How to Read SnapBet's Confidence Scores

A confidence score is the model's probability that its picked side will win the match. A 65% confidence "Home Win" means the model believes the home side wins 65 out of every 100 times — and crucially, that the away side or a draw still wins 35 out of 100. Confidence is not certainty.

What "high-confidence" means here. SnapBet flags a pick as high-confidence when at least one of our two models reports 50% or higher confidence. The premium tier requires more — typically 60%+ on V3 in leagues where V3 has historically been the stronger model. Premium-tier picks have been the strongest converter on the public audit at /performance — but past performance doesn't guarantee future results.

Sample size before headline. A 70% confidence score from a model that's only made 5 picks on this team is statistically weaker than a 58% score from a model with 30 picks. When you're looking at a team page, glance at the "n = X" number under each model's accuracy figure. Below 10 we don't even flag a Recommended model.

Confidence and value are different. A 70% confidence pick at decimal odds 1.30 has the same value (zero edge) as a 50% pick at 2.00. Confidence tells you accuracy; value comes from comparing accuracy to the market's implied probability. The premium tracker on /performance simulates flat $100 stakes regardless of odds, which makes hit-rate the dominant variable. Real-world bet sizing should factor in both confidence AND odds.

When the model says "no bet." Many matches don't get a picked side at all. That's a feature — we'd rather pass than push a coin-flip call. Fixtures without a meaningful pick still show up in the full fixture lists on /soccer/today, but they don't make the top-picks shortlist.