Anyone can claim their predictions are accurate. We'd rather show you.
Every prediction we publish is tracked, recorded, and measured against the result. We don't delete the ones that miss. We don't quietly adjust the numbers. What you see on this page is the real, unfiltered performance of our prediction model.
Our prediction accuracy is measured across a rolling 30-day window and broken down by market:
These figures are updated after every matchday.
A 67% accuracy rate on 1X2 predictions means that roughly two out of every three match result picks land correctly. For context, bookmaker implied probabilities on favourites typically sit around 55 to 60%, so consistently outperforming that threshold demonstrates genuine predictive value.
Our Over/Under 2.5 accuracy at 71% is particularly strong, reflecting the model's ability to read attacking and defensive patterns effectively.
BTTS accuracy at 63% exceeds the baseline you'd get from random selection (around 50%), indicating the model successfully identifies matches where both teams are likely or unlikely to score.
For each market, we track:
We measure the primary pick per market. For 1X2, that means whichever outcome (home, draw, away) our model assigned the highest probability. For BTTS and Over/Under, it's the Yes/No or Over/Under call we published.
Accuracy varies by league. Leagues with more predictable outcomes (fewer upsets, stronger home advantage) tend to produce higher accuracy rates. We publish league-specific breakdowns on each league prediction page so you can see how the model performs in the competition you care about.
We've maintained consistent accuracy across multiple seasons. While individual matchdays fluctuate (variance is part of football), the rolling averages remain stable over time. This consistency is what separates a robust prediction model from one that gets lucky for a few weeks.
If our accuracy drops, you'll see it here. We don't cherry-pick good weeks or reset the counter after a bad run. The numbers on this page reflect every single prediction we've published in the measurement window. That's the deal.