Prediction Accuracy
The scoreboard for the model itself: accuracy by competition and market, monthly breakdowns, and tracked value-bet ROI and closing line value.
Open Prediction Accuracy
This page grades the model in public, and it does so in two different ways. The Value Bets tab tracks what flagged value selections actually returned. The Overall Market tab ranks competitions by how often the highest-probability prediction turned out to be correct.
Value Bet Performance
This is the tab the page opens on, because it is the one that answers the money question rather than the accuracy question.
| Figure | What it tells you |
|---|---|
| Flagged bets | How many strict value selections are tracked in this window |
| Win rate | Wins divided by settled tracked bets |
| ROI | Profit in units per settled tracked bet |
| Avg edge | Model probability minus the market implied probability at signal time |
| Avg CLV | Average closing line value on the bets with a tracked closing price |
| Positive CLV rate | The share of tracked bets that beat the close |
Performance Trends underneath plots the cumulative ROI curve and the cumulative hit rate over seven days, thirty days, or all time, and a CLV audit sets out the data checks behind the published figures, including late signals that are tracked but excluded from the headline numbers.
Why CLV appears next to ROI
Over short windows, results are dominated by variance. Consistently beating the closing price is the earlier signal that an edge is real, which is why both are published side by side.
Overall Market accuracy
The second tab is the league ranking. It answers a narrower question: how often was the model’s highest-probability call the outcome that happened?
What accuracy means here
Accuracy is the share of predictions where the highest-probability outcome was the one that happened. It is not a profit figure: a market can be highly predictable and still unprofitable if the prices are short.
The ranking controls
- Market
- Which market to grade: match result, BTTS, total goals, the individual goal lines, double chance, draw no bet, half-time markets, half-time/full-time, cards, and corners.
- Line
- For over/under markets, which exact line to grade, so a 2.5 result is not blended with a 3.5 result.
- Period
- Season total or a single month, which is how you find out whether a league’s number is holding up or drifting.
- Season
- Which season the table is drawn from.
- Outcome filter
- On match result, restrict the table to home, draw, or away calls only.
Reading the table
Each row is a competition with its accuracy, the number of correct calls, and the sample size behind it. Clicking a row expands a monthly breakdown for that competition. The colour legend runs from excellent through good, average, and below average to poor, and the sample column is there so a small-sample spike does not read as a strong league.
Always read the sample size
A league at the top of the table with a few dozen graded fixtures is noise. Sort your attention by sample first, then accuracy.
Related guides
- Value BetsSelections where the model probability exceeds the probability implied by the price, filtered so recent trends support the same market.
- Odds and valueThe arithmetic behind every tool on the site: converting a price into a probability, spotting the bookmaker margin, and knowing what an edge does and does not promise.
- Bet TrackerLog every bet with its price, stake, and closing odds to get honest ROI, win rate, profit by market, and average closing line value.