Do Youth Baseball Rankings Predict Who Wins? We Tested It
The most common note we get is some version of "these rankings look odd." Fair enough. A rating can't be checked by looking at it, so rather than argue, we tested it. We hid the future from the rankings, let them make predictions, and counted how often they were right.
The short version (Fall 2026 baseball, ages 8U–14U). We rebuilt the rankings as they stood on October 4 and used them to pick the winner of 11,379 games played afterward. The higher-ranked team won 67% of the time. When two teams were far apart on the board, the favorite won 81% of the time. When two teams were within 25 spots of each other, it was a coin flip. The rankings aren't odd so much as they are honest about where the evidence runs out.
How the test works
The usual way to check a ranking is to look at the top ten and decide whether it feels right. That tests your intuition, not the ranking. A fair test needs one rule: the ranking can't see the games it is being graded on.
- Pick a cutoff date, say October 4.
- Rebuild the rankings using only games played on or before that date. Every later game is removed, as if the season stopped there.
- Take every game played after the cutoff between two teams the board already knew, in the same age group.
- For each game, the team the board ranked higher is the prediction. Count how often it won.
Ties are skipped. We used the same published formula on the same data you see on the site, with nothing tuned to these games. We repeated the test with four different cutoffs to make sure one date wasn't flattering us.
Result 1: the rankings beat chance, every time we checked
Same games scored three ways. "Our rating" is the composite on the board. The other two pick the team with the better win percentage or the better average run differential. Fall 2026 baseball, 8U–14U.
The ratings called 66% to 68% of later games correctly across all four cutoffs, and the number gets slightly better as more games come in. That is well above the 50% a coin would get. It also edges out the two obvious alternatives. On the October 4 cutoff, the rating beat the plain win-loss record by 1.7 percentage points (95% confidence interval: +1.0 to +2.3) and beat run differential by 1.1 points (+0.3 to +1.9). Those are small edges, and we won't oversell them. They are real, and they show that adjusting for schedule helps instead of hurting.
Why isn't it higher? Because youth baseball is a noisy game. In 27% of these games the margin was two runs or fewer, a result that a better team loses all the time. No rating, however good, gets to 90% when a fifth-grader's bad inning decides the game.
Result 2: the bigger the gap, the more reliable the pick
This is the check that matters most for "directionally correct." If the rankings capture real strength, then teams far apart should be easy to call, and teams close together should not. Here is accuracy sorted by the gap in rating between the two teams, on the 0–100 scale the board uses:
Fall 2026 baseball 8U–14U, board as of Oct 4, 11,379 later games. Bars show the share of games won by the higher-rated team. Whiskers are 95% confidence intervals.
It climbs the whole way. A gap under two points is a coin flip (52%). By five to ten points the favorite wins about two games in three. At 15 or more points it is better than four in five. A ranking that was noise wouldn't produce this staircase. Neither would one that was wrong in some systematic way.
Result 3: the same thing in ranks, which is how you see them
Rating points are abstract, so here is the same chart in the unit you see on the site: how many spots apart two teams are.
Spots apart on the age-group board as of Oct 4. Same games as above.
| If the teams are this far apart | Higher-ranked team won |
|---|---|
| Fewer than 25 spots | 51% |
| 25 to 50 spots | 55% |
| 50 to 100 spots | 58% |
| 100 to 200 spots | 66% |
| 200 to 400 spots | 73% |
| 400 or more spots | 81% |
This is the part we most want people to take away, because it explains most of the "odd" feeling. Teams within about 25 spots of each other are, as far as the data can tell, the same team. If #40 and #55 are in the wrong order, the ranking hasn't failed. Nothing, including the games themselves, can reliably tell those two apart yet. Rankings are a good guide to which neighborhood a team lives in, and a poor guide to who wins a given Saturday between neighbors.
Does it hold at every age?
| Age group | Games scored | Our rating | Win-loss record |
|---|---|---|---|
| 8U | 867 | 70.9% | 70.7% |
| 9U | 1,137 | 70.4% | 69.0% |
| 10U | 1,608 | 68.8% | 67.4% |
| 11U | 2,104 | 66.0% | 63.0% |
| 12U | 2,119 | 64.8% | 63.9% |
| 13U | 1,967 | 65.6% | 63.1% |
| 14U | 1,577 | 66.6% | 65.3% |
The rating is ahead of the raw record at every age. Accuracy is lower at the older ages, where talent gaps shrink and more teams are similar, and higher at the younger ones, where one strong team can run away from a weak one.
Where the rankings are weakest
- Early in the season. With few games, a team's rating leans on who it played, and that is itself uncertain. Accuracy was lowest on the earliest cutoff (65.7%) and rose as more games arrived.
- Teams in the same neighborhood. See the table above. This is a limit of the sport, not just the formula.
- Teams that don't play each other's opponents. If two groups of teams never cross paths, the board has to infer how they compare. Tier information (NCS, USSSA, Perfect Game) helps, and some mismatches remain.
- Games we can't see. The rankings only know about games in our data. Scrimmages and unreported games aren't there.
What this proves, and what it doesn't
It proves the rankings are directionally correct: a team ranked well above another wins more often, the advantage grows with the gap, and it holds across ages and dates. It does not prove the order is exactly right, and it can't. A pair of teams 20 spots apart has roughly a 51% chance of going the way the board says. If you're comparing teams that close, trust the schedule, the roster and the day over the number.
It also doesn't show the formula is the best one imaginable. It shows it beats the two obvious simple rules on out-of-sample games, which is the bar we can reasonably test. For how each input is weighted, see how rankings are calculated, and for why record alone is a weak guide, why win-loss record misleads.
About these numbers
All figures are Fall 2026 baseball, ages 8U–14U, from games between two teams the board already knew, scores decisive, same age group. The charts use the October 4 cutoff (11,379 games). Earlier and later cutoffs gave 4,125 (Sep 6), 12,080 (Sep 20) and 7,283 (Oct 18) games with accuracy of 65.7%, 66.8% and 67.8%. Later windows overlap with earlier ones, so they are not independent tests. Confidence intervals treat games as independent, which understates uncertainty slightly because the same team appears in many games. Ties and games between teams without a rating were excluded. To keep the comparison fair, pairs where the two teams had identical win percentages or run differentials were also dropped, for every method. We did not tune the formula to these games.
Frequently asked questions
How accurate are youth baseball rankings?
In Fall 2026 baseball for ages 8U to 14U, rankings rebuilt from games through October 4 picked the winner of 67% of the 11,379 games played afterward. Accuracy was 66% to 68% across four different cutoff dates. A coin flip gets 50%.
Why do the rankings look wrong to me?
Often because the two teams you're comparing are close. Teams within about 25 spots of each other were separated by the rankings only 51% of the time in our test, effectively a coin flip. Rankings are reliable about which neighborhood a team belongs in, and unreliable about the order inside it.
Are the rankings better than win-loss record?
Slightly, and measurably. On games after October 4 the rating picked the winner 1.7 percentage points more often than win percentage (95% interval +1.0 to +2.3) and 1.1 points more often than run differential (+0.3 to +1.9).
Why can't rankings predict more than about two games in three?
Youth baseball is a high-variance game. In 27% of the games we scored the final margin was two runs or fewer, and weaker teams beat stronger ones regularly. Even a perfect rating would not approach 90% accuracy.
How do you test a ranking without cheating?
By hiding the future. We rebuild the rankings using only games up to a cutoff date, then score them on games played after it. The rankings never see the games they are graded on.