Confidence Interval (95% CI)
A range around a measured average that accounts for sampling noise; LDBD treats a predictor as Verified once its 95% interval clears 0.
Formula
95% CI = mean ± 1.96 × std / √n (n = number of resolved predictions)
The average score you see on the leaderboard is computed from a sample, so when only a handful of predictions have resolved, a good number might be luck rather than real skill. A confidence interval accounts for that uncertainty by reporting the true average not as a single point but as a range. The more predictions in the sample (the larger n), the narrower that range gets, and the more you can trust the estimate.
LDBD computes a 95% confidence interval for each predictor and stamps them ✓Verified once the interval sits entirely away from 0. If the interval still straddles 0, the record can't be told apart from 'no skill' (a true value of 0); once it clears 0, there is an edge that luck alone struggles to explain. This drives only the confidence display and the Verified badge — it never removes anyone from the ranking.
Example
With mean a = 0.40, std = 1.2, and n = 40, the margin is 1.96 × 1.2/√40 ≈ 0.37, giving [0.03, 0.77] — it clears 0, so the predictor is Verified.
How LDBD uses it
The confidence interval is the final gate in LDBD's three confidence tiers (Rookie → Calibrated → Verified). Once enough predictions resolve and the 95% CI clears 0 entirely, the predictor earns the ✓Verified badge, meaning the track record is unlikely to be pure chance. The interval never shifts a ranking position or drops anyone from the board — it powers only the confidence display and the Verified check.
FAQ
If the interval includes 0, does that mean no skill?
Not necessarily. It usually just means there aren't yet enough resolved predictions to separate a real edge from luck. As the sample grows the interval tightens, and a genuine edge will eventually pull clear of 0.