How Often Does the Over Hit in the NBA? Eight Seasons of Closing Totals

The final score went over the closing total 4,761 times and under 4,640 times, with 107 pushes: 50.6% overs. Scoring climbed about nine points a game over the eight seasons, and the totals climbed with it, so the split stayed close to even every year:

Consensus closing totals and final scores, NBA regular season. Pushes are left out of the over rate. Totals can land on a quarter point, so the line-size bands below each run up to the next one.

The final score averaged at or slightly above the line in every season, by up to 1.2 points — small enough that the over still split close to 50/50 every year. At the standard -110, either side needs 52.38% of decided bets to break even, and neither side reached it over the eight seasons as a whole.

The closing total is right on average and wide on any one night. A game missed its total by 14.4 points on average (median 12). Only 23.1% of games finished within five points of the line, 43.4% within ten, and 25.9% missed by more than twenty.

That is the nature of the market, not a flaw in the line: two teams trade around 100 possessions each, and a few points of shooting luck in either direction compound over four quarters. It also means a single result tells you almost nothing about whether a total was set well. A 20-point miss happens about one game in four.

The highest totals went under slightly more often than the lowest, but the spread from top to bottom is under two points across thousands of games — the kind of difference chance produces. A total is high because both teams score and play fast, and the line already knows it.

223 games played outside October to April (the 2020 restart in July and August, and May 2021) are left out of this table.

Six months sit within about two points of 50%. April does not: 431 overs to 346 unders, 55.5%, with a 95% range of roughly 52% to 59%. Six of the seven seasons that had April games went over more than half the time in April. In the six seasons whose regular season ended in April, it was 311-226 (57.9%).

A plausible reason is the end of the regular season: teams out of the race and teams resting starters for the playoffs tend to defend less. But be careful with it. We found April by looking at seven months after the fact, which is exactly how chance patterns get mistaken for rules. In 2020-21, when the season ran to mid-May, May went 68-64 (51.5%) — not the late-season bump the explanation predicts. And a pattern this visible is one the books can price. Treat it as something to watch next April, not a system.

We tested our own totals model the same way, on full seasons with the closing line. Three versions of it — the one we were running, one with its box-score inputs repaired, and one built on game results alone — each hit about 45% at the same confidence threshold in the 2025-26 season. Always betting the under went 50.9% on the same games. Last season's graded record, 54-45, came from a late-season and playoff stretch (March to June); over whole seasons the same approach ran 45% to 48%.

So from opening night of 2026-27 our NBA board shows a totals projection for every game with no badge, and it is never graded. It is there as a reference for the game's likely scoring range, not as a pick. Our graded NBA markets are on the NBA performance page, and the same exercise for spreads is in how often NBA favorites cover.