Burun Farkı
A transparent AI horse-racing analysis platform. Model rankings, their reasoning and past results appear together. Explore the dated performance records and measurement scope in the evidence section below.
01 — What it does, and doesn't
It reads the race; it doesn't hand you a slip.
The platform takes a race's data and shows the model's ranking along with the factors that produced it. It generates no odds, recommends no bets, guarantees nothing. The decision is the user's.
Every prediction comes with its reasoning.
Rule-based contribution factors are listed per horse: which variable pushed the ranking up, which pushed it down. When the model returns something you dislike, you can see why.
02 — Method
The backtest runs on later-dated races the model never saw in training. Excluded races and the data range are listed in the method note.
Verified claims
Across 1,225 live-tracked races, Top-1 ~29% (28.6%) and Top-3 ~62% (62.0%) — honest post-deployment tracking. A separate training-time backtest measured ~69% Top-3 (a different measurement).
measuredverified 2026-08-25sourceThe 69-day/1,225-race live post-deployment aggregate displayed on 2026-08-25; not the training-time backtest. The evidence freshness window is 30 days.A rule-based contribution ranking is listed per prediction: which variable moved the ranking, and in which direction.
verified in source codeverified 2026-08-19internal record: ventures/burun-farki/web/src/components/HorseExplainCard.tsxThe 'top 3 factors' shown to users is a fixed-weight, rule-based ranking — NOT a formal explainability computation over the model output (a Shapley-based method, for instance); no such claim is made.
03 — Interface
Publication status needs re-verification
Check access through the Burun Farkı link, or send your questions by email.
Last publication check: 2026-09-06. You can open the link to check its current status.
Burun Farkı is not a betting product and does not broker bets. Past performance is no guarantee of future results.