Formintel is a three-part system. Each tool does a specific job and they work together seamlessly.
Here is exactly what happens under the hood — from loading a racecard to placing a smarter bet.
01
📄
PDF Extraction
You load an ATR form guide PDF. Formintel reads every page, groups the text by position on the page, and extracts each runner — stall, draw, name, age, weight, jockey, trainer, form string, and official rating. No typing required.
The Predictor
02
🔢
The Scoring Model
Each horse is scored across 12 signals: market implied probability (overround removed), recency-weighted form, official rating, pace classification, draw bias, jockey tier, trainer tier, jockey-trainer combo bonus, going suitability, distance suitability, class, and days since last run.
The Predictor
03
⚖️
Model vs Market Blend
The form score is blended with the market implied probability. When odds are available, the market gets 40% weight and form gets 60%. This grounds the model in reality — if a horse is drifting to 33/1 the model won't wildly back it on form alone.
The Predictor
04
💎
Value Detection
Once model probability and market implied probability are both calculated, any horse where the model says a higher probability than the market is flagged as a value overlay. That gap — in percentage points — is your potential edge.
The Predictor
05
💰
EV & Kelly Calculation
For each horse, Formintel calculates the exact Expected Value of a Win, Each-Way, or Place bet at the current odds. A positive EV bet is one worth taking long term. Kelly criterion then suggests the optimal stake size based on your edge and bankroll.
The Predictor
06
🎯
Lucky 15 Optimisation
Load multiple meetings and Formintel searches thousands of 4-horse combinations across all races. Three strategies: Likely (highest win probability), Value (biggest overlays), Mixed (maximum Lucky 15 expected return across all 15 bets).
The Predictor
07
🧠
Building Your AI Model
The Trainer takes a CSV of historical race results and learns from it. It calculates real win rates per jockey, trainer, and course. It builds draw bias statistics, going preferences, and trainer home-course maps. The output is a model.json file loaded into the Predictor to sharpen the scoring.
The Trainer
08
📊
Tracking Your Edge
Every bet you log in the Tracker is measured against Starting Price. CLV (Closing Line Value) tells you whether you got better odds than the market settled on — the single most reliable indicator of long-term profitability. The accuracy tab shows whether your model predictions are well calibrated.
The Tracker
