How We Grade Every Pick
Every finished pick is graded automatically and in public, wins and losses alike. This page explains the rules and shows the current numbers.
The rules
- Published before kick-off. Only picks published before the game starts and never edited after it are counted.
- A real price. A pick is counted only when a real sportsbook quote was stored with it. Picks without a market price are listed as excluded, not hidden.
- Graded against the official result. When the game finishes, the pick is graded automatically against the final score from our data provider.
- Wins, losses and pushes all shown. Picks on markets that cannot be graded automatically are ungraded, never wins.
- Flat stakes. Profit and ROI assume a flat 1-unit stake at the sportsbook price published with the pick.
The full text of the rules is in the methodology below. The results are on the track record.
The numbers, last 90 days
- Record (W-L-P)
- 1881-1409-2
- 3292 graded picks
- Hit rate
- 57.2%
- wins / (wins + losses)
- Profit
- -48.28u
- flat 1-unit stake
- ROI
- -1.5%
- flat 1-unit stake
Closing line value
Of 3277 graded picks with a closing price, 25.6% were priced better than the closing line, and the average closing line value was -1.4%. Closing line value compares the price of a pick with the last price before kick-off. It measures price, not profit.
At the best price
Scored at the best sportsbook price shown with each pick, 1773 picks returned +0.29u, an ROI of +0.0%.
Headline picks, those that passed the publish rule
- Record (W-L-P)
- 0-0-0
- 0 graded picks
- Hit rate
- —
- wins / (wins + losses)
- Profit
- 0.00u
- flat 1-unit stake
- ROI
- —
- flat 1-unit stake
Calibration
Calibration checks whether a stated confidence means what it says. Graded picks are grouped by confidence, and the share that actually won is shown next to the win rate the sportsbook price implied. 3548 graded picks are counted over the last 365 days.
| Confidence | Picks | Actually won | Price implied |
|---|---|---|---|
| 0-9% | 189 | 46.0% | 51.3% |
| 10-19% | 8 | 0.0% | 6.7% |
| 20-29% | 144 | 13.2% | 11.6% |
| 30-39% | 191 | 25.7% | 26.4% |
| 40-49% | 486 | 38.1% | 40.5% |
| 50-59% | 833 | 49.3% | 50.8% |
| 60-69% | 778 | 60.4% | 63.1% |
| 70-79% | 507 | 75.3% | 75.2% |
| 80-89% | 408 | 87.3% | 87.7% |
Methodology
Every SprtGenie primary pick for a game that has finished is graded automatically against the official final score from our data provider. Only picks that were published before kick-off and never edited after it are counted, and only picks priced at a real sportsbook quote stored with the prediction (picks without a market price are listed as excluded). Wins, losses and pushes are all shown; picks on markets that cannot be graded automatically (e.g. player props, MMA round totals, tennis game handicaps, retired tennis matches) are counted as ungraded, never as wins. Hit rate = wins / (wins + losses). Profit and ROI assume a flat 1-unit stake at that sportsbook price. Closing-line value (CLV) compares each graded pick's price with the closing no-vig consensus price of the same side from our data provider (spreads and totals only at the same line): CLV = pick decimal odds / closing decimal odds - 1; picks without a matching closing price are left out of it. Best price: the same picks at the best available bookmaker price shown with each pick (line shopping), counted only for picks published on fresh odds. Headline picks: each new primary pick is checked once, when it is generated before kick-off, and the result is stored with the pick and never changed after the game. The date shown is the earliest checked pick in this period. Picks that fail are left out of the upcoming picks, match pages, picks emails, texts and push notifications. Picks generated before the check existed were never checked and can still appear, and lists of past results, including the recent picks in this record, show every graded pick. A pick passes only if it backs one side to win (no totals, spreads, both-teams-to-score or draws) and is priced at a real sportsbook quote longer than -1000. When a no-vig fair price for the same side is available, the price must also be within 25% of it; when none is available, only the real-quote check applies. In three-way markets the home, draw and away prices must all be present and add up to a normal bookmaker margin. The figures above cover every pick; the headline block repeats them for the picks that passed. The rule removes picks we could not stand behind at the posted price; it is not a claim that the remaining picks beat the market. Past results do not guarantee future outcomes.
Questions
Are losing picks included in the record?
Yes. Every graded pick counts, wins, losses and pushes. Picks on markets that cannot be graded automatically are shown as ungraded and are never counted as wins.
Can a pick be edited after the game starts?
No pick that was edited after kick-off is counted. Only picks published before kick-off and never edited after it are graded.
What is the headline publish rule?
Each new pick is checked once, when it is generated before kick-off, and the result of the check is stored with the pick and never changed after the game. Picks that fail are left out of the upcoming picks, match pages, emails, texts and push notifications. The rule is not a claim that the remaining picks beat the market.
Get SprtGenie
Snap any live game in the app to see the AI pick, the best price across sportsbooks and the fair price. Informational only, no outcome guaranteed. 18+.
Picks are AI-generated research, not betting advice, and past results do not guarantee future outcomes. SprtGenie is an informational research tool, not a sportsbook. 18+ only. If gambling is causing you problems, call the National Council on Problem Gambling at 1-800-522-4700.