Preseason model · OCT 20 – OCT 22 · 16 games
Our numbers next to the books. Analysis, not advice.
Model vs line
Tap a game for the detailPreseason model. Big gaps here mostly mean our model hasn't caught up with offseason changes yet, not that the books are wrong.
| Game | Book ML (home) | Book win % | Our win % | Edge | Kalshi win % | Polymarket win % | Spread (home) | Our cover % | Total |
|---|---|---|---|---|---|---|---|---|---|
| -115 / -105 | 51.1% | 61.6% | +10.5 pts | 49.5% | 49.5% | DET -1.5 | 54.3% | 222.5 | |
| -200 / +165 | 63.9% | 81.5% | +17.6 pts | 63.5% | 63.0% | NYK -5.5 | 64.6% | 231.5 | |
| -134 / +111 | 54.7% | 58.9% | +4.2 pts | 55.0% | 55.5% | SAS -2.5 | 49.2% | 231.5 | |
| -144 / +119 | 56.4% | 62.2% | +5.8 pts | 57.8% | 58.0% | ORL -3 | 50.7% | 233.5 | |
| -218 / +180 | 65.7% | 41.3% | -24.5 pts | 66.5% | 67.5% | WAS -6 | 25.8% | 238.5 | |
| -122 / +104 | 52.9% | 58.1% | +5.2 pts | 51.7% | 51.0% | MIA -1.5 | 51.2% | 243.5 | |
| +108 / -130 | 46.0% | 22.3% | -23.7 pts | 46.0% | 45.5% | BKN +2 | 28.3% | 221.5 | |
| -350 / +275 | 74.5% | 86.2% | +11.7 pts | 76.0% | 75.0% | TOR -8.5 | 63.8% | 235.5 | |
| +106 / -128 | 46.4% | 55.7% | +9.4 pts | 47.0% | 47.0% | MEM +1.5 | 57.4% | 242.5 | |
| +130 / -155 | 41.7% | 65.2% | +23.5 pts | 40.3% | 41.5% | NOP +3.5 | 70.5% | 242 | |
| -318 / +255 | 73.0% | 83.9% | +10.9 pts | 73.9% | 74.5% | HOU -8.5 | 60.1% | 228.5 | |
| -199 / +165 | 63.8% | 70.8% | +7.0 pts | 63.9% | 64.5% | LAL -5.5 | 52.1% | 234.5 | |
| -161 / +136 | 59.3% | 53.5% | -5.7 pts | 60.1% | 60.0% | POR -4 | 40.4% | 234 | |
| -155 / +129 | 58.2% | 73.2% | +15.0 pts | 58.5% | 58.5% | LAC -4 | 58.7% | 230 | |
| -150 / +125 | 57.4% | 46.8% | -10.6 pts | 58.2% | 56.0% | PHI -3.5 | 36.4% | 234.75 | |
| -295 / +228 | 71.0% | 77.0% | +6.0 pts | 71.0% | 70.0% | OKC -7.5 | 53.7% | 234 | |
Book win % removes the sportsbook's cut from the consensus moneyline. Example: -150 / +130 imply 60.0% and 43.5% (103.5% in total), so the fair split is 58.0% / 42.0%. Edge is our win % minus that. Cover % uses our projected margin with a spread of 14.4 points, from our margin error on 2025-26. Totals are shown without a model number. Kalshi win % is the midpoint of Kalshi's bid and ask for each team, scaled so both teams add to 100%; tap it to open the market on Kalshi. Prices come from US-licensed sportsbooks only.
Model leans
None openLeans start once teams have played about 20 games. Until then our model runs on last season's ratings and this summer's roster moves, and it disagrees with the books far more than a real edge would (sometimes by 30 points), so we don't publish picks from it. Model vs line above still shows where it stands.
Track record
Starts opening night- Record
- 0-0
- Units
- +0.0
- Return per bet
- –
- Moneyline / spread
- 0-0 / 0-0
Every lean is frozen at tip-off at the price shown and graded after the final, 1 unit per lean. Nothing is removed after the fact. A good record over a few dozen bets can still be luck; judge it over hundreds.
How leans work
A moneyline lean needs our win chance to beat the no-vig book number by at least 4 points (6+ is a strong lean). A spread lean needs our cover chance at the consensus line to reach 54% (57%+ is strong); at -110 you need 52.4% just to break even. Leans update with the lines until tip-off, then freeze.
Leans only run once both teams have played 20 games. Before that the model leans on last season's ratings and this season's rosters, which we have not been able to test, and its gaps with the books are mostly its own error. Win probability accuracy and calibration are on the win probability page.
Building a parlay? The parlay builder shows the real payout next to our chance of hitting it. Player props: check any line against our projection on the projections page.