Published record
Our Prediction Accuracy, Published in Full
Every settled prediction across ten sports, won and lost, counted from the same results that mark the fixture tables, with nothing selected and nothing removed.
869 predictions have settled in the last 30 days, 869 won and 292 lost, a strike rate of 66.4%. A percentage is printed once 20 calls have settled in the window. Until then the counts are shown as they are, and every settled call stays on its fixture page next to the result.
- 1,639 predictions published across ten sports since launch, and the counter moves with the fixture list rather than with us.
- 869 settled in the last 30 days, 577 won and 292 lost.
- Nothing is removed. A prediction that loses stays published next to the result.
- No paid tier, so there is no separate record you are not seeing.
- We do not claim sure predictions. A probability that was never wrong would not be a probability.
What Does the Headline Figure Cover?
Every settled prediction in the window, not a selection.
Not a best run, not a featured tip record, not a figure from a good month. When a call loses it stays on its fixture page and it is counted here. The window is 30 days because a single day is too small a sample to mean anything: on a twelve fixture Tuesday, two results either way moves the figure seventeen points.
How Accurate Is Each Sport?
Where the model is doing well, and where it is not.
Every sport name links to its predictions. A sport showing too few has settled fewer than 20 calls in the window, which is not enough to state a percentage from. An empty row would read as a zero, and a zero is a result. Nothing is not. Football carries most of the volume, so the all sports figure is mostly a football figure; read the smaller sports on their own rows.
How Do I Check a Prediction Myself?
You do not have to take this page's word for it. Three steps.
- Open any settled fixture. Yesterday's card on football predictions, or any sport hub, lists fixtures that have finished with the final score next to the call we made.
- Compare the tip to the result. The tip names its market, so there is no guessing which outcome it was on. A won call shows a green mark, a lost call a red one, and the probability we published sits beside both.
- Add it up for yourself. This page is nothing more than that count across every settled fixture in the window. If you find a settled fixture with no mark, or a mark that disagrees with the score, tell us through the contact page.
How Should a Strike Rate Be Read?
The number on its own is close to meaningless. Here is why.
What Do We Not Claim?
Stated plainly, because the category is full of the opposite.
About Our Prediction Accuracy
What we measure, how it is calculated, and what it cannot tell you
Our prediction accuracy is 66.4% over the last 30 days, across 869 settled calls. That figure covers every settled prediction in the window rather than a chosen stretch. Each call is marked won or lost on its own fixture page the moment the result is in, and this page is the sum of those marks. Nothing is excluded, and the window does not move to flatter a bad week.
How the Calculation Works
Every prediction carries one tip. When the fixture settles, the tip is compared to the result and marked won or lost. There is no void, no push and no half win in this count: a handicap that lands exactly on the line is rare in the sports we cover and is recorded as it finished. 577 won and 292 lost of 869 settled is the whole arithmetic. You can check any individual call by opening its fixture page, where the prediction sits next to the final score.
Why a Rolling Window Rather Than a Daily Figure
Because a daily strike rate is noise. On a twelve fixture card two results either way swings it seventeen points, so a daily number would spend most of its life either flattering us or unfairly damning us, and neither tells you anything about the model. A 30 day window is long enough to be stable and short enough to reflect how the model is performing now rather than how it performed last season. The all time figure, 66.4%, sits alongside it for the longer view.
Why a Prediction Site Should Publish Losses
Because a record you cannot audit is a marketing claim. A site that shows only its winners is showing you a filtered sample, and the filter is doing all the work. Keeping the losses in place is the only way the number means anything, and it is the reason the figure here is n/a rather than the ninety plus that this category routinely advertises.
Accuracy Is Not the Same as Value
A tipster who only backs heavy favourites can post a very high strike rate and still lose money, because short odds return little when they win and cost everything when they do not. A lower strike rate at longer odds can be the better model. We publish the win probability on every call precisely so this is visible: a 55 percent call and an 85 percent call are not the same bet, and a record that averages them together hides the difference.
Where the Numbers Come From
Results are taken from the same API-Sports feed that supplies our fixtures, across all ten sports we rate. The settlement is automatic rather than manual, which removes the opportunity to quietly not settle something that went badly. The method is on the methodology page, the people who check this page are on the team page, the rules they work to are in the editorial policy, and the current card is on the all sports hub.
Frequently Asked Questions
What people ask about a published accuracy record.