Skip to content

Win Rate Confidence

Your record shows one number. This shows the range of true win rates that could plausibly have produced it.

Optional — puts the break-even rate on the chart so you can see whether the interval clears it.

Pushes are excluded — enter only decided bets. The interval treats your bets as identical independent trials, which real betting is not: varying prices, varying edges and correlated same-day results all make the true uncertainty wider than this.

Your true win rate is plausibly

47.1% – 60.5%

The range straddles the 52.38% break-even rate, so this record is consistent with having an edge and with having none. That is where nearly every betting record sits, and more bets is the only thing that narrows it.

| observed▬ plausible range| break-even
Observed win rate
53.85%
Sample size
208

Decided bets. Pushes do not count.

Lower bound
47.06%
Upper bound
60.49%
Interval width
13.4%

Narrows with the square root of the sample — four times the bets to halve it.

Break-even rate
52.38%

What the price requires.

Runs entirely in your browser · nothing is sent or saved

How it works

A win rate is a measurement, and every measurement has error. Winning 112 of 208 gives an observed 53.85%, but a bettor whose true rate is 50% produces that record reasonably often, and so does one at 58%. This calculates the range of true rates consistent with what you observed.

It uses the Wilson score interval rather than the textbook normal approximation. The naive version fails in exactly the cases bettors run into: at 8 wins from 10 it produces an upper bound above 100%, and at 0 wins it collapses to a single point. Wilson stays inside 0–100% by construction and behaves properly on small samples, which is why it has been the statistical recommendation for decades.

Enter your typical odds and the break-even rate is drawn on the chart. That comparison is the whole point. If the entire interval sits above break-even, you have statistical evidence of an edge. If it straddles the line — which it almost always will — your record is consistent with having an edge and with having none, and no amount of staring at it will resolve that.

The intervals are wide, and that is the finding rather than a limitation. Two hundred bets is a full season for many people and it buys a range roughly fourteen points across. Narrowing it is slow: precision improves with the square root of the sample, so cutting the width in half takes four times the bets.

One honest caveat. The interval assumes identical independent trials, and betting is neither — prices vary, your edge varies, and bets on the same day share weather, injuries and market moves. All of that makes the true uncertainty wider than what is shown. Read this as a lower bound on how uncertain you should be.

The formula

n  = wins + losses
p̂  = wins / n
z  = the two-sided normal quantile
     (1.645 at 90%, 1.960 at 95%, 2.576 at 99%)

centre    = (p̂ + z²/2n) / (1 + z²/n)
halfWidth = z/(1 + z²/n) × √( p̂(1−p̂)/n + z²/4n² )

interval  = centre ± halfWidth

width shrinks with √n:
  4× the bets  →  half the width

Wilson stays inside [0,1] by construction, which is why it works on small samples where the normal approximation runs past 100%.

A worked example

112–96, an observed 53.85% over 208 decided bets, at 95% confidence.

The Wilson interval runs from 47.1% to 60.5%. Break-even at −110 is 52.38%, which sits comfortably inside that range.

So the honest reading is: this record is consistent with a strong edge, and equally consistent with no edge at all. A full season of betting has not settled the question — which is the normal outcome, not a sign that anything went wrong.

To get the lower bound above 52.38% at this observed rate you would need roughly a thousand bets. That is the real cost of proving an edge from results alone, and it is why closing line value is the more practical diagnostic.

Common questions

Why is my interval so wide?
Because betting samples are small relative to the effects being measured. Two hundred bets buys a range about fourteen points across, and narrowing it is slow — precision improves with the square root of the sample, so halving the width takes four times the bets.
What does it mean if the interval crosses break-even?
That your record cannot distinguish having an edge from not having one. It is not a negative result; it is an undecided one, and it is where nearly every real betting record sits.
Why Wilson and not the standard formula?
Because the standard normal approximation breaks on small samples and near the edges — it happily returns bounds above 100%. Wilson is constrained to [0,1] by construction and is the interval the literature has recommended for a long time.
Should I use 90%, 95% or 99%?
95% is the convention. Higher confidence buys a wider interval for the same data — you are asking to be more certain, so the range of possibilities you must admit grows.
Do pushes count?
No. Enter only decided bets. A push did not test your edge in either direction, so including it would dilute the sample rather than inform it.
Is there a faster way to know if I have an edge?
Closing line value. It gives a reading on every bet regardless of the result, so it converges far faster than win rate — a hundred bets of consistent positive CLV says more than a hundred winning bets.

The guide behind this calculator

Related calculators

For informational and analytical purposes only. These tools do not predict outcomes and do not recommend wagers.