How to Verify a Sports Picks Site's Record
1. Was every pick locked before the start?
A pick only counts if it was on record before the game or fight began. Look for a lock time on every pick, next to the start time, and for a record of whether a pick was changed before the start. A pick that can be edited, posted after kickoff or quietly dropped tells you nothing. Screenshots and chat posts are hard to check because they can be chosen after the fact.
2. Is every pick in the record, misses included?
Ask for the whole list, one row per pick, in a form you can download and count yourself. The number of rows should match the headline record. Watch for a start date that moves, records shown only for a hot stretch, separate records for each sport that leave the weak ones out, and picks labelled differently after they lose.
Backtests (picks a model scores after the fact on past games) can be useful, but they belong in a separate, labelled table and never in the live record.
3. Compare it with picking the favorite
A high winner rate is easy to reach by taking favorites: in most sports the favorite wins well over half the time. The fair question is how the picks did against the betting favorite on the same games. If the two records are about the same, the picks are mostly repeating the market. That can still be a useful, honest probability, but it is not evidence of an edge.
4. A winner rate is not a profit figure
A winner rate leaves out the price. Favorites pay less than even money, so a record can win most of its games and still lose money.
- Risk to win 100
- 250 to win 100
- Break-even win rate (250 / 350)
- 71.4%
- Winning 70% of such picks
- loses money
Any profit or units figure is only checkable if each pick carries the price that was available when it was locked. Without the prices, a units number cannot be verified.
5. Is the sample big enough, and are the percentages honest?
Short records swing a lot. A range around the win rate shows how much.
Ten picks say almost nothing; a few hundred start to say something.
If a site gives a probability with each pick, check calibration: of the picks it called about 70%, did about 70% win? A score such as log loss or the Brier score, set beside the market's own on the same games, is the fairest single test of the probabilities.
The checklist, and where to find each answer on Probetrics
The same tests apply to Probetrics. Its records sit close to the betting favorite's, because since late September 2026 a priced game's winner call is built on the market's no-vig price; the results page says so in its own numbers.
See it on tonight's props. Pro puts each player's hit rate beside the books' chance on every line.
Start 7-day free trialQuestions and answers
- How do I know a picks record is real?
- Check that every pick has a lock time before its start, that the full list can be downloaded and counted, and how often the picks simply took the betting favorite.
- Is a 70% win rate good?
- Not on its own. Favorites often win that often, and at a price of -250 a pick has to win 71.4% just to break even. Compare it with the favorite's record and the prices involved.
- Why do units or profit claims need prices?
- Profit depends on the price each pick was made at. Without the price recorded at lock time, a units figure cannot be checked.
- Can I check Probetrics' record myself?
- Yes. The public ledger lists every counted pick with its lock time, start time, probability, result and whether it was the favorite, and can be downloaded as CSV or JSON.
Read next
- CalibrationCalibration is whether stated probabilities match what happens: of the picks a model gave about 70%, about 70% should win.
- Hit rateA hit rate is the share of a player's games that finished on one side of a prop line, and over a small sample it is a wide range rather than a single number, which a Wilson interval makes visible.
- Market-anchored probabilityA market-anchored probability starts from the sportsbook's no-vig price for a game and lets a model change it only where testing has shown the model adds information.
- Walk-forward backtestingWalk-forward backtesting scores a model on games it has not seen by pricing each game only from what was known before it, moving forward through time.
The public ledger →Ledger as CSV →Track record →Disagreements with the favorite →How the models are tested →All terms →
This page explains arithmetic and vocabulary. It does not recommend a bet, a side, an amount or a sportsbook, and the examples are illustrations. Data and analytics only, not betting advice: any bet is your own decision, and we are not responsible for losses. 21+ only. Gambling problem? Call or text 1-800-GAMBLER.