1. From form to probabilities

Our models read a deep history of Australian racing — results, sectionals, barriers, going, class, jockey and trainer records — and turn it into an objective win probability for every runner in a race. The output is a model price: the fair odds we believe a runner is genuinely worth, independent of the market.

2. Fair price vs the market = edge

We compare each runner's fair price to the market price. When the market is offering more than our fair price, there is positive expected value (a positive edge). That gap, not the favourite's name, is what we are hunting. Each pick on the Picks page shows its fair price, the pre-race market price and the edge between them.

3. The filters

A raw edge is not enough. Before a selection qualifies as a play, it has to clear a set of discipline filters, which we publish alongside each model:

Model rank — typically the model's top-rated runner in the race.
Jockey form — a minimum recent (90-day) strike rate, so we are not riding cold hoops.
Field entropy — a cap on how open the race is; the more chaotic the field, the less reliable any single read.
Positive SP-EV — at settlement, only selections that were genuine value at their starting price are counted as live plays.

4. Staking and settlement

We keep staking deliberately simple so the model's skill — not a staking trick — is what you see. Every qualifying selection is a flat $1 win bet at the starting price (SP). A daily stop-loss caps how much any single day can bleed, so one bad afternoon can't distort the record. Some models also run a wet-track variant that only fires on rain-affected going.

5. The receipts

Every settled selection flows into the Results ledger: outlay, return, profit, win rate and ROI, tracked per model with a running cumulative P&L chart. Nothing is cherry-picked — losing days are on the page next to the winning ones. Click any day in the ledger to see the exact selections behind that line.

A note on honesty