Data inputs
Each match prediction starts from a structured snapshot built at SSG time:
- Match metadata: start time, venue, surface, tournament, round.
- Recent form for each player (last 5 matches, weighted by recency).
- Head-to-head record between the two players over their last meetings.
- Surface split — how each player performs on hard, clay, and grass.
- Probability distributions over the match winner, total games, and over/under markets.
For background on this metric, see Elo rating.
Models
Predictions are produced by an ensemble of large language models that consume the structured inputs above and output a calibrated probability distribution plus a short reasoning paragraph.
No single model decides a prediction. The pipeline aggregates outputs across models and falls back to the best-calibrated source if a candidate disagrees beyond a configured threshold. For the underlying probability model, see probability theory.
Tournament scoring
When you submit predictions in tournaments your score is computed deterministically:
- 1 point for a correct outcome (player 1 win / player 2 win).
- 3 points for the exact final score.
Calibration
Prediction confidence is calibrated against historical results, not against the model's internal certainty. A 60% win probability means the model is right roughly 60% of the time on similarly framed matches — not that the favourite will win. For the statistical concept, see Calibration (statistics).
Known limitations
Tennis has irreducible variance. The model has no view on injuries, withdrawals, or last-minute fitness news unless those signals are present in its training inputs. Predictions are best read alongside current news, not as a replacement for it.