NBAI / Research

NBA / Evidence before outcomes

The research board.

A clear view of what we know, what is missing, and what still needs to be tested.

Historical totals / Coverage only

Historical totals candidates

Operator-reported · independently unqualified

Collection-stage snapshot: . These are earlier collection-stage candidates, not the later 6,059-game evaluated population.

6,083 / 6,150 candidate games

Five regular seasons · 30 teams · 150 team-season cells. Candidates are not betting signals or a validated strategy.

67 excluded games: 63 with no accepted pregame quotes (operator category; pregame status remains unverified), and 4 with ambiguous provider identity.

2021-22
1,226 / 1,230 candidates; 4 excluded
2022-23
1,175 / 1,230 candidates; 55 excluded
2023-24
1,230 / 1,230 candidates; 0 excluded
2024-25
1,226 / 1,230 candidates; 4 excluded
2025-26
1,226 / 1,230 candidates; 4 excluded
All 150 team-season cells
Operator-reported candidate / scheduled team appearances. Each game appears for two teams.
Team2021-222022-232023-242024-252025-26
ATL82 / 8280 / 8282 / 8281 / 8282 / 82
BKN81 / 8281 / 8282 / 8282 / 8282 / 82
BOS81 / 8281 / 8282 / 8282 / 8282 / 82
CHA82 / 8278 / 8282 / 8282 / 8282 / 82
CHI82 / 8281 / 8282 / 8282 / 8282 / 82
CLE82 / 8279 / 8282 / 8282 / 8282 / 82
DAL81 / 8279 / 8282 / 8282 / 8281 / 82
DEN82 / 8277 / 8282 / 8282 / 8282 / 82
DET82 / 8281 / 8282 / 8282 / 8281 / 82
GSW81 / 8280 / 8282 / 8282 / 8282 / 82
HOU82 / 8278 / 8282 / 8281 / 8282 / 82
IND82 / 8280 / 8282 / 8281 / 8282 / 82
LAC82 / 8279 / 8282 / 8282 / 8282 / 82
LAL81 / 8277 / 8282 / 8282 / 8282 / 82
MEM82 / 8274 / 8282 / 8282 / 8281 / 82
MIA82 / 8281 / 8282 / 8281 / 8282 / 82
MIL82 / 8280 / 8282 / 8281 / 8282 / 82
MIN82 / 8274 / 8282 / 8282 / 8282 / 82
NOP82 / 8280 / 8282 / 8282 / 8282 / 82
NYK82 / 8276 / 8282 / 8282 / 8281 / 82
OKC82 / 8278 / 8282 / 8281 / 8281 / 82
ORL82 / 8279 / 8282 / 8282 / 8280 / 82
PHI81 / 8280 / 8282 / 8282 / 8282 / 82
PHX81 / 8276 / 8282 / 8282 / 8282 / 82
POR82 / 8276 / 8282 / 8282 / 8282 / 82
SAC82 / 8279 / 8282 / 8282 / 8282 / 82
SAS82 / 8278 / 8282 / 8281 / 8281 / 82
TOR82 / 8276 / 8282 / 8282 / 8282 / 82
UTA81 / 8276 / 8282 / 8282 / 8282 / 82
WAS82 / 8276 / 8282 / 8281 / 8282 / 82
Source snapshot and hash provenance

Source generated at . Static historical snapshot, not live coverage.

Unchanged originals are staged outside the public web folder. Original SHA-256 hashes:

  • coverage.json: 77760d8d3b89105141b111aacfdac73690a154a90386ee54683470315a331d79
  • team_season_coverage.csv: bad5924d8fd6d16888862c88cd1f7695809f93629bc71f382a36c4a63fbb44ab
  • missing_or_unpaired_games.csv: 80d507b96e40a024470b35306c751d27ef9c8ba8909a5218b794f2c46c461a7d

Derived public aggregate (JSON) · Original versus derived hash manifest

Hashes bind bytes, not source truth. Team appearances reconcile to twice the game count; gap identities reconcile to team-season shortfalls.

Historical totals / Retrospective results

Both betting strategies lost

Operator-reported retrospective results · Independent acceptance pending

Five regular seasons, 2021–22 through 2025–26, targeting all 30 teams. This is not full league completeness.

6,059 / 6,150 evaluated games

91 exclusions: 55 with no provider quotes + 36 with ambiguous provider identity. Separate from the earlier collection-stage coverage above.

Corrected model

Eligible opportunities
6,059
Bets
4,548
Wins
2,291
Losses
2,219
Pushes
38
Net return
-136.49 units
ROI (net / units staked)
-3.00%
Reported 95% ROI interval
-5.89% to -0.21%
Max daily-close drawdown
168.21 units

Half model / half market

Eligible opportunities
6,059
Bets
3,155
Wins
1,615
Losses
1,515
Pushes
25
Net return
-46.66 units
ROI (net / units staked)
-1.48%
Reported 95% ROI interval
-4.92% to +2.05%
Max daily-close drawdown
101.14 units

No-vig market baseline

Eligible opportunities
6,059
Bets
0
Wins
0
Losses
0
Pushes
0
Net return
+0.00 units
ROI (net / units staked)
Unavailable (zero bets)
Reported 95% ROI interval
Unavailable (zero bets)
Max daily-close drawdown
0.00 units

One unit staked per bet, at most one bet per game per method. The baseline placed zero bets; its ROI and interval are unavailable, not 0%.

Reported intervals: 2,000 game-date cluster bootstrap resamples, fixed seed 417. They do not capture all model-selection or season-regime uncertainty. Drawdown uses daily-close equity, not intraday peaks.

Original summary and derived hash provenance

Run 1503e03be0fc4892a621ea215f4ecf07 · snapshot .

Original finished.json SHA-256: ef6168eaeb7d3d8fdf844509db58f4f740994dd6e6cdd8ccf5fac203e16b0be1. Unchanged original staged outside the public web folder.

Derived public results (JSON) · Original versus derived hashes

Hashes bind bytes, not truth. Counts and returns reconcile across summary seasons and periods; intervals and drawdown are copied from the reported aggregate, not independently recomputed from bets.

Data unavailable — waiting for a valid snapshot. No picks loaded.

01 / Historical evidence

Retrospective research

No historical summaries loaded. Minutes-model evidence is not betting validation.

02 / Research signals

Quote & forecast board

No research signals. Actual quotes and forecasts are unavailable.

03 / Forward evidence

Forward paper record

0 verified forward observations

No persisted ledger is connected. Performance is unvalidated.