AI consensus

Where the models agree

Today's horses scoring in the top tier across multiple AI variants. We run ten scoring algorithms (A–J) over every runner. When several variants agree, that's a stronger signal than any single score in isolation. Disagreements are interesting too — they're the runners with the biggest model spread.

How this works

For every runner on today's racecard we have ten AI scores — one from each variant. Each variant scores 0–100. A score ≥ 65 is "top tier" (the same threshold used elsewhere on the site).

Consensus: 4 or more variants put this horse in the top tier (≥ 65).
Strong consensus: 6 or more variants agree — high conviction.
Contrarian: spread between the highest and lowest variant ≥ 70 points and the top score is ≥ 70 — i.e. at least one variant rates the horse highly while others don't. These are the model-disagreement picks.

Use the per-variant grid on each card to see which variants like the horse and which don't. What each variant measures →