Accuracy
How accurate is it?
Every July from 2019 to 2025 we rebuilt the board with today’s method, using only data from before that date, then checked what happened a year later.
×1.3our picks vs. similar players, a year later305 pairs · ahead in all 7 seasons (2 within the margin of error)
70.8%right on whether a value moves up or down by more than 5%saying “down” every time: 63.5% · 26,691 cases
67%of our picks rose in valuesimilar players: 48%
These are backtests, not live results.
Backtest 2019–2025
Season by season
Each July’s shortlist, compared a year later with similar players we didn’t pick: same league, position, age, minutes and price.
2019 · 40 pairs×1.391.16–1.68
2020 · 43 pairs×1.290.98–1.67
2021 · 33 pairs×1.231.04–1.48
2022 · 40 pairs×1.250.97–1.58
2023 · 43 pairs×1.381.10–1.73
2024 · 52 pairs×1.351.06–1.69
2025 · 54 pairs×1.421.10–1.84
All seven seasons · 305 pairs×1.31.22–1.46
Diamond: result. Bar: 95% confidence range. White line: ×1.0, no difference from similar players. In 2020 and 2022 the bar crosses ×1.0.
Limits
Where it’s weaker
- Players above €20mAs a price estimate, it does worse than assuming the price stays put.
- The top five leaguesOn direction, it’s no better than assuming every value goes down.
- Players 21 and underThe price estimate does worse than assuming the price stays put.
- GoalkeepersThe performance rating barely works for them.
- 89 leagueshave no accuracy test of their own yet; 66 of them aren’t in the backtest at all.
From August 2027
Live record
Dated shortlists go back to 26 August 2026; since 7 October 2026 each one is saved with a timestamp. Grading starts on 26 August 2027, and the results go on this page, good or bad.
Download the backtest results (CSV)See it on your own squad.
A free two-page report on your club, before any call.