The public record
Performance
Every signal we published, settled against real results. Headline figures reflect backed signals only — the one band we suggested. The wider edges, the longshots, and everything backed before 6 August 2026 are shown below for information and are not tracked.
Read this before the numbers
An audit on 5 August 2026 found that three of the five architectures behind this record were not forecasting anything — one reproduced the market’s own de-vigged price, one was a five-value lookup keyed to price bands, and one was consensus odds put back through a Poisson. All three have been switched off and no longer publish. The figures below are therefore a record of what we published, not a claim about what we publish now: they are dominated by the withdrawn market-consensus signal, which returned a yield of −47.0% and closing-line value of −24.9% over 193 settled selections. They stay up because deleting them would be the same mistake in the other direction. The one model still publishing is the Dixon-Coles goals sheet; what a model has to prove sets out the bar anything else has to clear to come back.
- Win rate
- —
- Yield
- —
- ROI
- —
- Adj CLV
- —
- 0/30 settled — needs 30 to mean anything
- Signals
- —
Signal distribution by edge
Every signal detected to date. The backed band — 4–10% edge at odds 1.40–3.00 — is the only one we suggest and the only one counted above.
Methodology & definitions
- Sample evaluated
- Every signal recorded from engine launch to date. One signal is one detected value opportunity: a single outcome where our model assigned a higher probability than the bookmaker’s implied odds.
- Win / loss
- A signal is a win if the recommended outcome occurred at match completion, a loss if it did not. Void and push outcomes — cancellations, postponements — are excluded from win-rate and yield entirely.
- Yield
- Total profit ÷ total staked, at flat 1-unit stakes on every signal. Yield is the primary measure of long-term profitability; positive yield is rare and meaningful.
- Adjusted CLV
- The price improvement between the odds at detection and the final closing odds, with the margin removed. Stored as the log-ratio ln(signal ÷ closing); the percentage shown is e^clv − 1. Sustained positive CLV is the strongest available proof of a real edge — more reliable than a short-run win rate.
Past performance does not guarantee future results. This data is for informational and model-evaluation purposes only. Not financial advice. 18+, please gamble responsibly.
Signal history
Unbacked edges & longshots
These are not in the headline win rate, yield or ROI above — that scoreline is backed signals exclusively. They are published so you can see how the wider, non-suggested edges are landing.
No edge results yet — they’ll appear here as picks settle.