICATION OF A PREDICTION -- the withdrawal at |
||
|
builtbykev
|
6ae11f1193 |
LODO-gated provisional calibration: total_bases deploys, hits withdrawn
PHASE 0 — I applied factorGate's >=40 date-cluster floor to a calibration layer without challenging the binding. That floor is a cluster-robust interval bar for a CAUSAL claim. Calibration makes no causal claim, has a bounded failure mode (it can only over- or under-shrink) and consumes no Bonferroni slot. Its real risk is that the correction is DATE-DRIVEN, and leave-one-date-out tests that directly -- a STRICTER bar, since a cluster count cannot detect a single day carrying the effect. The >=40 floor is retained, correctly scoped as the PROMOTION bar. PHASE 1 — both guards codified, 11 tests, green before Phase 2. Demonstrated on live data: raw population violated=true, mean_p 0.4962, both_sides_share 0.9763; after dedup violated=false, mean_p 0.6694. The null guard's test demonstrates the trap explicitly, since (null-1)**2 is 1 and (null-0)**2 is 0 so a Brier over nulls equals the win rate. PHASE 2 — LODO: hits n=1140 dates=17 2 reversals (07-22 n=20, 07-26 n=25) FAIL total_bases n=1050 dates=7 0 reversals, 0 sign flips PASS rbi n= 630 dates=5 1 reversal (08-01 n=99) FAIL runs n= 597 dates=5 2 reversals (08-01 n=86, 08-05 n=244) FAIL Threshold sensitivity reported because the verdict moves: total_bases passes at every held-size threshold, runs fails at every one, and hits fails ONLY when 20/25-row dates are admitted. I fixed MIN_HELD_ROWS=20 before seeing which stats passed and did not move it afterwards to preserve a deploy. Honest caveat: a per-date Brier delta on 20 rows has a standard error several times the effect, so the instrument is underpowered per-drop -- an argument for pre-registering a higher threshold, which is a Roundtable call, not one to make while holding the results. PHASE 3 — total_bases DEPLOY-PROVISIONAL, band [0.6-0.8]. hits, rbi and runs REFUSE. HITS WAS BEING SERVED CALIBRATED AND IS NOT ANY MORE. snapshotService hardcoded it since S91; it fails LODO, so it is out. A stat that cannot survive dropping one day was never calibrated, it was fitted to that day. The consequence is real -- hits props become unstackable for chain.chainAcross -- and it errs toward withdrawing a claim rather than preserving one on a fragile verdict. Deployment is now driven by a frozen, tested CALIBRATION_DEPLOYED set, not a hardcoded stat name. PHASE 4 — calibrationRegistry, 14 tests. Deploy needs BOTH gates, neither waivable. reverify auto-demotes on the first breach (CI stops excluding zero, or the favourite bias flips sign) and logs the breaking date. Promotion needs the original >=40 bar. A provisional deploy that cannot be taken away is just a deploy. PHASE 5 — TB bands rebuilt on calibrated values, 625 eval rows. The two-bar rule still bites: calibrated YES, proven NO, so they stay a base-rate read, now honestly numbered. Every archetype still collapses to one band -- calibrated p_win separates within archetype no better than raw. PHASE 6 logged only: the dead gradient is buried (hits~TB > runs > RBI, and RBI has the SMALLEST bias, so the skill-driven-gradient mechanism did not survive); the refused set is a map of missing inputs; a low-parameter calibrator is queued unbuilt. p_win never mutated; calibration rides as p_win_calibrated with calibration_status provisional. No Bonferroni slot consumed. Counter and frozen clusters byte-identical. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
b2e4c6c4fb |
Link 2 at the coarse grain: pen QUALITY proves, archetype does not
The refinement was right. Naming the individual reliever failed; the same
question at the grain the chain needs passes, and it transmits more than
anything else measured in this chain.
WHY IT WAS WORTH RE-ASKING: last session's null (the pen is on average no
softer, +0.0010 on 35,760 PAs) does NOT rule this out, and treating it as
though it did would have been the error. An average washing out is fully
consistent with quality VARIATION mattering. It does -- actual arm quality
moves the hit rate monotonically across quartiles, 0.2244 / 0.2293 /
0.2410 / 0.2501, a 2.57pp spread, larger than the whole times-through-
the-order effect.
CLUSTER UNIT CORRECTED, THEN CHECKED RATHER THAN ARGUED. Last session
refused Link 2 partly as team-borne (30 bullpens, the park ceiling). My
first re-check was that 76% of pen-quality variance is within-team -- but
that is a statement about TREATMENT variance, not about where errors
correlate, and stopping there would have been picking the convenient
answer. Measured the actual thing: ICC of prediction error by team =
0.0261, design effect 1.41, SEs inflated ~19%. So the verdict was run
three ways:
unclustered CI [-0.0067,-0.0010] excludes zero
team-clustered (30) CI [-0.0086,-0.0003] excludes zero (below the
40-cluster floor -- indicative, not a pass)
design-effect adjusted CI [-0.0072,-0.0005] excludes zero
QUALITY GRAIN PROVES on the concentrated elevated-early-exit subset:
n=501 team-games, 426 clusters, MAE 0.0294 -> 0.0260, delta -0.0034, CI
[-0.0063,-0.0005] at 110 cumulative tests. Pooled also proves, so it is
not a subset artefact.
ARCHETYPE GRAIN DOES NOT: 0.5669 vs a 0.5309 modal-guess baseline,
corrected interval [-0.1073,+0.0268] spans zero. Two grains tested, one
earned a place -- penQuality.js exposes no archetype and a test asserts
it.
WHAT LINK 3 RECEIVES, which is the number that actually matters -- not
the MAE gain but realized outcome separation, prediction strictly
point-in-time:
predicted BEST pen 167 games 2,044 PAs hit rate 0.2231 +/-0.0180
predicted WORST pen 167 games 1,799 PAs hit rate 0.2501 +/-0.0200
2.70pp separated, intervals non-overlapping, capturing nearly all the
2.57pp available at the quartile grain. Caveat stated not buried: the
tercile cut is chosen in-sample; the prediction driving it is not.
BUILT: penQuality.js + 9 tests. Abstains below 5 prior club games and 40
arm appearances -- a league-average stand-in would assert "this is an
ordinary bullpen", which is a claim, and usually the wrong one for exactly
the clubs whose pens just turned over.
Link 3 is unblocked on a proven Link 2 at the quality grain only. Not run
here; this order scopes to building and gating Link 2.
Parallel track logged unchanged: TB n=948 pooled, BOMBER x TB 340, short
by 160.
Counter and frozen clusters byte-identical.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9
|
||
|
builtbykev
|
e4dae0e6b0 |
Reliever chain: Link 1 proves, Link 2 does not, and the premise inverts
The causal insight is right -- the game is a sequence and the matchup does shift mid-game. The direction is backwards, measured on 93,663 plate appearances from 1,238 games pulled free from statsapi. LINK 1 PROVES. Starter batters-faced, point-in-time from his own prior starts only, clustered on the pitcher: MAE 3.2226 -> 2.7990, delta -0.4236, CI [-0.6006,-0.2731] at 0.9995 corrected for 107 tests, 1,706 starts across 204 pitchers. It finds the tail the chain needed -- early exits are a 23.2% base rate, model-flagged starts are 34.0% early, lift +10.8pp. Scope correction inside Link 1: the order specifies fatigue x GAME SCRIPT, but game script is not available at grade time -- whether he gets hit tonight is the thing being projected, not an input to it. Only the workload half is measured; the in-game half is recorded as a live feature, out of scope, rather than quietly folded in. LINK 2 DOES NOT PROVE, twice over. Model accuracy 17.2% vs an 8.6% baseline -- doubling it sounds good and is not, since naming a specific arm is wrong five times in six. And structurally the entity is the BULLPEN: 39,629 post-starter plate appearances across 30 clubs is 30 readings, below the 40-cluster floor, the same permanent ceiling as park geometry and team defence. LINK 3 NOT RUN, per the order's own rule. THE PREMISE IS REFUTED, and this chains on nothing so it was safe to measure: vs STARTER n=48,492 hit rate 0.2444 +/-0.0038 vs BULLPEN n=35,760 hit rate 0.2373 +/-0.0044 The pen is 0.7pp HARDER. The specific effect the chain exists to exploit -- early exit making later at-bats softer -- is +0.0010 on 35,760 PAs. A well-powered null, not a sample problem. What IS real is times through the order: TTO1 0.2351 -> TTO2 0.2515 -> TTO3 0.2518. A starter does decay as the lineup sees him again, but that advantage is SURRENDERED when he leaves, not extended -- the pen is harder than his second and third time through. A modern bullpen is a queue of fresh specialists throwing one inning each; there is no tiring arm to punish. So the insight survives inverted, and Link 1 stays valuable for the opposite reason it was built: a likely early hook predicts the hitter LOSES his third-time-through look (0.2518 -> 0.2373 on that PA). The mispricing is on hitters who get an EXTRA look at a starter going deep. BUILT: predictionGate.js + tests -- the two-part gate for a continuous prediction. factorGate binarises outcomes for Brier, which would destroy a target like batters faced. Same discipline, same THEATER verdict, real scale. PRE-REGISTERED NOT RUN: Link 2' using a PA-weighted bullpen AGGREGATE rather than a named arm. Recorded rather than substituted in -- running Link 3 on a swapped-in Link 2 is the assumed-link failure the order forbids. Given the premise result its expected value is now low. PARALLEL TRACK logged: total_bases n=948 pooled, BOMBER x TB 340, short by 160. Sample-readiness only, not a verdict. Counter and frozen clusters byte-identical. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
3081c92e00 |
Per-archetype grade bands: built, gated, and the rescale blocked twice
The premise does not hold. proven-status.js run fresh: PROVEN_SET is EMPTY, no archetype x stat reaches the gate. pitcher_contact_profile has a CI upper bound of exactly 0.0000 and platoon_severity is held on 4.5%-contaminated splits, so the proven set is one factor, pooled, not three archetype-conditioned ones. The specific pattern the order names -- defense strong for GHOST/BRUSH, null for BOMBER -- is the one I measured running the OTHER WAY yesterday, both noise-dominated. But the second blocker is new and matters more, because it would stop the rescale even if the factors had proved: the grade does not separate within any archetype. Every archetype collapses to ONE band at the corrected bar, because bands merge when their intervals overlap and publishing two letters we cannot tell apart is a distinction we have not measured. Uncorrected, so the ranking is visible rather than hidden by the bar, this INVERTS the order's design. The order gives contact types the factor-rich treatment and power types honest base-rate, reasoning that single-game hits are variance for a power profile. Measured: BOMBER n=466 corr(p_win,outcome) +0.207 quintiles 0.75 0.62 0.60 0.48 0.48 GHOST n=192 corr(p_win,outcome) -0.007 quintiles 0.47 0.63 0.74 0.58 0.45 BOMBER is the one archetype the model ranks, and it splits into a real A 0.660 / B 0.481 at 95%. GHOST is flat, and non-monotone -- its most confident reads hit 47% while its middle reads hit 74%. Shipping as specified would have given the factor-rich treatment to the archetype the model reads worst and left base-rate on the one it reads best. That is mechanically sensible in hindsight: a power hitter's hit tracks whether he can damage the arm, a contact hitter's depends on balls finding holes. BOMBER's split does not survive the cumulative correction at 106 tests. Exposing it by loosening the correction is the curve-to-make-A's the order forbids, so it stays one band. BUILT: gradeBands.js -- lift against the archetype's OWN base rate (the same 62% is lift for a 45% profile and a deficit for a 68% one), indistinguishable neighbours merged, thin bands PROVISIONAL not dropped, Wilson intervals widened by the cumulative correction. The two-bar rule is structural: proven-alone, calibrated-alone and neither all return base_rate with the reason stated, so with nothing proven no factor-informed band can be produced at all. reasoning() is built and tested but NOT wired to the card -- there is no per-archetype band being served, so attaching the copy now would ship product language for a rescale that does not exist. NOT BUILT: the specified power-type reason "the matchup edge is in total_bases". total_bases is recorded INCONCLUSIVE (+0.0038, CI [-0.068,+0.075]). Wiring it would assert an edge measured as indistinguishable from zero -- the exact fabricated-reason failure this module exists to prevent. BOMBER x hits is 29 rows short of the gate and is the archetype the model actually reads. That is the first slot to test, not GHOST. Counter and frozen clusters byte-identical. No letter was moved. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
6b17f79367 |
Per-archetype re-audit: no slot reaches 500, and the replication unit
decided everything The premise does not hold. prove-hit-factors.js has no date filter anywhere in it and pages the full table -- there was never a window to widen. Full clean history is 1,266 rows, not 2,715. platoon was not "proved" last session, it was explicitly held on 4.5%-median-contaminated season-to-date splits, and pitcher_contact_profile was demoted. The proven set going in was one factor, not three. STEP 1: no archetype slot reaches n>=500 on full history. Best is BOMBER at 408, and BOMBER is the most common archetype on the board. GHOST 173, BRUSH 64, DRIVER 43, CATALYST 16. These are confirmed genuinely short, not artifacts. STEP 2 is where the real finding is. park_hits initially PROVED at 619 rows across 45 games -- but those games only ever visited 14 distinct park values. A park effect is replicated across parks, and unmodelled park heterogeneity is confounded with the thing being estimated. Each factor is now clustered on the coarser of the game and the entity its treatment rides on. That flipped two verdicts and confirms Kev's causal-correctness thesis from a new direction: defense_by_direction has 442 hitter-team units of replication where crude team defense has 26. The correct atom is not just more accurate, it is the only one measurable at all. park_hits (14) and defense (26) can never be validated however long the ledger runs -- the same ceiling as park dimensions, reached independently. Also fixed a bar I got wrong last session: I transplanted the 500-row floor onto clusters, which refused a factor with 1,059 rows over 85 games while answering neither question. Two floors now -- rows>=500 for a stable estimate, clusters>=40 for a trustworthy interval. Not a lowered bar: park_hits and defense are still refused. PROVEN: defense_by_direction only, pooled, [-0.0054,-0.0012] at 99 tests. It stays POOLED-ONLY -- no per-archetype reasoning wired, nothing grandfathered. The card must not say "GHOST: defence matchup strong" because we have not earned that sentence. The predicted fingerprint did not appear either: BOMBER -0.0036 vs GHOST -0.0024, the opposite direction, both noise-dominated. Recorded so it is not claimed later. RESCALE: NOT READY. One proven factor worth -0.0031 Brier. Rescaling on that is relabelling. Counter and frozen clusters byte-identical. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
7b85934dc3 |
Under-querying vs out of data: the answer depends on the unit
The platoon test's n=452 described how much of the JOIN survived, not how much data exists. There are 1,266 clean settled hits rows and zero quarantined ones. platoon_splits had been ingested from tonight's lineups only (315 players), so any hitter who settled a prop without appearing in an ingest-day lineup was silently absent from every test. Backfilled all 380 hitters (81 fetched, 0 unresolved). Re-ran on 1,059 rows, up from 452. THE DEMOTION IS THE HEADLINE. pitcher_contact_profile, the strongest proven factor in the programme (-0.0064, CI [-0.0113,-0.0014]), roughly halved to -0.0034 on more than double the sample and its corrected interval now spans zero. The Bonferroni denominator also rose to 55, which widens every interval -- but a denominator cannot move a point estimate, and that halved on its own. platoon and platoon_severity now clear the bar and are NOT promoted. Upper bound -0.0001, on season-to-date splits that contain the games they predict: measured contamination is 4.5% median, 12.4% at p90, 137% worst. I had assumed ~1%. They stay CANDIDATE pending point-in-time splits. GAME-LEVEL IS A DIFFERENT PROBLEM. game_context held zero weather rows ever -- not because the fetcher was wrong (it correctly targets Open-Meteo's archive) but because ledger_entries keys a game as mlb:2026-08-03:Away@Home and game_context keys it as mlb:823437. Every lookup missed and NULL columns read as honest absence. Third occurrence of that class. Fixed the join: 96/101 settled games now carry actual archived weather, park dimensions backfilled 15 -> 30 venues. But 928 total_bases rows sit on 47 games at 17.6 rows per game. Park and weather assign one value per game, so resampling rows would have manufactured a pass. factorGate now resamples clusters when rows carry one and judges sample against effective_n; unclustered rows keep the original path byte-for-byte. Verdict: 47 clusters < 500, and the point estimate is +0.0011 -- worse, not merely unproven. Weather needs ~57 more days. Park dimensions need never: there are 30 ballparks in MLB, so a venue-constant factor can never reach 500 independent units. That bar was built for player-level factors and does not transfer. Wind is refused. We have speed and bearing for all 96 games; we lack park orientation, and 220 degrees is blowing out at one park and in at another. Using speed alone would assert an effect while discarding the sign that decides what it is. Counter and frozen clusters untouched. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
de0077f6f9 |
Causally-correct platoon + park-dimensions ingest
Applying the method that worked for defence to the two factors the code flagged as still crude. PLATOON. The flat version is 'lefty versus righty, add a boost', and it failed the two-part gate for the same reason team-average defence did: it is not the unit the causal story runs through. The advantage is only worth what THIS hitter's split is actually worth -- measured on a real hitter, .284 against left-handed pitching versus .221 against right-handed, a 63-point split, where the flat factor applied the same six percent to him and to a hitter with none. Most of the work is sample discipline, and the second rule matters more than the first. Severity shrinks toward the league split weighted by the SMALLER side's plate appearances, because a 500-against-40 split is a 40-PA read. And below a floor it REFUSES outright rather than shrinking, because a heavily-shrunk severity is indistinguishable from a measured league-average one and those are different claims -- without the refusal the atom would quietly assert a league-typical split about every September call-up in the league. Switch hitters turn out to be the easy case misread as the hard one. He bats opposite by choice so the direction is never in doubt, but the per-side value of his swing is a different question and one this sample cannot answer, so he is unreadable rather than credited with an automatic edge. PARK DIMENSIONS. Free from statsapi's venue endpoint, which carries fence distances, roof, turf and elevation outright -- Wrigley returns 355 down the left line, 400 to centre, 353 to right, at 595 feet. parkFactors holds run COEFFICIENTS, which structurally cannot express a park that turns outs into hits without scoring, and that is why the crude park factor failed. The park join is by the venue the game is ACTUALLY at, carried from the schedule feed, never inferred from the home team -- neutral-site and international games break that assumption and they break it silently. A venue with no geometry at all is absent rather than a park with zero dimensions. Both tables dated in the primary key. Venue geometry changes rarely but it does change, and by now that is the default rather than a lesson. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
405180e791 |
Build the causally-correct defence atom: spray x positional OAA
Team-average defence failed the two-part gate for hits, and the reason was the unit rather than the signal. A left-handed pull-ground hitter meets the first baseman and the second baseman and almost nobody else, so a team total averages in five fielders who will never touch his ball. Both halves were already free on the host we pull from. Statcast publishes spray x trajectory per hitter -- pull/straight/oppo crossed with ground/air, 608 hitters -- and the OAA feed already carries each fielder's position, so per-position defence is a regrouping of data ingested last week rather than a new source. Zero new sourcing, as the order expected. Handedness is what joins them and getting it backwards would be invisible: pull for a right-handed hitter is the left side, pull for a left-handed hitter is the right side, so a model ignoring bats would send half the league's grounders to the wrong infielders and still look like it was reading defence. A switch hitter bats opposite the pitcher, which this does not resolve, so he is unreadable rather than guessed. Two properties the crude version could not express, both locked by test: two teams with the SAME total defence read differently for a pull hitter, and a ground-ball hitter and an air hitter read the same team in opposite directions. Unmeasured zones are renormalised away rather than contributing a zero, which would assert an exactly-average fielder standing there, and states honestly what share of a hitter's contact we could actually read. Nothing readable at all returns null, so the caller falls back to the base rate instead of to an invented 1.0 that looks measured. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
a9ee55550b |
Build the two-part factor gate: one factor proves, and zero are theatre
The question was whether the hit grade reads tonight's game or just says he is due. Answering it needed a gate that correlation cannot provide, because correlation cannot separate the two ways a factor looks alive: it reads the game, or it moves the number and reads nothing. The second is what a product ships by accident -- arch-v1 moved 76% of rows by 2.5 points, changed resolution by 0.0000, and was live for months, and no user could have told. So a factor must now clear both conditions: move the prediction off the player's own leave-one-out base rate, AND improve out-of-sample Brier. Brier rather than correlation, because correlation asks whether the ordering improved and this asks whether the NUMBER got closer to what happened -- and for a graded probability the number is the product. The correction applies to the interval itself, which turned out to matter more than expected. A plain 95% CI is the right bar for one test; at fifty cumulative tests roughly two or three intervals exclude zero by chance alone. Widening to 1 - 0.05/tests, currently 99.9%, flipped both defence and platoon out of "proves". A 95% interval would have shipped two unproven factors into the grade, with reasoning text explaining them to users. That forced a distinction I had initially collapsed. Defence and platoon have FAVOURABLE point estimates whose corrected intervals merely span zero, and calling that THEATER would repeat the error this codebase keeps correcting: insufficient evidence is not evidence of absence. THEATER is now reserved for its one real meaning -- moves the number, reads nothing -- and NOT_PROVEN_AT_CORRECTED_BAR names a real candidate held to a bar that rises with every hypothesis the programme tests. Result on 741 settled hits rows: pitcher_contact_profile PROVES, improving Brier by 0.0066 with a 99.9% interval of [-0.0114, -0.0016]. Defence (-0.0043) and platoon (-0.0039) are not proven at the corrected bar. Park is sample-blocked at n=405. Zero factors are theatre, which is the genuinely good news: nothing decorative is being wired. Per-archetype every slot is sample-blocked (BOMBER 252-294, GHOST 67-125). Two spec gaps worth recording. The approach identities the order names -- SPRAY, DAMAGE-DEALER, COUNT-WORKER -- do not exist in the registry; the MLB batter archetypes are BOMBER, GHOST, TORCH, BRUSH, DRIVER, FLEX, ALPHA, HYBRID and CATALYST. And parkFactors maps hits to run_base, so there is no hits-specific park factor at all: a park that turns outs into hits without producing runs is invisible to the input we have. The grade rescale is NOT run. It was explicitly gated on the factor proving, and one pooled factor worth 0.0066 of Brier is not a factor-informed distribution -- rescaling on it would dress a base-rate model as a matchup model, which is the exact thing this gate was built to prevent. 4,286 tests green (340 suites); web build exit 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
4d1803f6d7 |
Calibrate hits point-in-time: partial pass, and an honest ceiling of 0.667
Fitted the isotonic map on game_date < 2026-08-02 (n=589) and evaluated it on everything from that date forward (n=383). The map never saw the evaluation rows, which is the only thing that makes the result mean anything -- fitting and evaluating on the same rows always looks perfectly calibrated, because the map is reciting the answers it was built from. It works, on most of the distribution. Held-out after correction: 0.477 comes back 0.506, 0.587 comes back 0.580, 0.667 comes back 0.603 -- against raw errors of +0.191, +0.279 and +0.246 in the same bins. Ordering survived, and that was verified pairwise rather than assumed, because a broken map would silently destroy the one thing this model does well. Two findings matter more than the pass. First, the honest ceiling is 0.667. Once the numbers are truthful this model has no 80%-plus hit reads at all -- the top of its range was miscalibration, not confidence. A four-leg ticket at the ceiling is 0.198, where the raw numbers implied 0.686. The high-floor parlay is a two-thirds-per-leg proposition, and that is the number to say out loud. Second, calibration is certified BY BAND rather than by a blanket flag. Held-out error was -0.029 and +0.007 through the middle but -0.167 at the bottom and +0.063 at the top: the model is trustworthy over most of its mass and untrustworthy at both edges. A single true/false would either throw away the 72% that works or ship the edges that do not. Only a probability inside a certified band is marked stackable, and that flag is what chainAcross requires before it will compound anything. The certified band is 0.40 to 0.60, n=276. A methodological catch on the way: my first pass condition demanded honest bins at 0.70 and above -- but honest calibration REMOVES those bins, since the ceiling drops to 0.667. The gate would have failed the repair for succeeding. It now tests the highest remaining band instead of a fixed threshold. Wired forward with the same discipline: calibrationService fits strictly before today, splits by time rather than at random, and returns null on thin history so that "no calibrator" means nothing is stackable rather than "trust the raw numbers". p_win is never mutated -- the calibrated value rides beside it as p_win_calibrated, because a calibration map is a correction to a forecast, not a different forecast, and the counter stays byte-identical. 4,275 tests green (339 suites); web build exit 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
9c5b968351 |
chaining-v1: the portable chain, and the gate that blocks the parlay surface
The order's own prerequisite for the hit-parlay surface was to verify the hit probability is calibrated. It is not, and the failure is exactly the shape that destroys a parlay. Measured on 972 settled hits props: the model is monotonically over-confident at the top and flat above 0.70. Predicted 0.911 comes back 0.630. Predicted 0.844 comes back 0.630. Predicted 0.747 comes back 0.605. There is no discrimination at all in the range a parlay is built from, and the error runs in the flattering direction. Four "91%" legs are 0.686 by the model and 0.157 in fact -- a 4.4x overstatement that compounds with every leg added. Single props survive a calibration error of that size. A parlay multiplies it. So chainAcross REFUSES to compound atoms not marked calibrated, and refusing is the feature rather than a limitation: a ticket built on these numbers would be confidently wrong in the direction the user pays for. calibration.js provides the reliability table, the gate (tolerance 0.05, weighted to the high end because that is where tickets live) and an isotonic fit. Isotonic is the honest repair here because it is monotone: the model's ordering survives untouched while the numbers move to what actually happened. The fitted map says 0.65 -> 0.594, 0.85 -> 0.639, 0.91 -> 0.639. chain.js is the portable core -- base events plus context, through a chain function, into a PLUGGABLE aggregator: across players for a compound ticket, up to the team for expected scoring. The sport-specific parts are inputs rather than code paths, so basketball plugs in as content. The archetype redistribution hook is there now, dormant in baseball because a nine-run lead does not change who bats next, and live in basketball where a blowout fades the star and feeds the bench. Two judgement calls worth naming. Treating same-game legs as independent errs in the FLATTERING direction, since they share pitcher, park and weather -- so correlation shifts the compound toward the weakest leg, bounded, and is labelled an approximation rather than a joint distribution. And market divergence does NOT downgrade confidence: it flags a contested script whose props are either the best or the worst on the board, and which one is unknown until settled. Internal inconsistency does downgrade it, because per-entity reads failing to sum to the team read means one of them is wrong and we do not know which. Not built: the independent game-script projection. It needs proven team-level atoms and out-of-sample validation against actual margins, and no atom has passed the gate yet. Building it now would produce something plausible rather than something proven, which is the failure mode this whole programme exists to avoid. 4,269 tests green (339 suites); web build exit 0; counter and frozen clusters byte-identical. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
08276c0880 |
Ingest lineup + baserunner context: the input RBI and runs always needed
RBI is power TIMES opportunity. The same swing drives in one run or three depending on who is on base, and a hitter batting with the bases empty cannot drive anyone in however hard he hits it. Every context-free model of RBI here has failed, and the failure kept being read as 'skill inputs don't work for RBI' when the truth was that we were modelling half the stat. Both halves are free from statsapi.mlb.com, which we already call for game logs, schedules and probable pitchers. No new provider, no key, no quota. RUNG 1, batting order: schedule?hydrate=lineups returns homePlayers and awayPlayers as ORDERED arrays of nine, and the order IS the batting order -- index 0 is the leadoff hitter. That single fact gives CATALYST its identity and supplies lineup-position context for every context-dependent stat. RUNG 2 turned out cheap, which the cheapest-first rule did not expect. It looked like it would need play-by-play reconstruction across a season; statsapi serves situational splits directly, so 'how often does this hitter bat with runners to drive in' is ONE call per player rather than one per game. Measured on a real hitter: 87 plate appearances with runners in scoring position producing 25 RBI, against 302 with the bases empty producing 17. That ratio is the opportunity half of the stat and it is the thing no amount of exit velocity can tell you. Both tables are dated in the primary key. statcast_aggregates was built upsert-in-place and that silently made every backtest leak the games it was predicting; a lineup is worse still, because it is a PRE-GAME fact that changes by the hour, so an in-place table would overwrite what we knew at grade time with what turned out to be true. Absent stays absent throughout: no lineup posted is an empty slate rather than a guessed order, a short lineup records fewer slots rather than padding to nine, and a hitter with no splits is null rather than a zero RISP share -- which would assert he never bats with runners on, a strong claim and usually a false one. Wired into the snapshot best-effort, so a context failure can never break the pipeline it rides in. The three pre-registered theories are now marked input-ready rather than input-blocked: DRIVER's power x runners-on and power x lineup-position, and CATALYST's speed x on-base x power-behind. They are sample-blocked from here, and the proofs run under native cumulative correction as sample accumulates -- ingesting is not proving. Counter and frozen clusters byte-identical. 4,250 tests green (338 suites); web build exit 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
ff037e40c2 |
Re-adjudicate: nothing to demote, and close the hole that would have mattered
There is nothing to re-adjudicate. The proven set is empty and always has
been -- verified three ways: proven-status reports EMPTY, validatedSkills()
returns {} for every archetype, and zero conditioning entries have ever
reached PROVEN. The one PROVEN feature is recent_frequency_prior, which is the
incumbent counter itself, proven by the S78 ablation as ~100% of the
champion's resolution. It is the baseline every challenger is measured
against, not a conditioning interaction, and demoting it would leave the model
with nothing to grade from.
A correction to the premise: the cumulative gate did NOT catch a false
positive last session. It caught nothing, because there was nothing in the
proven set to catch. What it did was tighten alpha from 0.0026 to 0.0013
within one session, which demonstrated the mechanism working rather than a
demotion. So steps 3 and 4 -- demote, recalibrate -- are vacuous here, and
readjudicateAll says so plainly rather than glossing a no-op.
But the worry behind the order was well founded, and the audit found the real
exposure: promote() did not require the cumulative denominator. It checked n,
lift and CI, and nothing stopped a future session from testing eight
hypotheses, correcting by eight, and promoting on a p-value that would not
survive the programme's real denominator. That is precisely the hole that
makes a retroactive re-adjudication pass necessary later, so it is closed at
promotion time instead. isSufficient now refuses evidence carrying no
correction, evidence corrected against fewer tests than the cumulative count,
and any p-value that does not clear 0.05 over its own test count. The same
rule guards a PROVEN conditioning entry.
The second audit found two of four analysis scripts still correcting
per-session; pitcher-prove-k and tb-solo-and-interactions now use the
cumulative ledger, so the correction is native on every path.
reAblation.js is the standing second line: pure and injectable, so the
decision rule cannot drift from the gate's, and every verdict records both
p-values and both test counts so a demotion is re-derivable by anyone. A
feature promoted at alpha 0.05/20 can demote on the same p-value once the bar
is 0.05/60 -- correct, because the bar rose only after the programme had more
chances to get lucky. No fresh measurement is PENDING_RETEST and never a
demotion: absence of a re-test is not evidence, and demoting on it would
punish whichever stat happens to be off-season.
Net effect on the proven set is zero. No demotions, no recalibrations, and no
public ledger event -- announcing "recalibrated after re-adjudication" when
nothing changed would itself be a false signal of rigour.
4,238 tests green (337 suites); web build exit 0; counter byte-identical.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9
|
||
|
builtbykev
|
ece2b9f5f9 |
Ingest defence, and make Bonferroni cumulative across the programme
Two things shipped that stand regardless of sample. DEFENCE. Statcast Outs Above Average is free on the host we already pull six feeds from, so there was nothing to decide. 514 fielders, aggregated to team level -- the unit a batter's prop actually needs, the defence behind the pitcher he faces -- and persisted as 31 team rows. Verified in production. Cubs +56 best, Mariners -29 worst. Unknown is not zero, and it bites unusually hard here: an OAA of 0 is a REAL reading meaning exactly average, so coercing absence to 0 would assert that every unmeasured fielder is league-average, which is the commonest defensive profile there is. team_defense also carries as_of_date in its primary key from the first row -- statcast_aggregates was built upsert-in-place and that silently made every backtest leak the games it predicted, so point-in-time is available here before it is needed rather than after a wrong answer. A bug worth recording as a class: BASE already ends in /leaderboard, so the new feed built a doubled path and 404'd. Because a failing feed degrades to an empty index by design -- correct, so one broken source cannot fail the whole pull -- it surfaced as "fielding_oaa: 0 rows", which reads exactly like "Statcast has no fielding data". Graceful degradation makes a wiring bug look like an honest absence. CUMULATIVE CORRECTION. Bonferroni had been applied per session throughout: a run testing eight features corrected by eight. Across a programme's lifetime that is wrong in the dangerous direction, because every order gets a fresh generous alpha and the false-positive rate compounds quietly. Correcting by 8 when sixty have been tried is how a noise result eventually gets recorded as PROVEN with a p-value to point at. The denominator is now distinct hypotheses ever tested, persisted, and it moved 19 -> 38 within this session alone, alpha 0.0026 -> 0.0013. Re-tests deliberately do not inflate it: re-asking the same question on more data is not a new shot on goal, and counting it would punish the discipline of waiting for sample. THE MEASUREMENT. The differential the theory predicted is present: defence correlates with the counter's residual at +0.130 for GHOST, the contact and speed archetype, and -0.018 for BOMBER, the power archetype. A GHOST's hits depend on whether anyone can range to the ball; a BOMBER's barrels clear the defence entirely. So a flat BOMBER result is the theory working rather than the test failing. It is not a result. GHOST is n=104 against a 500 bar, with p=0.188 against a corrected alpha of 0.0013 -- three orders of magnitude short. Both are recorded as CANDIDATE with their measured lift, tagged contact-skill, so the re-run at full sample compares against a recorded baseline. Nothing proved, so nothing was recalibrated and nothing shipped. 4,228 tests green (336 suites); web build exit 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
843c8c6d4b |
Build the pitcher engine, and find the cap was eating the whole board
Strikeouts are NOT proven -- n=57 against a bar of 500. But the finding that matters is not a correlation. THE CAP. Measured on the live slate via the refusal diagnostic: 1,244 unique gradeable props exist, the 500 cap graded about 334, and because dedupeProps takes first-row-wins in FEED ORDER, what survives is decided by feed position rather than value. Pitchers are 2.6% of a batter-dominated feed, so we were grading SIX strikeout props a slate against 32 available -- putting n>=500 three months away for every pitcher stat. Pitcher props were never being refused (graded 5, refused 0, suppressed 0); it was truncation. Raised 500 -> 1500 on measured cost: 721ms per prop at concurrency 5 is about 179 seconds for the full board, against a cron that runs five times a day and a fire-and-forget caller that never holds an HTTP response. statsapi is free and unlimited. Concurrency stays at 5 -- one variable at a time. This unblocks every n-blocked stat in the programme, not just pitchers. THE ENGINE. pitcherEngine.js is its own engine, not the batter engine pointed at pitchers: the batter model asks whether contact becomes a hit and reads contact quality, the pitcher model asks whether the plate appearance ends without contact at all and reads stuff. Archetypes are FLAME (whiff-led), SCALPEL (chase-led), SINKER (pitches to contact) and DEFAULT, and a test asserts the weight keys are not the batter engine's. The projection is K% by log5 against THIS lineup, times batters faced, through a binomial. An unclassifiable arm gets the balanced map, never a guessed archetype. THE MEASUREMENT, at n=57 and contaminated. Four solo features clear the 0.15 effect bar and fail only on sample: arm angle at -0.250 -- the largest correlation measured anywhere in this programme -- then whiff +0.213, k rate +0.206, chase +0.195. The batter cluster's best was 0.135. Head to head, pitch-v1 resolves 0.1285 against the counter's -0.0639, delta +0.192 with a CI spanning zero. That negative is the interesting number. The counter is ANTI-PREDICTIVE on strikeouts: counting a pitcher's recent Ks is worse than useless, because his recent totals track which lineups he drew and how long he was left in rather than his skill. It is the one stat where the incumbent has no defensible edge. A bug caught on the way. resolveTeam wants an abbreviation and the game log supplies full team names, so the roster join silently resolved nothing and the first run reported 0% lineup coverage -- the theorized stuff x lineup carrier was never being tested, not failing. Fixed; coverage is now 94.7%. The carrier still shows no incremental signal over whiff alone, and adding the lineup term lowered head-to-head resolution, which is recorded rather than dropped. Calibration was not reached: nothing passed the first bar. The batter model and the counter are byte-identical, verified by diff. 4,221 tests green (335 suites); web build exit 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
4aab18096f |
Prove both on total bases -- and find that my own fix destroyed the backtest
Nothing passed. Nothing promoted. Counter byte-identical. THE BLOCKER, which is the real finding. statcast_aggregates is upserted in place and holds exactly one as-of date. Yesterday's skill backtest was honest only by accident: the nightly refresh was unreachable code, so the profiles sat frozen at 2026-07-21 -- before the settled window. Repairing that cron was right for production and it refreshed them to today, destroying every prior version. Scoring a 2026-07-25 game now uses a season aggregate that contains that game. Point-in-time validation is structurally impossible from that table, so every number in this run is contaminated and directional, and none of it is a gate verdict. Fixed forward: statcast_history retains a dated snapshot on every refresh, so point-in-time becomes "as_of_date < game_date, most recent". Retention is best-effort and cannot fail the refresh; both properties are unit-tested. It has one day of data, which is not yet a window. SOLO BASELINE, n=383, Bonferroni across 12 tests (alpha 0.00417): nothing passes. hard_hit_pct is closest at marginal r 0.135 with p 0.0080, failing both the 0.15 effect bar and the corrected alpha. And it drifted DOWN from 0.153 at n=295 -- an estimate regressing as noise averages out, not an effect firming up. I called that number encouraging yesterday; on 88 more rows it is fading, and it should not keep being quoted at its best value. INTERACTIONS, each scored by partial correlation against the counter residual controlling for both of its own components: none pass. Only barrel x power archetype has an incremental exceeding its parts (-0.101 against 0.019) at n=260 -- the shape Discipline 2 predicts, but a lead, not a finding. A methodological catch worth keeping. The archetype conditioner was first built as barrel_pct over league barrel -- a monotone transform of one of its own components -- so the "interaction" was barrel squared, measuring nonlinearity in barrel rate rather than any archetype effect, and it produced this run's only positive result. A Gauss-Jordan pivot test does not catch that, because the two columns differ by a scale factor. Fixed with a scale-free collinearity check plus real archetype labels joined from model_snapshots. Without it this document would have reported a fabricated interaction as the session's finding. COMBINED vs COUNTER on total bases: 0.2718 against 0.2647, delta +0.0071, CI [-0.065, +0.079] -- inconclusive, and the first time a challenger has not lost. The same engine on hits was -0.116 with a CI excluding zero. That contrast is the whole argument for total bases, and it is what the physics said: contact quality governs extra bases, not whether a grounder finds a hole. Also built: the compound TB projection. skillProjection no longer refuses total bases -- a deterministic bases-per-hit multiplier had made P(TB>=2) exactly P(hits>=1), a relabelled hits curve. It is now a convolution over per-PA base outcomes with hit-type shares shifted by skill. Non-degeneracy is locked by test. 4,204 tests green (334 suites); web build exit 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
c7cc8f5e52 |
Build the gate, run it, and find we were proving things on the wrong stat
PREMISE CORRECTION FIRST. statModel.js and correlateValidator.js do not exist in this repository. The validation spec's only prior form is src/services/python/blueprints/unconventional.py -- a Flask blueprint in the Python service that is offline in production, scoring NBA factors against a warehouse that was never populated -- and tests/unit/supplementSystems.test.js requires only fs and path while defining its own validateFactor inline at line 368. Those tests assert a re-implementation of the thresholds, not an implementation, which is exactly why they passed for months while nothing was connected. The diagnosis behind the order is right -- every challenger was measured without a gate -- but the cause is that there was no gate on the Node side to import. So it is built, to the exact spec. correlateValidator: n>=500, |r|>=0.15, p<0.05, Bonferroni across the sweep. The p-value is exact rather than approximated (t-transform through a regularized incomplete beta) and is verified in the suite against known values, because scipy is not available here. Pairs with an unknown side are dropped, never zero-filled -- a zero-fill inside a correlation does not add noise, it invents a point at the origin. THE RUN, hits, n=570, Bonferroni-8: every skill feature fails, and not narrowly. The strongest marginal correlation against the counter's residual is 0.062 against a 0.15 bar. That is an effect-size failure at a sample that would have found a real effect comfortably -- a clean, well-powered negative. The head-to-head agrees: value engine 0.0499 against the counter's 0.166, delta -0.116 with CI [-0.189, -0.043]. Not promoted. THE RUN, total bases, n=295: cannot be tested, and that is the finding. hard_hit_pct shows a marginal r of 0.153 -- above the threshold -- and exit velo 0.124, refused solely because n is 205 short of 500. It is the most encouraging number this work has produced, and it is what the physics predicts: contact quality governs extra bases, not whether a grounder finds a hole. We have been testing skill inputs on the one stat where they should not matter much. Two things the run forced. Feature verdicts are now PER STAT, because marking these DEAD sport-wide on hits evidence would have killed, for total bases, the features that look most alive there -- per-sport doctrine one level deeper. And the gate now reports r and p even when underpowered, because "not enough data yet" and "nothing here" demand opposite decisions and a bare refusal was hiding the best signal on the board. Next: build the compound TB projection (skillProjection still refuses total bases by design, since a deterministic bases-per-hit made P(TB>=2) identical to P(hits>=1)), accrue to n>=500, re-run this gate. Leave hits alone. 4,200 tests green (334 suites); web build exit 0; counter byte-identical. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
258d8a6655 |
The skill engine: built, gated by construction, and Stage A honestly lost
Built src/services/model/ -- the forward, archetype-selected, skill-based projection, as a challenger. The champion is untouched. featureRegistry makes "earn its place or it's out" structural rather than aspirational: CANDIDATE / PROVEN / DEAD per feature per sport, liveFeatures() returns PROVEN only, promotion requires n>=200 with positive lift and a CI excluding zero, and there is deliberately no override argument. It ships with exactly ONE proven feature -- the incumbent counter, because it is the only one with a measurement. A test asserts that with only PROVEN features allowed the projection returns null, so an unproven model cannot reach a user by accident. The three champion adjustment layers are registered DEAD with their reasons so they cannot be silently rebuilt. skillProjection is a PA outcome tree: K and BB combined by log5 odds-ratio against league (both identities unit-tested), then archetype-weighted contact quality against contact allowed, then Binomial(PA, p_hit) mixed over a PA distribution. Archetype is a FEATURE SELECTOR, not a nudge -- BOMBER reads barrels at 0.50 and ground-ball speed at 0.00, GHOST inverts it -- and a test locks that the same hitter read two ways moves more than 0.15. STAGE A: IT LOSES. Out-of-sample on 570 settled hits props with 91.9% opposing-pitcher coverage, resolution 0.0499 against the champion's 0.166, delta -0.116 with CI [-0.189, -0.043]. It is not selective either: its eight most confident picks hit 50%, a lift of -0.065. Not promoted. The gate did its job on its first real test, which is the point of having built it that way. Two false starts, both recorded because they nearly produced a wrong verdict: statcast_aggregates stores PERCENTAGES, so raw rows made bip = 1-29.6-17.1 and refused 568 of 576 -- the honest-absent guards made a units bug loud instead of silent, and the conversion now lives at one chokepoint. And the first run resolved an opposing pitcher for 1 of 570 rows, because ledger team/opponent are NULL, so it would have reported "skill-v1 loses" while measuring a batter-only model with no matchup in it at all. The verdict above is from the corrected run. The loss is real but partial: park was passed as 1.0, handedness and opportunity_drift never fired, PA is season-PA over a constant, and the skill profiles carry no recency at all while the champion has a last-5 term. Also fixed: the Statcast nightly refresh was unreachable code. It sat inside tick() below "if (!HOURS_UTC.includes(h)) return" while testing h === 11, so it had never run once; the aggregates were 13 days stale and both of its alerts were in the same dead branch. It now runs on its own tick, and the test that passed happily throughout -- it only checked the string existed -- is replaced by one that asserts it is not behind the guard. 4,182 tests green (333 suites); web build exit 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9 |
||
|
builtbykev
|
b06a84af80 |
Settlement has been dead since 2026-08-01: a 500-id filter overflowed the URL
The self-learning loop stopped two days ago and reported success the whole
time. 1,444 ledger rows from 2026-08-01 sit unsettled with settle_attempts=0
-- never even attempted -- and every accruing challenger has been starved of
settled sample as a result.
ROOT CAUSE. settleLedger fetched open ids, then REFETCHED the full rows with
.in('id', ids). PostgREST puts filters in the URL, so 500 UUIDs became an
18,499-character request that the fetch layer rejects with "TypeError: fetch
failed". The result was destructured as `const { data: rows } = ...` with NO
error binding, so rows came back null, the loop body never executed, and the
function returned {settled:0, voided:0, unrecoverable:0, pending:0} --
byte-identical to a clean "nothing to settle". Reproduced against prod before
changing anything.
WHY IT HID FOR TWO DAYS. It is volume-triggered. Daily volume ran 20-260 rows
and settled perfectly for weeks; 2026-08-01 was the first day past the 500-row
fetch limit. And the zero-settle ops alarm reads these very return values, so
pending:0 told the watchdog the backlog was empty -- the alarm built to catch
exactly this could not see it.
THE FIX. The refetch existed only to add game_date/settle_attempts/
dclv_computed_at. Selecting them in the first query removes the id list
entirely, so there is no URL to overflow at any volume. A failed fetch now
surfaces its error instead of being reported as an empty backlog.
captureClosing carried the same shape one level down -- .in('id', g.ids) on an
UPDATE, which fails identically once a single line|odds group gets large on a
big slate. Its id filters are now chunked at 100 (~3.7 KB).
Tests: the regression is locked by asserting settlement issues NO id-list
filter at 500 rows, and that a failed fetch is never reported as an empty
backlog -- the two properties that would have caught this. Two existing
suites asserted the old two-query shape and were updated to the real one.
4,159 tests green (332 suites); web build exit 0.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9
|