Every pick our models make this season, game bets and player props. A pick lands here once its game is over. Nothing is ever deleted.
MEM @ SAC
DET @ PHI
TOR @ MIN
ORL @ NYK
IND @ BKN
LAL @ UTA
GSW @ PHX
NOP @ DEN
HOU @ ATL
BKN @ CLE
MIA @ CHA
POR @ SAC
POR @ SAC
LAC @ LAL
LAC @ LAL
HOU @ SAS
NYK @ BOS
TOR @ WAS
DAL @ IND
CHI @ MIL
DEN @ OKC
SAC @ LAC
GSW @ LAL
CHI @ TOR
CHA @ BKN
OKC @ SAS
PHI @ NYK
DET @ PHI
IND @ BKN
ORL @ NYK
ORL @ NYK
DET @ PHI
DET @ PHI
LAL @ UTA
TOR @ MIN
LAC @ OKC
DET @ PHI
DET @ PHI
DET @ PHI
DET @ PHI
DET @ PHI
GSW @ PHX
GSW @ PHX
GSW @ PHX
GSW @ PHX
SAS @ DAL
SAS @ DAL
GSW @ PHX
SAS @ DAL
SAS @ DAL
GSW @ PHX
GSW @ PHX
MIL @ PHI
MIA @ CHA
GSW @ PHX
GSW @ PHX
GSW @ PHX
GSW @ PHX
GSW @ PHX
MIL @ PHI
MIA @ CHA
GSW @ PHX
HOU @ ATL
GSW @ PHX
SAS @ DAL
NOP @ DEN
MIA @ CHA
MIA @ CHA
MIA @ CHA
SAS @ DAL
WAS @ CHI
MIA @ CHA
NOP @ DEN
NOP @ DEN
NOP @ DEN
GSW @ PHX
GSW @ PHX
GSW @ PHX
WAS @ CHI
BKN @ CLE
NYK @ BOS
NYK @ BOS
NYK @ BOS
DET @ MIA
HOU @ SAS
CHI @ MIL
NOP @ UTA
LAC @ LAL
POR @ SAC
NYK @ BOS
Before going live, we tested the model on five past seasons. Each season, it only learned from the seasons before it, so it never saw the games it was picking.
Pinnacle is the sharpest betting market in the world. On average, our price was 3.2% better than Pinnacle's final price before tip-off. Example: we took 2.00, and the fair price at the close was 1.94.
Bet $100 on each of the 1,938 picks and you would end up about $8,800 ahead. Spreads alone: +6.0%, about $7,800 on 1,306 picks.
| Season | Picks | Beat the close | Profit |
|---|---|---|---|
| 2021-22 | 378 | +0.7% | +14.5% |
| 2022-23 | 356 | +5.3% | -3.4% |
| 2023-24 | 413 | +2.1% | +2.8% |
| 2024-25 | 311 | +4.8% | +7.6% |
| 2025-26 | 480 | +3.6% | +2.2% |
| All five | 1,938 | +3.2% | +4.6% |
| Bet type | Picks | Beat the close | Profit |
|---|---|---|---|
| Spreads | 1,306 | +3.1% | +6.0% |
| Over / under | 491 | +3.0% | +2.8% |
| Winners | 141 | +4.9% | -2.2% |
Why do the profits jump around so much, even when we beat the close? Luck. A few hundred bets is too few: a handful of close games going the other way swings the result. Beating the closing price is the better sign of real skill.
Before going live, we re-played the props model on three past seasons exactly as a bettor would have: when a line was posted, using only what was known at that moment, one pick per player line at the best of DraftKings, FanDuel, BetMGM and Caesars, 1 unit each, with today's betting rule. The model only ever learned from seasons before the one being tested. 2025-26 uses each book's first posted price; for 2023-24 and 2024-25 the data allows the price 14 hours before tip, which is a little harsher.
5,255 picks, 52% won. Bet $100 on each and you would end up about $41,500 ahead. Every season was positive; the weakest was +5.1%.
2025-26: 1,150 picks at the books' first posted prices, 54% won. Opened once, at the end, with the rule already fixed.
| Season | Picks | Won | Profit |
|---|---|---|---|
| 2023-24 best of 4 books, price 14h before tip | 2,403 | 52.4% | +8.8% |
| 2024-25 best of 4 books, price 14h before tip | 1,702 | 51.4% | +5.1% |
| 2025-26 best of 4 books, first posted price | 1,150 | 53.7% | +10.3% |
| All three | 5,255 | 52.3% | +7.9% |
Nearly all picks are unders, about 97 in 100. Books tend to set player lines a little high.
An example. Our model picks Celtics -4.5 on Monday at a price of 1.95. By tip-off on Wednesday, the market has moved: the same bet now costs 1.88. We got a better price than the market's final word, by +3.7%.
Our price is the price when we made the pick (Pinnacle, the biggest betting market). Closing price is the fair price for the same bet just before the game starts, with the bookmaker's cut removed.
Why "beat the closing price" matters most. Wins and losses are mostly luck over a few hundred games. The closing price is the market's best guess with all the news in. Bettors who keep getting a better price than the close win in the long run, even through bad weeks.
Profit counts 1 unit on every pick. A win at 1.95 pays +0.95u. A loss is -1u.
Player props. Example: our model guesses Jaylen Brown makes 1.7 threes, and a book offers under 2.5 at 1.91. We take it. If he makes 0, 1 or 2, the pick wins +0.91u. If he makes 3 or more, it loses -1u. Our guess is the model's number for that player tonight. Our price is the book's price when we made the pick.
Game picks are spreads, totals (over / under) and winners. Player props are points, rebounds, assists, threes, steals and blocks. NBA regular season, 2026-27. This is a record of our opinion, not betting advice.