Bets OS — Status

01What to Know

Component	Last Run	Notes
Pulse (board refresh)	07/14/2026 18:01:51	3× daily
Grade (AJAI_BetsGrade)	—	08:00 daily
Digest (AJAI_DailyDigest)	2026-07-14T16:15	10:45 + 17:15
Drip pull	78/120 nav (65%)	replay archive
Board deployed today	4	wrangler pages
Loop locks	idle	watch=idle
Odds API budget	28 used / 472 remaining	2026-07

Proven-positive cappers (W/L record only, no CLV):

Capper	Units Net	Record
TheGamblingGawd	+98.4u	n=172 · 58.2% hit
Anders Picks	+18.4u	n=121 · 61.3% hit
Smart Money Sports	+11.9u	n=125 · 54.1% hit
Nicky Cashin	+10.2u	n=181 · 54.7% hit

All four have sport-concentration caveats. In-sample only. 0 / 33 backtest variants survive.

CLV Report · 2026-07-14 12:00 UTC

INSUFFICIENT DATA

3 closing lines in closing_lines.jsonl. CLV is the honest signal; W/L is a lagging proxy.

Fade gate (paper trade opened 2026-07-12):

Capper	Status	Notes
MrBigBets	ACTIVE	n=0 · verdict 2026-08-09
ThisGirlBetz	ACTIVE	n=0 · verdict 2026-08-09
Underdog Sniper	ACTIVE	n=0 · verdict 2026-08-09

Forward lab (7 hypotheses):

ID	Name	Status	Progress
H1	dog-bucket	ACCUMULATING (n=2)	n=2 / 49d left
H2	spread-over-ML	ACCUMULATING (S=0/ML=23)	n=0 / 49d left
H3	late-post	DEFERRED (30 time_unknown)	n=30 / 49d left
H4	NBA-pool	PAUSED (n=0 — no NBA until Oct	n=0 / Noned left
H5	conviction	OFF-TRACK	n=— / 49d left
H6	consensus-card	ACCUMULATING (n=0)	n=0 / Noned left
H7	fade-the-inflated	DELEGATED → fade_gate.py	n=0 / 26d left

Replay-30: 9W-9L +0.137u · n=18 · ROI +0.76% · CI [−45.3%, +46.7%] · NOT SIGNIFICANT

04Net Net

The system has 5,730 graded picks across 33 archive days, 4 proven-positive cappers by W/L record (no CLV verified yet), and 3 closing-line entries captured so far. The 33-variant backtest returned 0 survivors — no edge was found on historical data. The forward lab has 7 hypotheses accumulating; none has met its minimum-n gate. The fade-gate paper-trade opened 2026-07-12 with verdict expected 2026-08-09. The honest product right now is infrastructure and transparency, not alpha. CLV accumulation is the critical path: once 3 grows to ~50–100 per capper, the tier system can be rebuilt on signal that actually means something.

02What to Do

5 open decision cards in queue/:

buddy-activation-card.md

Buddy Cast -- Activation Card

Rec: see card

decision-card-actionability.md

Decision Card — Bets Stack Next Investment of Attention

Rec: [A]

[A] / [B] / [C] / [D] / [E]

decision-card-sources-next.md

Decision Card — Source Expansion: capperinsights.com EVALUATED + Discord / Reddi

Rec: [A] — wire nightly baseline cron.

[A] / [B] / [C] / [D]

overnight-clv-report.md

Overnight CLV Report — Beat the Close

Rec: see card

[A] / [B] / [C]

upgrade-wave-report.md

Upgrade Wave Report

Rec: see card

[A] / [B] / [C]

03Architecture / How It Works

Daily Flow

Time	Task
08:00	AJAI_BetsGrade — grade yesterday · CLV capture · fade-gate
08:05	AJAI_ForwardLab — classify picks → hypotheses
08:15	AJAI_BoardPulse_AM — pull → grade → gen → deploy
08:20	AJAI_StatusPage — this page
10:45	AJAI_DailyDigest — board_play.json + today_card.json
12:00	AJAI_BoardPulse_Noon
17:15	AJAI_DailyDigest (eve)
18:00	AJAI_BoardPulse_Eve

Entry points AJ ever touches

1 — Nothing pulses run automatically
2 — card_day.cmd on game nights: starts live_watch + board_refresh loops
3 — /bets in Claude digest skill → ranked card from today_card.json

Data: X/Twitter (12 accts) → capper_feed.json → ESPN slate + odds → board_gen.py → public/index.html → wrangler pages deploy → aj-board.pages.dev
Logs: pulse.log · grade_morning.log · digest.log · capper_pull.log

05Next Steps

#	Action	Detail
1	[A] decision-card-actionability	Scale replay to 30 days when grader is stable ≥3 mornings
2	[A] decision-card-sources-next	Wire CI baseline nightly cron — 60-capper long-run records as sanity check
3	[A] upgrade-wave-report	Reduce Match Point Bets weight to 0.5 floor (32pp ROI gap on same sport)
4	Monitor	Fade-gate verdict date: 2026-08-09 · n=0 for all 3 candidates
5	Monitor	CLV: 3 entries in closing_lines.jsonl · target 50+/capper before rebuild
6	Monitor	Lab hypotheses: check back 2026-09-01 when verdict windows close

aj-board.pages.dev/status · status_gen.py · 2026-07-15 01:44 UTC
Data honesty: n + receipts only. Descriptive ≠ predictive. No +EV claims.