Prediction track record

This page is the part of the site we cannot touch. Every tip that appears on a day page or a league page is written to a log before the match starts. At kick-off the entry is frozen. After the final whistle a script reads the 90-minute score and marks the tip as a hit or a miss. The tables below are built from that log and nothing else.

What is logged and when

A tip enters the log the moment it first appears on the site, with the match, the market, the selection, the probability and the fair odds at that moment. If the probability changes during the day, the log keeps the version that was live at kick-off. Tips that fall below the threshold before kick-off, and so disappear from the page, are removed from the log too, because they were never a live tip at the moment that mattered.

How settlement works

Settlement is mechanical. The final score after 90 minutes decides every market: match winner, double chance, over and under goal lines, both teams to score and correct score. Matches that are postponed or abandoned are void. Matches decided in extra time or on penalties are settled on the 90-minute score only, because that is what the tip was about. Accumulators are settled leg by leg and count as a hit only when every leg lands. Nothing is edited after settlement, and there is no process for appealing a result.

Last 30 days

Every tip is logged before kick-off and settled automatically from the final score. Nothing is edited after the match.

Last 90 days

Every tip is logged before kick-off and settled automatically from the final score. Nothing is edited after the match.

How to read a hit rate

A hit rate on its own tells you less than it seems to. Three things change what a number means:

• Sample size. Ten tips at 80% can land eight, ten, or five times without the model being wrong about any of them. Over a hundred tips the rate settles; over thirty it bounces. Read the 90-day table before drawing conclusions from the 30-day one.
• Average probability. A market full of short-priced favourites should hit more often than a market of draws. A 75% hit rate on 1X2 tips averaging 65% probability is good; a 75% hit rate on tips averaging 85% probability is not.
• Variance. Even a perfectly calibrated model has losing weeks. If the long-run rate sits close to the average probability of the tips, the model is doing its job, whatever the last seven days say.

The honest test of a prediction model is calibration: do 70% tips land about 70% of the time? That is what we are watching for in these tables, and it is what you should look for too. A hit rate well above the average probability over a long period would be surprising, and we would be as suspicious of it as you should be.

What the record does not show

The record shows whether the tips were right. It does not show whether following them was profitable, because that depends on the odds you got, your stakes and the bookmaker's margin, which the log does not know. It also does not include anything the site did not publish: there are no private tips, no premium list and no retroactive additions.

Thursday 1 October – Wednesday 7 October: our model analysed 5 matches across 2 competitions and 0 meet the threshold for this market.

The week ahead is listed on the day pages, starting with the prediction for today and the weekend prediction. The method behind every tip in this log is described on the methodology page. Any bet can lose, and this page exists to show how often ours did. The site is for adults aged 18+ only; see responsible gambling for limits and support.