Trader Size › Correct score

Correct score predictions with artificial intelligence

Trader Size estimates the correct score of every match with five machine learning models trained on the history of the teams and publishes the distribution of the most likely scores. The real hit rate is measured every day and published on this page, compared with a simple baseline.

Correct score is the final result with the number of goals scored by each team, for example 2-1. It is the hardest market to predict: even the most common score in a league rarely exceeds 15% of matches.

How it works

  1. We gather recent results, goals scored and conceded, and each team's home and away performance.
  2. Random Forest, Neural Network, Gradient Boosting, XGBoost and LightGBM each estimate the probability of every score.
  3. The consensus follows the match result (1X2) picked by at least 3 of the 5 models and, among those, the most-voted score; with no majority, the result with the highest combined probability across the five models applies, together with the most-voted score across all five. We also show the five most likely scores in the models' combined probabilities.
  4. After the match, we compare the prediction with the official score and update the rates.

Real numbers

The numbers are being updated. Please check back in a few hours.

Limitations

Predictions are statistical estimates, not guarantees. Past results do not ensure future results. Trader Size is not a betting site and does not accept bets. Content for people aged 18 and over.

Frequently asked questions

What is the hit rate for the correct score?

The real rate for the last 90 days is in the table on this page, next to the baseline. It is recalculated every day.

Why show the 5 most likely scores?

Because the correct score has many possible combinations. Knowing whether the result landed among the five most likely shows whether the model read the match well.

Is the score prediction free?

The page for each match is free and shows the general reading. The full score distribution by model is exclusive to subscribers.

WhatsApp