What counts as a good ROI in sports betting
By George Boyle · Updated 2026-09-19 · The Sport Stack
For a bettor beating the market on their own, a sustained return on investment of 1–3% of amount staked is good and 5% is exceptional. Published figures far above that are almost always short samples, cherry-picked windows, or graded against prices that were not really available. Sample size matters more than the number: below roughly 500 bets, a return rate is mostly noise.
The honest benchmark
Professional sports bettors operate on thin margins. Beating a market that prices in roughly a 4.5% hold on standard -110 lines is hard, and the people who do it sustainably are typically making 1–3% on amount staked across thousands of bets. A long-run 5% is exceptional rather than typical, and anything advertised well beyond that should prompt a question about the sample and the grading, not admiration.
Note the denominator. Return on investment here means net profit divided by total amount staked — not divided by bankroll, which produces a much larger and much less comparable number. Two records quoting "ROI" can be measuring different things, and the one dividing by bankroll will always look better.
Why sample size dominates the number
A return rate computed over 20 or 50 bets carries an enormous confidence interval — wide enough that a genuinely losing process can print a large positive number and a genuinely winning one can look broken. This is not a technicality; it is the single most common way betting records mislead, and it usually happens without anyone intending to deceive.
Roughly 500 settled bets is where a return rate starts to carry real information about a 2–3% edge, and detecting a small edge with confidence takes more than that. Our own surfaces apply this rule to themselves: a per-sport record with fewer than 25 settled bets publishes its win-loss counts but deliberately does not publish a return rate, because at that size the rate would be a worse guide than no number at all.
What to check before believing a published record
Ask when the record starts and whether that start date moved. A record beginning at a convenient point is a different claim from one beginning when the system did. Ask whether losing plays are still visible, or whether only the graded winners survived. Ask what price each bet was recorded at — a record graded at the best price available anywhere, after the fact, is not a record of bets anyone could have placed.
Ask whether the rules were fixed in advance. A system whose staking or selection criteria changed mid-sample is being fitted to its own results, and its historical return describes the fitting rather than an edge. Pre-registration — writing the rules, stakes and shutdown conditions down before the first real bet — is the defence, and it is checkable by a reader in a way that a headline percentage is not.
How do you verify a betting model’s track record?
Run the claim through four questions, in order. Is the whole ledger visible, or only screenshots and highlights? Is every play recorded at the price it was struck at, before the event started, or graded afterwards at the best price anyone found? Are the losing plays still there? Were the rules — what counts as a play, how it is staked, when the system stops — written down before the record began, and have they changed since? A record that fails any one of these is describing its own presentation, not an edge. Watch for the softer signs too: a “unit” that is never defined, a win rate quoted without prices, a start date that moves, and a return above roughly 10% on fewer than a few hundred bets.
On this site the test is meant to be run against us. Every staked play appears on the public record at the price it was struck at and stays there whether it wins or loses; every per-sport summary states the sample it is drawn from and, where the record was floored, the date it starts; the grading rules for the MLB and NFL records are published on their methodology pages; and no return on investment is printed below 25 settled bets. Pre-registration, price of record and visible losers are the whole test, and any record — this one included — can be checked against them in a few minutes.
Common questions
How many bets do you need before ROI is meaningful?
Roughly 500 settled bets before a return rate carries real information about a 2–3% edge, and considerably more to distinguish a small edge from zero with confidence. Below about 25 bets a return rate is close to meaningless — the confidence interval spans both large profits and large losses — which is why records at that size are better read as win-loss counts than as a percentage.
Is a 50% win rate good in sports betting?
It depends entirely on the prices. At standard -110 odds you need roughly 52.4% to break even, so 50% is a losing record. On longshots at +200, a 50% win rate would be extraordinary. Win rate on its own says nothing without the odds attached, which is why return on amount staked is the more useful measure.
What ROI do sports betting tipsters claim, and why is it higher?
Advertised returns of 15–30% are common and almost never survive scrutiny. The usual explanations are a short sample, a start date chosen after the fact, removal of losing plays, or grading at prices that were not obtainable at the time of the pick. None of these require deliberate fraud — a small sample alone will produce a number that large often enough by chance.
How can you tell if a betting track record is real?
Look for the full ledger rather than highlights, the price each bet was recorded at and when, losing plays that are still visible, and rules that were fixed before the record started. A record that shows all four can be checked; one that shows only a percentage and a screenshot cannot, and should be read as marketing until it does.
Written by George Boyle, who builds The Sport Stack — the models, the public ledger and these explainers. Corrections and questions: hello@thesportstack.io. Who runs this.