PREMISE CORRECTION FIRST, because it defines the bar. total_bases has not
passed BAR 1. Its head-to-head is inconclusive at parity -- delta +0.004 to
+0.007 with a CI spanning zero -- and it is contaminated, and no feature of
its passed the gate. It was described last session as the first challenger
that did not LOSE, which is not the same as proven. If it is installed as the
frozen proven reference and every other stat is held to "the identical bar
total_bases cleared", the bar becomes "be inconclusive at parity" and the
whole cluster passes on a null result. The proven set is EMPTY.
HITS IS NOW A FINAL ANSWER. At n=803 it clears the gate's sample requirement,
so its features were properly TESTED rather than refused: every one fails on
effect size (max marginal |r| 0.053 against a 0.15 bar), every interaction's
incremental contribution collapses to about zero, and the model loses
head-to-head by 0.096 with a CI excluding zero. That is a well-powered
negative and hits should be closed rather than retried.
The rest are n-blocked: total_bases 383, rbi 391, home_runs 228, runs 188,
against a bar of 500. Two leads are worth carrying. home_runs barrel rate has
a marginal r of -0.135, and the sign matters -- higher barrel rate goes with
the counter OVER-predicting, which would be a correction rather than a new
predictor. And runs batterK x pitcherK has the largest incremental in the
cluster at +0.132, with a clean mechanism: strikeouts destroy plate
appearances, and a PA that never happens cannot score.
RBI deserves a caveat rather than a verdict. It is power times OPPORTUNITY,
and we ingest no baserunner state at all, so half its mechanism is missing. A
weak RBI result is evidence that we are modelling half the stat.
total_bases was held frozen: git diff on skillProjection against the prior
commit is empty. The counter is untouched.
Also fixed and verified in production: the point-in-time retention shipped
after yesterday's refresh had already run, so statcast_history was empty, and
its first run then failed on a hand-enumerated schema that had already drifted
from its source ("could not find the 'swing_pct' column"). The refresh itself
still succeeded and wrote all 1,387 aggregate rows, which confirmed the
best-effort guard in prod. The table now mirrors the source via LIKE and the
writer passes rows through whole. Verified live: 1,387 rows retained at as_of
2026-08-03. A usable point-in-time window starts 2026-08-04.
Stage B has nothing to calibrate. Everything now waits on a point-in-time
window and on sample -- both waiting problems, not building problems.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9
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
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
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
READ-ONLY. src/ and web/ untouched.
Inventoried every forward-model component against the real objective -- a
forward matchup projection, not an edge number. The finding is that all of it
already exists and is already loaded in production, and 100% of it sits
DOWNSTREAM of the grade in challenger columns nothing serves. The served p_win
reads three features and a game log; it has never seen a pitcher.
Inputs are HAVE, not missing: statcast_aggregates carries 1,354 rows (750
pitchers, 604 batters) with exit velo, launch angle, barrel, hard-hit, whiff,
chase, pitch mix, GB/FB, arm angle, and handedness complete on every row. Real
gaps are team defense and catcher/umpire. So Stage A is a plumbing-and-
modelling job, not a data-acquisition job.
Found along the way: the Statcast nightly refresh is unreachable code. tick()
returns for any hour not in HOURS_UTC (14,19,22,1,3) and the refresh block
then tests h === 11, which that guard can never admit. The mechanism data has
been frozen at its 2026-07-21 backfill for 13 days, and the block's own
failure alert sits in the same dead branch -- the identical silently-guarded-
out shape as the settlement outage.
Design shows the counter: every factor label the SIGNAL BREAKDOWN renders is a
restatement of recent frequency (l5_hot_vs_line, l20_over_line, back_to_back,
home_game) plus several structurally-NBA labels (referees, coach pace,
starters out) inside a baseball product. Not one names a pitcher, pitch type,
handedness or park. The card's forward-read slots already exist and go
unfilled -- the surface needs feeding, not redesigning.
On what changes: the prior measurements were outcome-accuracy, not edge, so
the metric was right and the question was narrow. proj-v1.1 and hits-v1 stay
correctly refuted as DISTRIBUTION swaps on thin inputs -- neither tested a
matchup-fed projection. arch-v1 is a market-relative nudge by construction and
is the one component genuinely measured on the wrong axis. AT CEILING is
provisional: measured only against features the champion already reads.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9
READ-ONLY. src/ and web/ untouched; 4,159 tests still green.
WHAT THE CHAMPION IS. probabilityEstimator is five lines of arithmetic: the
empirical frequency of (stat > THIS line) over the game log, blended 0.6/0.4
with the last-5 frequency, then +/-0.03 opponent, +/-0.015 home/away, a
cv>0.40 pull toward 0.50, and a clamp to [0.10, 0.95]. It reads three
features. featureCache retains a dozen more that p_win never touches.
THE ABLATION IS EXACT, NOT A REFIT. Every adjustment is closed-form from
stored features and the consistency step is linear, so each layer subtracts
algebraically out of the stored p_win -- no re-estimation, no re-fetch, no
lookahead possible. Per stat, paired bootstrap:
removing ALL THREE adjustments changes resolution by NOTHING on every stat
hits -0.0059 total_bases -0.0015 rbi +0.0106 runs +0.0130 walks +0.0008
and rbi's home/away is mildly HARMFUL (+0.0053, CI excludes zero). So ~100% of
the champion's resolution is base+recency: how often this player has cleared
this number lately. Everything else is decoration.
A CORRECTION. Pooled, the champion resolves 0.46; per stat it is 0.196 (hits)
to 0.499 (rbi). Pooling stats with different base rates inflates correlation,
so 0.46 should not be quoted as the champion's resolution. Last session's
paired differences remain valid; only the absolute level was inflated.
THE BIGGEST LOSS IS NOT A MISSING FEATURE -- IT IS THE CLAMP. 358 of 1,741
settled rows (20.6%) sit on the boundary, so the model emits a constant there
and cannot rank a fifth of the book at all. And that constant hides two
opposite failures: 0.900 covers home_runs-under truly winning 99.5% (9.5pts
under-confident) next to hits-under truly winning 51.9% (38.1pts over-
confident). PROB_CEIL=0.95 makes the 99.5% case inexpressible. Global
over-prediction is +3.5pts, +7.6 on total_bases. None of this needs new data.
ONE REAL MISSING-WEIGHTING LEAD: opportunity_drift, residual corr +0.156 on
hits and +0.145 on total_bases -- it REPEATS across independent stats, unlike
the weather hits on TB which sit inside the expected false-positive count (70
tests at alpha .05 expects 3-4). And we already compute it: arch-v1's
opportunity axis uses it and extracts nothing (delta +0.0001). Wrong
implementation, not a missing feature -- opportunity must scale the rate, not
nudge the probability.
ARCHETYPE IS UNMEASURABLE, NOT REFUTED. Only 2 of 41 labels (BOMBER, GHOST)
reach n>=40 settled rows and every mean residual straddles zero. That is "we
have not measured it", and it does not license acting in either direction.
Why every challenger has failed is now legible: the ladder and hits-v1 REPLACE
the frequency question with a fitted distribution; the environment axis adds
inputs the champion ignores. Asking the frequency question at the traded line
is the thing that works.
Flagged, not fixed: model_snapshots.outcome is NULL on all 22,032 rows -- the
retention table built for exactly this replay was never settled, so labels had
to be joined from ledger_entries.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01W1sivYNqY2TS5ftykmHBU9
PROMOTE-THE-EARNED. Nothing was promoted, because nothing earned it -- not
because the bar was held high. Measured on the same bar that refuted hits-v1:
own rows only, direction-aligned, paired bootstrap, promote only on a CI
excluding zero.
arch-v1 n=1741 delta 0.0000 CI[-0.0050,+0.0054] inconclusive
contact-v1 n=1055 delta +0.0008 CI[-0.0052,+0.0069] inconclusive
proj-v1.1 n=1664 delta -0.0301 CI[-0.0543,-0.0060] reliably WORSE
matchup/tb-v1/hits-v1 n=0 genuinely pending (rows dated 08-02+)
arch-v1 is the interesting one: it MOVED 76% of rows by 2.5 points on average
and resolution is identical to the champion to four decimals, on the moved
rows too. That is active movement carrying no information -- a finding, not a
pending verdict.
These are true prospective holdouts: arch-v1 and contact-v1 wrote p_win at
grade time into their own columns before the game. Nothing recomputed.
THE 429, read-only. The premise was that we re-pull the full picture every
slot and blow the quota. Measured: PropLine is at 5 calls of 3,000/day --
0.17%. One snapshot is ONE PropLine call per sport, all markets comma-joined.
There is no request-pattern problem, so a change-based pull cannot fix it and
no tier upgrade is needed.
The 429 is odds-api: 478/500 MONTHLY, blocked at 95%. oddsService falls
through silently when PropLine returns empty, and the backup's quota gate
throws the error -- so an empty slate is indistinguishable from an outage and
the message names the wrong provider. Flagged for its own order.
Could NOT verify PropLine movement endpoints: docs are auth-gated and the keys
are production-only. Not asserted either way. The movement-as-data argument
stands on its own merits and should be justified that way, not as a quota fix
it isn't.
Book-breadth invariant written down: we never discard books. All are kept and
shown (DISPLAY_BOOKS = MODEL + REFERENCE + DFS); DFS pick'em is excluded from
PRICING only, because a fixed-payout shaded number is not a market price.
Verified this is already what bookRoles.js does.
Champion byte-identical; every challenger stays wired.
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
Hits was diagnosed as a family mismatch: 84% of hits rows trade at 0.5, so
the stat rides on P(0), and a negative binomial has unbounded support and no
notion of opportunity at all. hits-v1 models it as the bounded conversion it
is -- N ~ the player's empirical at-bat distribution, hits|N ~ Binomial(N,q),
with the multiplier scaling q (conversion) and never N (opportunity).
STEP 0 confirmed the inputs before the model existed: 30/30 real ledger
players, 100% combined-input coverage. Every read goes through knownRate --
a row with no atBats is dropped, never counted as a 0-at-bat game.
It FIRES: 158/159 hits props (99.4%) on the live production snapshot, through
the real attachProjection path. Scoping by book IDENTITY rather than price
shape kept 94 out-of-promotion-band props on the board, 93 of them modelled --
59% that a price rule would have deleted.
And it LOST. Point-in-time replay (game log truncated strictly before each
row's game_date, real grade-time multiplier), hits-only, direction-aligned,
n=242: resolution champion 0.195 / ladder 0.048 / hits-v1 0.026. Paired
bootstrap on the same rows: hits-v1 - ladder = -0.022, CI95 excluding zero.
Not promoted.
The value is in what it eliminates. The family was wrong AND the mean was not
the constraint -- hits-v1 moved the line-0.5 mean 0.554 -> 0.581 toward a
0.598 base rate while resolution fell. What is left is per-prop
discrimination: the ladder's inputs, not its distribution.
The pre-registered fallback is recorded as WRONG rather than deleted. It said
hits might be genuinely low-resolution for anyone; the champion scores 0.276
on the identical 189 rows, so there is real signal and the ceiling claim was
the comfortable reading, not the honest one. Its own control refuted it, and
that control was already in hand when the branch was written.
hits-v1 stays wired as a challenger writing its own ledger columns so the
forward accrual can confirm the backtest. Champion, ladder, ranking,
calibration, reference ruler and the four accruing verdicts are byte-identical
-- the diff has zero deleted lines.
Tests 4,156 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
MLB statsapi + WNBA ESPN live boxscores -> per-player current values
(live:{sport}:{date} TTL 90s, /api/live/:sport + Next proxy). Pure
propState math (HIT / ON PACE / NEEDS N / HOLDS / LINE PASSED — never
red in-progress), attachLiveProgress strip join on nameKey+statType,
proximity-to-hit slate float, StatStrip LiveTracker in the ROW-GRAMMAR
outcome slot (spec amended + lock test updated). Grades never change
in-game — tracking, labeled as such. 2698 -> 2757 tests, web build 0.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- specs/ROW-GRAMMAR.md: the row grammar law (slot order, color law, mark
law, mobile stacking, no-truncation), locked by tests/unit/rowGrammar.
StatStrip violations fixed: MovementChip before the grade (market
context before model output); ViabilityChips after the archetype
(identity is one contiguous run).
- Line-movement sparklines: intradayRefreshService.trackHistory captures
real {t,line} points per grade (seeded with the lock, deduped when
flat, capped 24) inside the snapshot write-back; StatStrip.LineSparkline
renders at >=3 points (green toward / amber against / dim flat).
- Last-10 dot strips: services/last10Dots (streaksService accessors) ->
/api/snapshot/:sport attaches last10_dots from rosterlogs:{sport};
StatStrip.DotStrip renders vs the LOCKED line, newest first.
- CLV distribution: getModelAggregate emits clv_distribution (7 signed
buckets, outliers clamped) only past the centralized n>=20 gate;
ledger MODEL header renders the bar strip.
- Global search: SearchModal (cmd-K via GlobalHosts + window.__search),
players via /api/players/search per sport + static lib/teams.js
(soccer deliberately absent); Nav search icon + Search first in the
mobile More sheet. Explore tab untouched.
- Landing LCP: fade-up floored at opacity .6 (hero h1 contentful on
first frame, S33 visible-floor rule) + IBM Plex Mono preload:false
(4 decorative font files off the slow-4G critical path).
2654 -> 2698 tests (226 suites) green; web build exit 0.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
tesseract.js (self-hosted WASM, Apache-2.0) + pure per-book layout
parsers (DK/FD/MGM/Caesars) with per-field confidence and needs_review
honesty — the reader never guesses. POST /api/slips/parse (auth, free
1/day paid 10/day, 4MB cap) + Next proxy. Gated /slip page: upload or
paste, manual-correct UI, per-leg grades through the normal engine
(refusals render honestly), add-all to Parlay Lab, share card. Vision
model upgrade logged post-revenue. 2574 -> 2608 tests.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Email capture + daily report assembly + operator-triggered Listmonk send.
- NewsletterCapture (dark terminal, mono) on landing (below FAQ) + /welcome
(the real signup success surface); double-opt-in note; 'Signups open soon'
when Listmonk env is unset.
- POST /api/newsletter/subscribe: public, 10/min IP limit, honeypot,
server-side email validation, forwards to Listmonk subscribers API with
preconfirm_subscriptions:false (Listmonk sends the confirmation).
No env -> calm 200 { ok:false, reason:'not configured' }. Next proxy
web/src/app/api/newsletter/subscribe/route.ts (S25 rule).
- newsletterService.buildDailyReport: signals from snapshot:{sport}:latest,
STREAK WATCH via rosterLogs -> streaksService -> streakLens, THE RECORD via
ledgerService.getModelAggregate (percentage only when hit_pct != null —
n>=20 gate — else 'RECORD BUILDING · N pending'). RG footer (21+,
1-800-GAMBLER, Listmonk-native {{ UnsubscribeURL }}) in html + text.
VOICE v1.1 lint locked by tests: no '!', no banned vocabulary, numbers
only from injected pipeline data.
- sendDailyReport: creates + starts a Listmonk campaign; env-gated no-op;
refuses an empty report. Deliberately UNSCHEDULED — only
POST /api/internal/newsletter/send (internal key) triggers it.
- docs/NEWSLETTER.md: box-side Listmonk runbook (install, double-opt-in
list, API user, Coolify env, test-send).
- Spec: specs/feature-a1-s7-newsletter.md.
Tests 2398 -> 2429 (207 suites, all green); next build exit 0.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- QA.20: one meaning per color, locked — STEAM amber (caution), VALUE
green (edge), red only at STALE/miss; revisions render the original
grade struck through on slate AND ledger.
- QA.21: every new data surface (PriorReads, PlayerStreaks, ModelRecord)
renders data in mono (tabular-nums via the .mono class).
- QA.22: waiting copy derives from the real cron schedule on every
surface (StatStrip, dashboard, Explore) — no passive mystery anywhere.
- Explore empty state joined the honest-waiting lib.
- Sweep result: no raw brand hexes and no sans-serif data in tonight's
components (sans only on prose/captions/hero letter — sanctioned);
glow confined to A-tier badges + reveal chrome; 390px containment
holds (wrapping chips, scroll-contained tables from Phase 3 rules).
Backend 2352 -> 2396 tests (205 suites), web build exit 0.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Overnight sprint for the Saturday 10 AM ET deploy gate — day one of the
public ledger record locks against freshly posted lines.
Task A — ledger team/opponent (migration 020, applied at 0 rows):
populated in both write paths from the real feed; opponent only when the
player's team matches a game participant (never guessed). Roadmap: Phase
4.5 WNBA ESPN-boxscore settlement (due ~Jul 24) + Phase 5 per-tier
calibration logged.
Task B — work-order 1.6 CLOSED (canonical player keys):
- searchPlayer resolves via nameKey; the old matcher deleted accents
("Sanchez" with acute -> "snchez") and substring-guessed onto the WRONG
player (the mismatched last-10 bug). Ambiguous -> null, never guess.
- Slate JOIN INVARIANT: a graded prop whose player's real team isn't in
the game is dropped (TB player can't render under MIL@PIT) — locked by
tests that fail the suite on regression.
- grades:{sport} TTL 2h -> 6h (expired between 5h cron gaps — the real
cause of /team "No active props" for slate players).
Task C — Phase 2 slate UX: tabs are THE filter (URL ?sport=, deep-linkable,
duplicate legacy tablist removed); cards cap at 6 graded props sorted
A+->F with ALL N READS in-place expander; waiting states show the real
next pipeline run ("Grades post ~6:00 PM ET").
Task D — Phase 3 mobile P0: root cause of vanished 390px nav was HIDE_ON
including '/' (landing had zero navigation) — fixed; html/body overflow-x
contained; GAME LINES collapses to best-line summary + "N BOOKS" expander
below 640px; venue drops before time/pitchers ever truncate.
Live verification: raw ESPN today STILL returns the Jun 13 NYK@SA Finals
game without a date pin; the pinned fetch returns 0 games, 0 off-date.
Backend 2327 -> 2352 tests (202 suites), web build exit 0.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
ledger_entries is live (migration 019 applied to prod, RLS + NULLS NOT
DISTINCT dedupe verified against the real database). Every grade now
persists, settles against the real result, and carries closing-line value.
- ledgerService: pipeline pre-grade upserts (public model record, user_id
null, idempotent), closing capture on every snapshot (last write before
game start = the close), settlement with SIGNED CLV (over = locked -
closing; beat/faded/flat), 30d model aggregate with the hard n>=20 rule.
- Write paths: snapshotService -> ledger (priority path); Next /api/scan ->
ledger for authenticated users only (anon never touches the public
record). Refused reads write nothing and don't burn a scan.
- Honest refusal (work-order 1.5): no projection => insufficient_data,
grade null, "INSUFFICIENT DATA - no read" UI. The web gradeAdapter no
longer displays the line as the model projection (the audit's
model==line / +0% edge degenerate); the card renders absent states.
projectionFor is sport-aware (l5 -> l20 -> {stat}_per_90 -> xG).
- /ledger: MY READS | MODEL tabs; model header shows hit% + beat-close%
only at n>=20, else RECORD BUILDING + live pending count. ModelRecord
deferred-render strip on landing + player hero. CLV + outcome chips,
revised_from_grade strikethrough (Phase 2.5 ready).
- SYNC (Task 5): thresholds vs SNAPSHOT_EXPECTED_INTERVAL (normal <1.5x,
amber >=1.5x, STALE red >=3x) via /api/snapshot/summary.
- Phase 2.5 logged in specs/vyndr-roadmap.md (build after Phase 3).
- Data-semantics hardening: strict null-safe numeric parsing everywhere a
market value is handled (Number(null)===0 would have fabricated lines).
Backend 2309 -> 2327 tests (201 suites), web build exit 0.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Work-order Phase 0 (Jul 10 live audit): every fabricated UI element deleted
or rewired to real data. Deletion sprint — no new product features.
0.1 Fake NBA game: root cause was scheduleService fetching the ESPN
scoreboard with no ?dates= param or date filter — off-season ESPN
returns the NEAREST slate (Jun 13 NYK@SA Finals rendered as tonight).
Now pinned to the requested ET date + defensive filter; undated events
dropped. Honest month-aware per-sport empty states (lib/emptyState.js).
0.2 Fake header counters: liveTick stripped to a bare 1s pulse (the
auto-incrementing "247 graded", sin-driven brain-%, aPlus/cascades are
dead). New GET /api/snapshot/summary (cache-only, before /:sport) +
Next proxy; HeartbeatBar shows the real graded count and SYNC =
elapsed since the last pipeline run (amber past 5 min).
0.3 Ticker: hardcoded fallback items deleted (real snapshot exhaust only);
MOVE kept — computeLineDeltas is real movement. <4 real items → no
ticker; bar publishes --ticker-h so the fixed header collapses cleanly.
0.4 /terminal retired: route redirects to /dashboard; VVI/injury-wire/
leaders layouts preserved unrouted as §12 content-engine templates.
Nav PRIMARY = Slate/Scan/Ledger; BottomTabBar Terminal→Explore; PWA
shortcut Terminal→Ledger; #terminal alias → /dashboard.
0.5 ›Query nav pill deleted (duplicate /scan link).
Backend 2289 → 2309 tests (199 suites), web build exit 0.
Spec: specs/phase-0-kill-the-lies.md. Next: work-order Phase 1 (ledger
persistence + settlement).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Product overhaul core — the two transformative, differentiated systems:
Self-learning loop (Phase 2): outcomeService settles locked snapshot grades
against real MLB Stats API results → hit/miss/push, rolling accuracy by grade
tier (30d window). Idempotent, injectable, unit-tested. New GET /api/accuracy +
/api/ledger/accuracy + internal settle triggers + cron hook. AccuracyBadge
(dashboard/scan/landing) is honest — "LEARNING" below MIN_SAMPLE, never a fake
number. Settled HIT/MISS chips overlay the live slate.
Real-time layer (Phase 1): Slate silent 60s auto-refresh (no flash, no wipe on
transient blips) + "SIGNAL LIVE · UPDATED Xs ago" freshness strip; Ticker LIVE
badge that flashes on fresh events.
Landing (Phase 3): TopSignals shows tonight's real top-3 A-rated grades + live
accuracy — the product shown, not described.
Founder pricing: FOUNDER_CODE_EXPIRY default 2026-06-30 → 2026-12-31 (had
lapsed, disabling every founder code + the ClaimMeter pitch). That expiry — not
a tier change — was the real cause of the 4 stripe test failures.
Backend 2255 (4 failing) → 2274 (all green; +19 new, +4 fixed). Web build exit 0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
P1 name edge cases (BOTH playerName.js copies, kept identical):
- normalizeName strips hyphens (display+key): "Jung-hoo Lee" === "Jung Hoo Lee".
- nameKey strips single-letter MIDDLE tokens: "Josh H Smith" === "Josh Smith"
(keeps first+last; real middle names + collapsed initials untouched).
- richie -> richard added to NICKNAMES.
P2 polish:
- Team Hub names normalized at the source (teamService.getTeamHub) so
"J.C. Escarra" renders as "JC Escarra" like the dashboard.
- snapshotService dedup keeps the highest-confidence GRADE but the richest
DISPLAY (accented "José" over "Jose") so prop rows match the pitcher line.
- correlationWarning names the game: "2 legs from the same game (NYY @ BOS)".
Backend 2246 -> 2255 tests (+9), 194 suites. Web build clean (exit 0).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The link preview still showed engineer-speak ("Bayesian intelligence", "kill
conditions") because the OG meta tags were never updated after the S44 landing
cleanup.
- layout.tsx: de-jargoned the openGraph + main description + titles → "Pre-graded
player props with proprietary archetypes. Correlation-aware Parlay Lab.
Real-time line movement tracking. Built in Detroit." (Used "proprietary
archetypes" not "45" — codebase has 41.)
- NEW app/opengraph-image.tsx (+ twitter-image.tsx re-export): dynamically
generated 1200x630 card — VYNDR wordmark, "The books have every advantage. /
We built this to give it back." + feature row. Node runtime (not edge —
self-hosted standalone). Both routes prerender to a real PNG.
- Removed the stale /og-image.png metadata ref so the file convention owns the
image (no duplicate og:image tag).
Backend 2239 -> 2246 tests (+7), 193 suites. Web build clean (exit 0).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Trace-first: the normalizer functions were correct (S47) but raw names still
flowed through paths that skipped them. Fixed each leaking path.
- 2a (source chokepoint): snapshotService.runSnapshot normalizes each grade's
player to the de-dotted display AND dedupes to one grade per nameKey|stat
(highest confidence) before writing grades:{sport} + snapshot:latest. Every
consumer (GameCard, Explore, leaders, profile) now gets clean merged names.
- 2b: buildPlayerStripsFromProps dedupes a player's props by stat (graded >
awaiting) → one row per stat (kills "Ks 5.5 AND Ks 3.5" variant dupes).
- 2c: scan tonightsPlayers grid groups by nameKey, displays normalized name.
- 3: profile VYNDR INTELLIGENCE "+0%"/"—" was buildIntel's defaults (separate
from the grade card's buildIntelFields, which already works). resolvePlayerStats
now attaches real usage (AB/G) + rest (B2B/Xd); buildIntel renders them; REST
default is now "—".
Backend 2149 -> 2156 tests (+7), 181 suites. Web build clean (exit 0).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Name normalization completed: NICKNAMES table (Matt↔Matthew, Mike↔Michael...)
resolved in nameKey, parenthetical team-tag strip "(STL)", verified accent-fold
(Iván/Ivan, José/Jose). Slate strip now DISPLAYS the normalized de-dotted name
("AJ Ewing" not "A.J. Ewing") via buildPlayerStripsFromProps.
- Complete MLB VYNDR INTELLIGENCE: mlbGameLogFeatures derives rest_days (days off
between latest games; 0=B2B) + ab_per_game (usage). buildIntelFields renders
usage as "X AB/G", rest as B2B/Xd, matchup from bvp_advantage fallback.
- Ticker SCAN dedup: pushTickerItems keeps one SCAN per sport (sport field or
text-prefix parse for legacy); MOVE/GRADE preserved; cap 50.
- BOMBER threshold prorated for mid-season (hr>=15 strong / >=10 mod) so June
sluggers classify BOMBER not FLEX/DRIVER.
Backend 2122 -> 2149 tests (+27), 179 suites. Web build clean (exit 0).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Three focused P1 fixes on the Session-45 snapshot model.
- Grade card intel ROOT CAUSE: gameLogService is NBA/WNBA-only (offline Python),
so MLB props never got l5_avg/l20_avg and buildIntelFields returned {}. Wired
MLB game logs into featureCache.gameLogFeatures via mlbStatsAdapter.getPlayerStats
(pure mlbGameLogFeatures + MLB stat_type->field map). buildIntelFields gained
playerStats/projection fallbacks for partial intel.
- Player name normalization: src/utils/playerName.js (+ web/src/lib copy):
normalizeName -> {display,key}. Strips periods, de-dots suffix, accent-folds
the key. Applied in snapshotService grouping, slateAdapter grade index +
player-strip merge (variants collapse, longest name shown), and
playerIntelService. "A.J. Ewing"/"AJ Ewing" + "Jazz Chisholm"/"Jr." now merge.
- MLB starting pitchers: new GET /api/schedule/:sport/pitchers (probablePitchers
service wrapping mlbStatsAdapter.getScheduleWithPitchers + best-effort ERA).
Slate fetches it, builds a team->pitcher map (full name + mascot match),
attaches pitchers to MLB GameCardData. + Next proxy.
Backend 2100 -> 2122 tests (+22), 176 suites. Web build clean (exit 0).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The on-demand "Read" grade model is RETIRED. A scheduled pipeline pre-grades the
slate, locks grades to the line, tracks movement; the dashboard shows them already
there. Orchestrates existing services — nothing rebuilt.
- snapshotService.runSnapshot(sport): getOdds → gradeAndCacheSlate → classify
archetype per player → lock gradedAt → line deltas vs previous snapshot → write
snapshot:{sport}:latest/previous + grades:{sport} → ticker events. Fully
injectable, zero-network unit tests. runAllSnapshots = cron entrypoint.
- Internal trigger POST /api/internal/snapshot/:sport + /all (requireInternalAuth).
In-process cron (SNAPSHOT_CRON=1, UTC 14,19,22,1,3) in server.js, no new dep.
- Public reads: GET /api/snapshot/:sport (cache-only) + GET /api/ticker (merges
TICKER_MANUAL pins) + Next proxies.
- GameCard swap: live Slate renders vyndr/GameCard (legacy kept for types only),
overlays locked grades onto game props → player name once + archetype badge +
"Graded Xh ago at -115 · Current 2.5 · ▲ TOWARD +1.0". Ungraded → "Awaiting next
scan", NO Read button. On-demand onGrade flow deleted.
- Ticker polls /api/ticker every 30s, graceful fallback to hardcoded items.
- NBA/WNBA: espnStatsAdapter free fallback (defensive parse → found:false on shape
mismatch) wired into resolvePlayerStats after the offline Python service.
Env: PROPLINE_API_KEY_1/2/3, VYNDR_INTERNAL_KEY, SNAPSHOT_CRON=1, TICKER_MANUAL.
Backend 2061 -> 2100 tests (+39), 173 suites. Web build clean (exit 0).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
P0 fixes + wiring real data into the S42 Player Intelligence architecture.
- P0 dropdown z-index: the nav's backdrop-filter stacking context let the
Ticker/HeartbeatBar paint over the avatar/More dropdowns and eat clicks.
nav now position:relative zIndex:2; menus zIndex:100. Avatar Settings ->
/settings.
- Real MLB stats: mlbStatsAdapter.searchPlayer + getPlayerStats (name->id->
season+gamelog). playerIntelService.resolvePlayerStats normalizes into the
archetype classifier; getPlayerIntel returns found:true + real season +
archetype classified from real stats. NBA via nbaStatsClient (degrades).
- Game cards: slateAdapter.groupPropsByPlayer (playerStrips, name once) +
mapPitchers (MLB probables), folded into mapScheduleToGameCards. Legacy
GameCard line grid renders BookChip (brand colors) not grey text.
- Grade card intel: analyzeViaEngine1.buildIntelFields computes stat-context +
form/usage/matchup/rest from the existing feature vector (zero extra I/O);
gradeAdapter lights up the card sections. Archetype deferred (needs season
line at grade time).
- Depth chart foundation: depthChartService (getLineup/getDepthChart/
getCascadeProjection) + /api/stats/lineup|depth|cascade, graceful + injectable.
- Mobile: player hero name overflow-wrap + 24px on <=640px (was clipping).
Backend 2011 -> 2045 tests (+34), 163 suites. Web build clean (exit 0).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Final phase of the VYNDR 2.0 conversion (Sessions 33-39). Verify -> fix -> lock.
Frontend-only; zero backend changes.
§13 automated checklist: all PASS or FIXED.
- QA.1 token resolution FIXED: ProcessingGrade #00ffb8 -> var(--g-ap);
game/[id] sport literals -> var(--s-*). Remaining hex documented as intentional
(var-with-fallback, Next metadata, bespoke intel-surface shades).
- QA.6 glitch discipline: ZERO glitch on data components.
- QA.4/5/8/11/16/18 verified; QA.9 (cmd palette) documented deferred;
QA.17 (AI slop) flagged for Kev's manual browser review.
De-flake: soccerFeatureExtractorCascade hit Jest's 5s default under full-suite
load (falls through to live adapters on cache miss) -> jest.setTimeout(20000),
same family as the S32 pipeline test. Verified stable across 3 full-suite runs.
New: tests/unit/vyndrParityQA.test.js (17 tests locking the parity invariants).
Backend 1890 -> 1907, 146 suites, zero regressions (stable x3). Web build clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
VYNDR 2.0 conversion, Phase G (the systems that make the design alive). All 5
wired. Frontend-only; zero backend changes.
- lib/parlayMath.js: correlation model (0.62/0.34/0.06/0) + parlayGrade penalty
+ grade->odds + combined odds (frontend; backend parlayService unchanged).
- lib/oddsFormat.js: fmtOdds across american/decimal/fractional/implied with the
totals-pass-through rule (safer than the prototype's parseAm, which would
mis-convert 228.5) + region presets.
- lib/prefs.js: applyPrefs sets <html data-*> (the S33 a11y CSS layer) + load/save.
- lib/liveTick.js: single tick engine (SSR/test-safe, no auto-start, fresh state).
- lib/checkout.js: checkoutUrl(plan).
- LiveLayer (useLive/LiveNumber/HeartbeatBar) under the Nav ticker; GlobalHosts in
layout applies prefs + registers __prefs/__goPaywall/__checkout + hosts the
Preferences and Paywall modals. Nav read-meter is now a paywall trigger.
Gotchas: useEffect can't return a Set.delete unsub directly (boolean != cleanup);
header grew to 124px so layout paddingTop + Slate sticky-top updated to match.
18 new tests. Backend 1872 -> 1890, 146 suites, zero regressions. Web build clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
VYNDR 2.0 conversion, Phase F (mobile is the PWA we launch first). Frontend-only;
zero backend changes.
- BottomTabBar rewritten to the §6 5-tab spec: Slate/Terminal/Scan/Ledger/More,
with Scan as the prominent raised grade-green action. Shown for anon too (only
mobile nav). Integrated More bottom sheet (sheet-up, backdrop dismiss, 48px mono
rows). iOS safe-area + 44px touch targets.
- Nav hamburger retired on mobile (tab bar owns nav).
- globals.css mobile section: tab-bar hidden >=768, main bottom padding,
grade-hero 80px, terminal-grid stacks, game-lines horizontal scroll.
- PWA: manifest shortcuts (Slate/Scan/Terminal) + categories; viewport-fit=cover.
Gotcha: `as const` on the TABS array broke type-check (distinct literal types);
fixed with a shared TabDef interface.
19 new tests. Backend 1853 -> 1872, 145 suites, zero regressions. Web build clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
VYNDR 2.0 conversion, Phase E. Frontend-only; zero backend changes.
- lib/slateAdapter.js: parseAmericanOdds, detectBestLines, mapScheduleToGameCards
(best/worst line detection — the Bloomberg pattern).
- Reskinned the LEGACY GameCard's game-lines grid with best/worst highlighting +
SportBadge, keeping inline grading intact (a wholesale swap to the display-only
vyndr/GameCard would have deleted the slate's grading interaction).
- compare/invite/help/about: RouteStubs -> real design-system pages.
- login reskinned (scanlines, system voice, new Wordmark); pricing + ClaimMeter.
Honest scope: the full GameCard swap needs inline grading ported into the new
component first; profile/settings/blog/game-detail reskins are light/deferred.
18 new tests. Backend 1839 -> 1853, 144 suites, zero regressions. Web build clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
VYNDR 2.0 conversion, Phase D (the screens users touch). Frontend-only; zero
backend changes.
- GradeResultCard + ProcessingGrade (the core product moment): intel-surface
grade hero, signal breakdown, kill conditions, best-book strip, alt ladder;
sections self-hide when empty.
- lib/gradeAdapter.js maps engine output -> §7 contract and tier-gates content
(free teaser / analyst kill-conditions / desk alt ladder) so the new card
doesn't give paid content away.
- Scan result wired to ProcessingGrade->GradeResultCard, preserving scan limits,
parlay add, reads tracking, and noopener sportsbook deep-links.
- GameCard (Bloomberg best/worst line cells) built + tested.
- Terminal page replaces its stub with a real league-intelligence screen.
- Landing gets the founder-seat ClaimMeter.
Honest scope: live dashboard/Slate swap onto GameCard, scan input -> TerminalInput,
full landing rebuild, and the blurred-paywall polish (Phase G) are deferred to
keep working flows stable.
22 new tests. Backend 1818 -> 1839, 143 suites, zero regressions. Web build clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
VYNDR 2.0 conversion, Phase C (the frame every page sits inside). Frontend-only;
zero backend changes.
- Nav rewritten: new .wm Wordmark, mono uppercase links, More dropdown, search/
bell/read-meter/avatar, Ticker under the bar. layout main paddingTop 64 -> 96.
- Routing: web/src/lib/routes.js (GATED/OPEN/HASH_ALIASES, isGatedRoute,
resolveHashAlias). Client AuthGate bounces signed-out users off personal
routes to /login?next=. HashRedirect maps #scan/#terminal to real routes.
- Footer rewritten to system voice + Detroit signature; mounted globally in
layout (removed per-page dup).
- 404 converted to the north star (scanlines, crt-sweep, glitch wordmark, amber).
- Stub pages for terminal/compare/invite/help/about/notifications via RouteStub.
Honest reconciliations: auth gate is client-side (no auth-helpers pkg; session is
client-side Supabase); GATED narrowed to protect the free-scan funnel; did not
stub over existing real pages; redirect param is ?next= (what /login reads).
26 new tests. Backend 1792 -> 1818, 142 suites, zero regressions. Web build clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- gradeSlateService writes grades:{sport} cache (closes content pipeline →
dataLevel full); fire-and-forget from oddsService.recordDownstream, gated
by shouldGradeSlate (off in test, GRADE_SLATE_ON_FETCH override)
- NFL/NHL wired: oddsService SPORT_KEYS/SPORT_MARKETS (correct the-odds-api
keys americanfootball_nfl/icehockey_nhl), proplineAdapter MARKETS, NHL
MARKET_MAP keys to avoid silent-zero
- rate limiting mounted on 8 public cached routers (odds/parlay 30/min,
rest 60/min)
- jsonlLogger writes to temp under test (no more dirtied tracked artifact);
5MB pipeline test given 20s timeout
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Add NFL keys to oddsNormalizer.MARKET_MAP (defensive; same silent-zero
class as the Session 30 MLB bug) + NFL surface test
- npm audit fix: ws/qs + Supabase transitives, 7 vulns -> 0 (semver-safe)
- Audit findings documented in BUILD-STATE: grades cache has no writer,
NFL/NHL not wired end-to-end, rate limiting only on /analyze, tests
mutate a tracked jsonl, leaked GitHub PAT in origin remote (rotate)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
builtbykev