1. Home
  2. Learn
  3. Our Accuracy
Transparency

Our Prediction Accuracy

We believe in transparency. Here's how our pG predictions actually perform against real match results.

4 minute read

How We Measure Accuracy

Since we predict goals (not bet outcomes), we measure accuracy using Mean Absolute Error (MAE) — the average difference between our predicted goals and actual goals scored.

Understanding MAE

Example Calculation
Predicted
2.1
vs
Actual
2
=
Error
0.1

We calculate this error for every prediction, then average across all matches. Lower MAE = more accurate predictions.

Why MAE Instead of Win/Loss %?

We predict goals, not bet outcomes. A "correct" prediction where we said 1.8 pG and the team scored 2 is more meaningful than simply tracking whether a bet would have won. MAE measures the quality of our goal predictions directly.

Current Performance

Based on walk-forward validation across 15,000+ matches in Europe's top 5 leagues:

Home Goals MAE
0.71
Average error in predicted home goals
Away Goals MAE
0.64
Average error in predicted away goals

What These Numbers Mean

Home MAE of 0.71 means our home goal predictions are typically within 0.71 goals of the actual result.

If we predict 1.8 home goals, actual results average between 1.1 and 2.5

Away MAE of 0.64 means our away predictions are slightly more accurate than home predictions.

Away goals are generally lower and less variable, making them easier to predict

Putting These Numbers in Context

Better Than Naive Baselines

Simply predicting league averages (e.g., "every home team scores 1.5 goals") produces MAE around 1.0+. Our model significantly outperforms this baseline.

Validated on Unseen Data

These numbers come from walk-forward validation — predicting matches the model hadn't seen during training. No data leakage or overfitting.

Football Has Inherent Variance

Even a "perfect" model can't predict exact goals. A team with 2.0 xG might score 0-4 goals. Some error is unavoidable — football is chaotic.

MAE Reference Scale

0.5-0.7
Excellent
0.7-0.9
Good
0.9-1.1
Average
1.1+
Poor

Our current MAE of ~0.67-0.71 places us in the "Good" to "Excellent" range

Verify It Yourself

We don't ask you to trust us blindly. You can verify our accuracy:

Results Page

Our Results page shows predictions alongside actual outcomes. You can see exactly what we predicted vs. what happened for every match.

View Results

Our Commitment

We publish all predictions before matches and never retroactively change them. What you see is what we predicted. This transparency is fundamental to how we operate.

Honest Limitations

Our model performs well on average, but every prediction system has limitations:

Individual Match Variance

MAE is an average. Individual predictions can be off by more. A match where we predicted 2.0 pG might end with 0 or 4 goals.

No Model is Perfect

We don't capture player injuries, tactical changes, or motivation factors. These can cause predictions to miss significantly on specific matches.

Performance Can Vary

Accuracy varies by league, team type, and time of season. Early season predictions (less data) may be less accurate than mid-season.

Key Takeaways

  • We measure with MAE — average error between predicted and actual goals
  • Current MAE: ~0.71 (home), ~0.64 (away) — predictions typically within 0.7 goals of actual
  • Validated honestly — walk-forward testing on unseen data, no cheating
  • Verifiable — check our Results page to see predictions vs. actuals
  • Honest about limits — individual matches vary, no model is perfect

Related Pages