ThoroughWiz

Handicapping methodology

The ranking anchor, the model, our figure and the selection rules — the method in plain terms.

The anchor: the morning line

Horses are ranked in morning-line order, and nothing reorders that ranking (ML-BASE). That is a measured choice, not a default: the lab's own benchmark work found the morning line's top pick as accurate as any model the lab built, and the alternatives tested — including pace re-ranking and market-free models — did not beat it. The morning line is the control every candidate method has to beat before it can move a pick, and so far none has.

The model: win chances, printed

Alongside the ranking, a blend model gives every horse a win chance, and those chances print on the past-performance sheets. The blend's probabilities are the best-calibrated numbers the lab produces — in the benchmark they score better than the morning line's implied chances at saying how often horses win, even though its top pick wins no more often. That is the honest use of the model: a probability to weigh against the price, not a ranking to follow. The benchmark itself is published on the record page.

Our figure and the power rating

Every sheet prints our own speed figure for each horse's past races (figure v5.0): final times and beaten lengths measured against track constants estimated from official charts, a race-day variant anchored on each horse's own scored history, and distance-band offsets and correction tables that finish the figure. Figures print on an anchored scale where 100 is a stakes-winning performance, so a number means the same thing at every track. The power rating (OUR_RATING v5.0) is the summary form of the same work — one number for comparing horses across tracks, ages and sexes. A negative figure prints in red, in parentheses.

The pick (Selection Protocol v3.19)

One horse per race is the pick, chosen under the lab's written selection rules, and the pick is the booked horse: the headline record follows the pick, and results publish win or lose. Pace figures describe each race's shape as annotation only. The protocol is versioned — changes ship as new numbered versions with their evidence recorded, never as silent edits.

Tested before adopted

Every idea the lab uses was written down, run against historical races under rules fixed before the results were counted, and kept only if the count said to keep it. The failures publish alongside the successes: What We Tested keeps the claims ledger, and the research tier of the record keeps the closed studies with their sample sizes. The method’s source shelf is The Books Behind the Method.