Public accountability layer · no login required
Every top pick, settled in public. Including the ones we got wrong.
The model's own record over the last two weeks — the runner it rated first, the price it started at, and where it finished. No filtering, no “selected bets”, no quietly dropped losing week.
Settled picks
Updates as results come inNo settled races to publish right now — results land after each meeting and this table fills automatically. In the meantime, the full walk-forward record covers 15,048 races.
“Top pick” is the runner the model rated first before the jump; “AI pick” is the retired AI analyst's own top selection for the same race, with the reason it gave at the time. It no longer appears anywhere in the product and moves nothing in the ranking. Neither is a betting recommendation. Races marked VOID were scratched or did not finish and are excluded from the strike and return figures. Over the full 15,048-race backtest, backing every top pick flat returned -17.1%. Two weeks is a small sample and will swing hard in both directions; that is the point of publishing it.
Step 1
Pre-race read
The board records model rank, probability, price-vs-probability context, and market notes before the jump. Nothing is edited afterwards.
Step 2
Market check
Late firming, drifts, and PULSE-style movement are reviewed against the original race shape instead of treated as standalone hype.
Step 3
Result reconciliation
When official results land, the original read is compared with the finish, the closing market, and how the race was actually run.
Step 4
Learning note
The receipt ends with what held up, what missed, and which lesson feeds the next race-day workflow.
Daily race intelligence brief
Get the next daily race intelligence brief.
Start with the free morning briefing, then use these receipts to judge whether the board is helping your race-day research.