Overall record
Model v3 — current rules, since 18 Sep 2026
The v2 record settled at −6.1% ROI over 933 graded picks — market-average results minus the bookmaker margin, which the published closing-line value (−0.6%) confirmed. Its own breakdown showed where the money went: Over-goals picks (−18% ROI), picks the model rated at 50–59% (they hit 44%), and the opening weeks of a season, when ratings still reflect last year. From 18 Sep 2026 the same model publishes only picks it rates at 55% or more, never an Over, and nothing until both teams have four recent matches. Roughly a quarter of the old volume.
Model v2 — 29 Jul to 17 Sep 2026, closed
Selective picks from our Poisson goal model blended with the market, published whenever the best price cleared a value threshold. It lost, and the numbers stay here as published — a record that only keeps its winning periods is not a record. CLV: -0.48% average, beating the close on 254 of 867 bets.
v3 tracks by market
Totals model
Draw value model
These are performance tracks of one and the same model, split by market — not people. Every number counts real settled tips only: archived before kick-off, graded against the final score, never edited. Small samples are labelled as such.
v2 tracks by market
Draw value model
Match Result model
Totals model
These are performance tracks of one and the same model, split by market — not people. Every number counts real settled tips only: archived before kick-off, graded against the final score, never edited. Small samples are labelled as such.
All-time tracks (incl. v1)
Draw value model
Match Result model
Totals model
These are performance tracks of one and the same model, split by market — not people. Every number counts real settled tips only: archived before kick-off, graded against the final score, never edited. Small samples are labelled as such.
Settled log
By date →Showing the latest 40 of 1234 settled tips — the rest live on the daily archive pages.
How to read this
Every tip is archived with its odds the moment it is published, then settled automatically against the official final score. Win rate, ROI and profit are computed from those graded bets only — nothing is edited or deleted afterwards.
A slice of the same model by market type: match-result picks, totals picks, draw value. Tracks are strategies, not people — we don't invent tipster personas.
Closing-line value compares the price we took with the last price we tracked before kick-off. Beating the close consistently is the standard early sign of a real edge — it converges much faster than ROI. Our snapshots are taken at each daily refresh, so this is the last tracked price rather than the true market close; we label it honestly.
Yes, and when it is, the page shows it. Betting outcomes swing hard on small samples — that's variance, and it's exactly why the full log stays public while the sample grows.
About this page
This page shows the honest, by-market track record of our own prediction model, broken down so you can see how it performs across match result, totals, and draw value markets separately. There are no tipster personas here, just data from a single model and a fully settled public log you can browse in our tipsters section. Every tip we publish is archived and settled openly, win or lose, which is the core trust promise behind this site. To get the most from it, focus on the markets where the model has shown consistent value rather than treating all picks the same, and compare its reasoning against the current odds. Remember that past performance does not predict future results, so stake within a budget you are comfortable with.
Performance FAQ
Every tip the model publishes is logged and settled openly, whether it wins or loses, so nothing is quietly deleted. You can review the full history in our public log and see performance split by market, including match result, totals, and draw value.
Performance varies by market, which is exactly why we separate match result, totals, and draw value on this page rather than showing a single blended figure. Check the by-market breakdown to see where the model has been most consistent, and keep in mind that historical results are no guarantee of future outcomes.
Betting always carries risk and no model can promise a profit, since results depend on uncertain outcomes. Our aim is to show transparent, settled records so you can judge value for yourself, then decide with a bankroll you can afford to lose.