How accurate is it?
Every scouting report is built from what the data says before a match. So we check that picture against what actually happens, match after match, using only data from before each one. No hindsight, no cherry-picking: these are all 2,038 Division II matches with box scores this season.
What wins D2 matches
Before we can tell a team what to do, we have to know what actually decides matches. From 2,295 Division II matches this season: how often the team that wins each stat wins the match.
| The team with… | Won the match | |
|---|---|---|
| Better hitting % | 91% | |
| More kills per set | 86% | |
| Better sideout % | 84% | |
| More point-scoring % (points won on our serve) | 84% | |
| More aces per set | 78% | |
| More digs per set | 78% | |
| Fewer attack errors per set | 75% | |
| More aces per service error | 68% | |
| More blocks per set | 67% | |
| More aces minus service errors | 66% | |
| Better passing (aced less often) | 66% | |
| Fewer service errors per set | 43% |
- Under .100: 3%
- .100 to .150: 18%
- .150 to .200: 46%
- .200 to .250: 70%
- .250 to .300: 88%
- .300 to up: 97%
- Winning 0–35% of points on your own serve: 11%
- Winning 35–40% of points on your own serve: 17%
- Winning 40–45% of points on your own serve: 42%
- Winning 45–50% of points on your own serve: 70%
- Winning 50–100% of points on your own serve: 85%
Point-scoring % (points won while serving) and sideout % (points won while receiving) are how the pros split the game. The better you score on your own serve, the more you win.
- More aces, more errors: 69%
- More aces, fewer errors: 64%
- Fewer aces, more errors: 34%
- Fewer aces, fewer errors: 29%
Compared with the D2 middle (1.6 aces, 2.0 errors per set). Low aces is the problem whether errors are high or low.
- Under -1: 37%
- -1 to 0: 44%
- 0 to 1: 54%
- +1 or better: 81%
Missing serves on its own doesn't lose matches: the team with fewer service errors wins only 43%, because tough serving brings aces. That's why our reports set serving targets on aces, not on errors.
Win chances you can trust
When we give a team a given chance to win, how often does it? A trustworthy model's 75% favorites win about 75% of the time.
| We said | Matches | Average chance | Favorite won | |
|---|---|---|---|---|
| 50–60% | 302 | 55% | 53% | |
| 60–70% | 280 | 65% | 63% | |
| 70–80% | 296 | 75% | 77% | |
| 80–90% | 412 | 85% | 83% | |
| 90–100% | 684 | 96% | 95% |
Gold: what we said. Purple: what happened. Our favorites win a little less often than we say at the top end (heavy favorites win about 93% of the time when we say 96%), so we treat those chances as slightly optimistic.
Is the picture of each team honest?
For each team in a match, both the scouted team and its opponent, we set a range for its key numbers that should contain the result 7 times in 10: a team's numbers swing from match to match, and a range that never misses is too wide to be useful. So the target is 70%, not 100%. A model that's about right lands near the gold line; ours lands at 67.8%.
| Range | Checked | Held | |
|---|---|---|---|
| Team hitting % | 3,131 | 68% | |
| Team sideout % | 2,961 | 68% | |
| Aces per set | 3,131 | 69% | |
| Service errors per set | 3,131 | 68% | |
| Reception error % (getting aced) | 3,129 | 68% | |
| Blocks per set | 3,131 | 66% | |
| Top attacker's share of swings | 2,917 | 66% | |
| Top attacker's hitting % | 2,794 | 67% |
Who gets the ball
| Call | Checked | Right |
|---|---|---|
| Top attacker leads her team in swings | 3,131 | 55% |
| Main setter leads her team in assists | 3,131 | 78% |
The top attacker leads in swings about half the time: a match usually spreads the ball among two or three hitters, so it's a lean, not a lock, and our reports treat it that way. Main setters lead in assists nearly 4 times in 5.
Locked in before the match
From October 3, 2026 on, the predictions for every upcoming D2 match are saved before the first serve and graded afterward: a record that can't be adjusted after the fact.
Do our targets move the needle?
Every report names the numbers that decide the match, with a target for each. They're levers, not predictions: the test is whether teams that hit them win more often than the pre-match odds said they would.
| Target | Met: won | (odds said) | Missed: won | (odds said) |
|---|---|---|---|---|
| Our hitting % | 100% 3 | 45% | 100% 3 | 73% |
| Hitting % we allow | 100% 3 | 59% | 100% 3 | 59% |
| Our sideout % | 100% 2 | 49% | 100% 1 | 43% |
| Our point-scoring % (on our serve) | 100% 3 | 65% | 100% 1 | 41% |
| Our aces | 100% 2 | 40% | 100% 1 | 57% |
29 targets from 6 reports so far, including a random weekly sample of full reports written for teams that aren't subscribers.
Every stat, what it measures and the research behind it: What wins volleyball matches ›
How we check
- Every prediction uses only matches played before it: the power rating is refit for each day, and each team's ranges come from its own earlier matches.
- Ranges are 70% prediction intervals from a team's match-to-match spread (five or more earlier matches). They're meant to miss 3 times in 10; that's what makes "held about 70%" the honest target.
- Results come from NCAA box scores and play-by-play. Rally share and sideout % need play-by-play, so a few matches without it aren't counted there.
- A random sample of full AI-written reports is also generated and graded each week, so the reports themselves, not just the numbers behind them, are measured.