NBA Line Watch
Where we think a game line is wrong, and whether the closing line came our way.
Open signals
| Game | Pick | Model | Market | Why |
|---|---|---|---|---|
PHI @ NYK 2026-10-20 · seen 14 Sept, 09:23 | NYK -5.5 1.95 bet365 | -10.7 | -5.5 | Model against the first line we saw The model was 5.2 points from the first line. |
PHI @ NYK 2026-10-20 · seen 14 Sept, 09:23 | Under 232.5 1.90 bet365 | 224.7 | 231.8 | Model against the first line we saw The model was 7.1 points from the first line. |
MIL @ WAS 2026-10-21 · seen 14 Sept, 09:23 | MIL +6.5 1.91 betmgm | -0.1 | -6.5 | Model against the first line we saw The model was 6.4 points from the first line. |
CHA @ BKN 2026-10-21 · seen 14 Sept, 09:23 | CHA -4.5 1.91 betmgm | +7.8 | +4.5 | Model against the first line we saw The model was 3.4 points from the first line. |
UTA @ MEM 2026-10-21 · seen 14 Sept, 09:23 | MEM +2.5 1.98 draftkings | -0.7 | +2.5 | Model against the first line we saw The model was 3.2 points from the first line. |
UTA @ MEM 2026-10-21 · seen 14 Sept, 09:23 | Over 241.5 1.91 betmgm | 249.6 | 241.5 | Model against the first line we saw The model was 8.1 points from the first line. |
IND @ NOP 2026-10-21 · seen 14 Sept, 09:23 | NOP +4.5 1.86 draftkings | -4 | +3.8 | Model against the first line we saw The model was 7.8 points from the first line. |
CLE @ PHI 2026-10-22 · seen 14 Sept, 09:23 | CLE +4.5 1.91 betmgm | -0.2 | -4.5 | Model against the first line we saw The model was 4.3 points from the first line. |
Picking winners is not an edge: last season our model called 68% of games, exactly as often as the bookmakers' favourite, and betting it at closing prices broke even. What does show an edge is beating the closing line, getting a better number than the market settles on at tip-off. Line Watch records every time we think a spread or total is wrong, at the best line on offer across five bookmakers, and grades each one twice: against the closing line, and against the result. Four kinds of signal are tracked. Team news: a player is ruled in or out, our number moves, and the market's has not yet. A slow book: one bookmaker's line sits away from the others, which on last season's closing lines was worth betting at that book in the others' direction. The opener: the model's own number against the first line posted. And two situational totals leads from last season's closing lines, followed here to see whether they hold up in a season they were not found in. These are tracked for measurement, not published as tips.
Frequently asked
The difference between the line a signal was recorded at and the line at tip-off, in points. Bet a side at -3 that closes at -5 and the market has moved two points your way. Results are noisy over a few hundred bets; closing line value is not, which is why professionals judge a method by it first.
A change in who is available that moves our projected margin by at least 1.5 points (2 for totals) more than the consensus line has moved; one bookmaker's line at least 1.5 points (2 for totals) from the average of the others; a first posted line at least 3 points (5 for totals) from the model's own number; or a game matching one of the tracked totals leads, a team on the second night of a back-to-back, or March and April of the regular season.
The best line available for that side across Bet365, DraftKings, BetMGM, Caesars and BetRivers when it was recorded, at that book's price. The closing line is the average of each book's last line before tip-off.