Best published result: the German Football Association (DFB) talent programme, Höner, Leyhr & Kelava, PLoS ONE, 2017 (14,178 players tested at under 12, followed to adult level).
Evidence
Checked against where players actually ended up.
Game1’s forecast is built on 10+ years of test days and checked against where the players ended up.
The result
More accurate than the best published result.
How this was validated
- In-house developed AI models, trained on outcomes.Predicting with custom-built neural nets, where players actually end up.
- A proprietary data set.Years of standardized youth test days, linked to adult careers. The largest of its kind.
- Rigorously backtested.Backtested out of sample, on data the model has not seen.
8 in 10
future professionals in the top quarter of the ranking.
Line chart. Looking at the top 25% of Game1’s ranking finds 80% of the future professionals; the top 34% finds 90%. A random order, the diagonal line of chance, would find 25% and 34%.
The two marked points are the measured results; the line joins them. The diagonal is a random order.
The method, written down
The figures on this page come from Game1’s 2026 model and are newer than those in the November 2025 paper.
-
Prediction model · November 2025
Predicting Elite Youth Athlete Outcomes from Standardized Testing Data
How the prediction model was built and tested.
whitepaper.game1.ai/potential (opens in a new tab) -
Mental strength
The psychology paper
How mental strength is measured, and how the results become training.
whitepaper.game1.ai/psychology (opens in a new tab)