Does defense vs position matter for NBA props?

Facing a more generous defense at his position (a higher rank number in this study, where rank 1 allows the fewest) was associated with a higher result relative to a player’s own average in our 2025-26 regular-season count. The pooled gap was largest for points and smallest for threes. It was not the same at every position.

That makes the matchup table useful context to investigate. It does not tell us whether a book has already accounted for that matchup in its line.

What changed against the highest-allowance defenses?

For every qualifying game, we subtracted the player’s average in his earlier games that season from his result. We then compared games against the most generous and stingiest allowance groups at his position.

Here are the results from the 2025-26 regular season, with measured games running from 29 November 2025 through 12 April 2026:

The gap is calculated before rounding the two means. These are averages across player-games, not an adjustment to add to every player’s next line. A player can appear many times, and several players can share one game.

Both groups finished above their prior averages on average. That matters: we required 20 minutes in the measured game, but the player’s baseline includes all his earlier appearances, including short ones. Changes in minutes and role can lift either group. Subtracting the player’s average does not remove those differences.

Above an average is not over a sportsbook line

The same 2025-26 comparison, counted as games strictly above the player’s prior average:

An exact tie with the prior average stays in the denominator but is not counted as above. For threes, that includes 148 games in the generous group and 172 in the stingiest group. An average is not a median, so even a generous group need not finish above its own average in half its games.

We hold no 2025-26 NBA sportsbook lines or prices. A player finishing with 24 points against his earlier average of 22 might still go under a posted 25.5. That is an illustration, not a measured bet from this sample. The percentages above cannot be used as prop hit rates or converted into a record of winning bets.

The difference was not uniform across positions

The 2025-26 gap between the generous and stingiest groups, again after subtracting each player’s prior average:

These are each stat’s own units. For example, the points gap was 1.94 for centers and 0.10 for small forwards, while the small-forward threes gap went the other way by 0.03 threes. The pooled result should not be read as the same adjustment for every position or player.

The player-game counts behind that 2025-26 breakdown are below, written as generous group / stingiest group, in points, rebounds, assists and threes order:

• Point guards: 403 / 424, 438 / 439, 416 / 400, 456 / 517.
• Shooting guards: 513 / 536, 531 / 624, 540 / 604, 587 / 568.
• Small forwards: 443 / 434, 459 / 496, 503 / 470, 505 / 506.
• Power forwards: 390 / 409, 417 / 426, 403 / 432, 469 / 476.
• Centers: 303 / 315, 302 / 319, 330 / 335, 369 / 426.

How we kept later results out of the ranks

We froze our recorded NBA logs, roster mapping and corrected defense-table calculation on 16 October 2026. The regular-season selection contains 26,640 player-games across 1,230 games, dated 21 October 2025 through 12 April 2026, with 82 games for every club. Play-in and playoff rows are excluded using the schedule’s season type, not just a date range. The 16 December 2025 NBA Cup championship is also excluded: unlike the earlier Cup rounds, its results do not count toward regular-season statistics. Its box score remains available for grading that game’s actual bets.

For each date, we did the following before adding any results from that date:

1. Add up what each defense had allowed to opposing players at each position on earlier dates. As on the Lab hub, two point guards in one game contribute to one combined position allowance, not two team games. The denominator counts games with a recorded player at that position.
2. Rank those allowances for points, rebounds, assists and threes separately. In this study, rank 1 means fewest allowed. Values are rounded to one decimal before ranking, and equal displayed allowances share a rank. The live NBA defense vs position table ranks from the other end: there, rank 1 allows the most.
3. Compare the five lowest and five highest allowance slots, including every club tied at either cutoff. There is no alphabetical choice between tied clubs. Depending on the date, position and stat, this made each group five to 12 clubs, not always five. The article’s tables use those tie-inclusive groups throughout.
4. Admit a player-game only after the opponent had 20 earlier team games, the player had 10 earlier appearances and he played at least 20 minutes in the measured game. The team pages’ separate display floor remains eight position-covered games. In the qualifying observations, the opponent’s position bucket held 16 to 81 earlier covered games.
5. Subtract the player’s average over all his earlier regular-season appearances, then count the result and whether it was strictly above that average. The current game’s result never enters its own baseline or rank.

The final-season calculation reproduces all 604 cells of the corrected Lab hub calculation, including values, ranks, tie counts and game counts. The historical test recalculates that same table before each date rather than borrowing its final ranks.

Which games were left out?

Only exact PG, SG, SF, PF and C labels qualify. Of the 26,640 regular-season player-games in the frozen 2025-26 selection, 24,305 had one of those labels. The rest were 218 generic-position rows and 2,117 without a position; we did not guess a more specific label for them.

Applying the rules in order gives this count. Each excluded player-game appears once, at its first failed rule:

Those 12,003 player-games came from 420 players. A game can belong to the generous group for one stat and the middle or stingiest group for another, which is why the stat-specific counts differ. Middle-group games are not in either extreme’s denominator.

What this cannot separate

The time split applies to the recorded results and averages, not to a historical roster archive. We used the 16 October 2026 position mapping that supplies the hub. It contains estimated positions: 5,988 of the selection’s regular-season player-games have labels derived from season-level assists or rebounds per minute. We did not rebuild those labels as they would have been known before every game.

We also did not hold minutes, injuries, teammates, pace, opponent schedule or role changes constant. The two groups contain different mixtures of players and games, and repeated appearances are not independent experiments. We have not established that a gap persists in another season or against posted lines. The cause is unknown.

The related L5 and L10 trend study explains another reason to distinguish a historical pattern from a betting rule.

Open the NBA defense vs position tables to compare the relevant stat and position with its season and game count attached. Check that context alongside the player’s workload and the actual line, rather than treating the rank as the answer by itself.