The Lab · The math · updated 2026-10-22
How many graded games until a win rate means anything?
A win rate needs thousands of games to mean much. Here's the arithmetic, our board's record as it has grown from 519 games, and the faster number to check while you wait.
Far more than you'd think. At 500 games a win rate can still be off by four points either way, and pinning it to one point takes nearly ten thousand. Calibration — does 60% happen 60% of the time? — answers in hundreds of games, and closing line value answers faster still.
Record figures marked as today's come live from the public record, refreshed as games are graded — as of Nov 17, 2026. Numbers we published on earlier dates stay as written, dated, so you can check them against today's.
On August 15 our board graded its 519th game. The favorite had won 286 of them, a 55.1% hit rate, every call published before the game and scored afterwards.
Five hundred sounds like a lot. It is the point where most people would say a record has proven something. So here is what 519 games actually bought — less than you'd think in one direction, more than it sounds in another — and, further down, what the board says now that it has grown.
What a win rate is worth at 519 games
A hit rate is an estimate, and every estimate has a width. For a percentage the standard
error is sqrt(p(1-p)/n), which at 55.1% over 519 games comes to 2.18 points.
Two standard errors either side is the usual bar for "we're fairly sure". That gives:
After 519 graded games, the true hit rate is somewhere between 50.8% and 59.4%.
Read that again, because it is the whole article. After more than five hundred published calls, the data could not distinguish "barely better than a coin flip" from "wins three out of five". Both were comfortably inside the range. Anyone quoting 55.1% as a fact was quoting the middle of an interval eight and a half points wide.
And it narrows slowly, because the n is under a square root. Halving that interval takes
four times the games. Cutting it to a single point takes about 9,500 — which, with only
baseball on the board in August at about fifteen games a day, was more than three seasons
away.
Here is the same arithmetic as a table. How many graded games before a calibration error of a given size becomes detectable at all:
| error to detect | games needed |
|---|---|
| 3 points | ~1,056 |
| 2 points | ~2,376 |
| 1 point | ~9,504 |
In August we were at 519 — not close to the top row.
It is worse for profit
Win rate at least has the decency to be a single number. Return on investment mixes in price, and prices vary, so the variance is larger still.
To establish that a strategy really returns +3% — a genuinely good edge, the kind a professional would take all day — from win/loss records alone takes roughly 4,300 bets at two standard errors, and nearly ten thousand at three. Not 500. Not 1,000.
This is why we say, repeatedly and to our own cost, that our published picks record proves nothing. In August it was 3 graded, 2–1, +97.9% ROI — arithmetically true and completely meaningless: three bets at plus money, two of which landed. Here is every figure we've published about it since, and today's:
| Published | Picks | Record | Avg CLV | Beat the close | ROI |
|---|---|---|---|---|---|
| Jul 27, 2026 | 1 | — | +16.43% | — | — |
| Aug 15, 2026 | 3 | 2–1 | +3.66% | — | +97.9% |
| Oct 21, 2026 | 14 | 6–8 | +6.58% | 64% | +55.6% |
| Today live | 15 | 7–8 | +6.52% | 67% | +56.5% |
Still nothing. The bar is thousands, not dozens. We report it because hiding it would be worse, not because it is evidence.
If a service shows you a 60% win rate over 50 picks, you now know exactly what to do with that number. The standard error on 50 picks is about 7 points. Their 60% is somewhere between 46% and 74%. They have not shown you anything, and if they are charging for it, the sample size is the tell.
So what does 500 games buy?
A different question, and a better one.
Instead of "is the model right?" — which needs thousands of games — ask "do the model's numbers mean what they say?" When it says 62%, does that group win about 62% of the time? That is calibration, and it is answerable much sooner, because every band is its own test.
On August 15, at 519 games:
| model said | games | actual | inside its own error bar? |
|---|---|---|---|
| 50–55% | 226 | 51.8% | yes |
| 55–60% | 178 | 54.5% | yes |
| 60–65% | 76 | 61.8% | yes |
| 65–70% | 27 | 63.0% | yes |
| 70–75% | 10 | 60.0% | yes |
Every band landed within one standard error of what it claimed. The ordering held too: higher confidence really did win more often, monotonically, across the 480 games in the first three rows.
Across the whole book the model claimed an average of 56.6% and delivered 55.1% — a miss of 1.5 points, which is 0.68 standard errors. Statistically invisible. The numbers meant what they said.
That was a real result. It was also not a claim that we beat anything, and the difference between those two sentences is the difference between this site and an advertisement.
Two months later
Since August the board has added hockey, basketball, football and, as a trial, college football. Today it has graded 2,079 games. The favorite has won 62.2% against 60.4% claimed — a gap of +1.8 points, which is +1.72 standard errors. The true rate is somewhere in 60.1%–64.4%.
Every figure we've published about the board, dated, beside today's:
| Published | Games | Favorite won | Market said | Gap (pts) |
|---|---|---|---|---|
| Jul 27, 2026 | 266 | 54.9% | 55.8% | −0.9 |
| Jul 28, 2026 | 311 | 54.7% | 55.9% | −1.2 |
| Aug 1, 2026 | 332 | 54.5% | 55.9% | −1.4 |
| Aug 15, 2026 | 519 | 55.1% | 56.6% | −1.5 |
| Oct 21, 2026 | 1,439 | 58.9% | 58.2% | +0.7 |
| Today live | 2,079 | 62.2% | 60.4% | +1.8 |
And the calibration table, every band, live:
| Market gave | Games | Price said | Favorite won |
|---|---|---|---|
| 50–55% | 678 | 52.4% | 52.2% |
| 55–60% | 554 | 57.3% | 61.2% |
| 60–65% | 355 | 62.2% | 62.3% |
| 65–70% | 185 | 66.9% | 69.2% |
| 70–75% | 136 | 72.1% | 75.7% |
| 75–80% | 66 | 77.4% | 84.8% |
| 80–85% | 52 | 82.6% | 90.4% |
| 85–90% | 39 | 87.1% | 82.1% |
| 90–95% | 12 | 91.6% | 100.0% |
| 95–100% | 2 | 95.8% | 100.0% |
The record grew, the interval narrowed, and the market kept its word at every date we published — the gap between what it claimed and what happened never left the noise. More games didn't reveal an edge. They confirmed, more precisely each time, that there wasn't one to find in the market's own numbers.
The part that stings
Being calibrated is not the same as being profitable, and we know precisely why ours isn't.
Our published probability is the de-vigged consensus of the sportsbooks. No signal has ever been promoted past shadow testing, so the board agrees with the market by construction. Being well calibrated to the market's own opinion is a fair description of "having no edge", stated politely.
We have spent this season trying to find one, and the trying is mostly a list of things that did not work. By August 15:
- A recent-form adjustment, carried over from an older version of this model. It turned out to predict where the price goes, not who wins — and adding it to our numbers made them slightly worse. The full story. Still shadow.
- A closeness offset, which assumed the market misprices near-even games. It doesn't. Across 178 coin-flip games the market was off by 0.1 points.
- A de-vig method bake-off — multiplicative against power against Shin. Near even money they differ by 0.30 points, against an effect we would need 15 points to matter.
- A price-neutral gap gate, meant to stop our screen filling with longshots. It selects the same longshots. Across 507 games the books never disagreed by more than 1.3 points, and the median gap is negative — the best price available is usually still worse than the field's own fair number. The full measurement is here.
Four hypotheses, four nulls. Since then the list has grown: seven popular price strategies tested on 26,000 games (what happened), and a stats model for player props that lost to the price by more than four standard errors (the props test). That is what the search for an edge actually looks like, and almost nobody publishes this part.
What to do with any record, including ours
Ask for the sample size before the percentage. It is the number that decides whether the percentage means anything, and it is the number least often volunteered.
Prefer calibration to win rate. "When they say 60%, does it happen 60% of the time?" is answerable in hundreds of games. "Do they beat the market?" needs thousands.
Grade probabilities, not picks. Comparing two forecasts game by game with the Brier score settles in hundreds of games what a win–loss record needs thousands for.
Treat closing line value as the faster instrument. Comparing the price you got against the market's final number is far less noisy than waiting for games to land — it is roughly 400 times more efficient than scoring against outcomes, which is why we grade every pick that way.
Be suspicious of a good number on a small sample, most of all when it is ours.
We publish the record because you should be able to check it, not because any sample has settled anything. In August, with baseball alone, the win rate looked set to become genuinely informative around 2030. With five sports grading, that's now a year or two away. The calibration table was telling the truth in August, and it still is.
Model output is informational and entertainment content, not betting or financial advice. If you bet, bet what you can afford to lose.
Log what you actually bet — units and dollars — and see it graded beside a model that publishes every call it makes, wins and losses alike. Free, no card.
Start a free tracker