How Much Should You Forgive a Bad Run?

The forty per cent rule of form reading

A horse that normally runs to 110 just ran to 102. Is it a 102 horse now? Still a 110 horse having a bad day? Something in between?

Most handicappers answer by feel. We answered by regression, across eleven jurisdictions: measure every horse's deviation from its own recent level, then measure how much of that deviation shows up in the next start. The construction needs care — the two baselines must not share races, or a bad run today mechanically drags down the standard it is judged against tomorrow — but done properly it turns the oldest argument in form study into a coefficient.

The forty per cent rule

A horse carries roughly forty per cent of any deviation into its next start. Run eight points below your level, and you are about three points worse next time — not eight. Run eight points above, and you keep about three of them.

The striking part is the symmetry. It is tempting to believe good runs are "real" and bad runs deserve forgiveness, and our own first measurement seemed to show exactly that: bad runs appearing to carry nothing forward at all. That result turned out to be an artefact of the extremes — a horse beaten twenty lengths was usually not ridden out, so its figure records a jockey's decision rather than the horse's ability, and those runs dominated the bad tail. Compare good and bad runs over the same range of magnitudes and the asymmetry vanishes in ten of eleven jurisdictions. Forgiveness is symmetric. The only exception is Japan, where good runs genuinely carry about twice what bad ones do — noted, and treated cautiously until it replicates.

The baselines by jurisdiction, for the curious: Australia 0.37, France 0.42, Britain 0.45, Japan 0.45, Hong Kong 0.46, Ireland 0.48. Looser betting markets carry more forward.

What changes the weight

The forty per cent is an average. Three things move it a lot, and each is a number you can apply directly.

Ground. A British run on heavy going carries 0.30 where an all-weather run carries 0.48 — the wet run is nearly forty per cent less informative about the future. This is the classic wet-track excuse, finally expressed as a coefficient, and it is the only excuse in our entire battery that survives testing.

Surface. Dirt form transfers worse than turf form. A Japanese dirt run carries a third less than a Japanese turf run (0.36 against 0.53), and Hong Kong shows the identical split. Weight dirt lines accordingly.

Time. Form has a half-life. A British run from the last fortnight carries 0.50; the same run four months old carries 0.24. Whatever a stale figure once said, half of it has evaporated.

The runs to trust most

Three kinds of run that handicappers routinely discount turn out to deserve the opposite treatment.

A run at a materially different trip from the horse's recent races carries more forward, not less. So does a first run back from a break. So does a run into a hostile pace — the closer in a crawl, the front-runner in a duel. The mechanism is simple once said aloud: a run in unfamiliar conditions reveals something durable — a distance preference, a fitness level, a limit of the running style — and what it reveals repeats. A run in familiar conditions mostly confirms what you already knew.

The standard advice to throw out the wrong-trip run, the comeback run and the wrong-pace run is backwards. Those are the informative ones.

How to use it

Next time you look at a form line, resist both extremes. Nobody is as bad as their worst figure, and nobody keeps all of a career-best. Take the deviation, keep forty per cent of it, discount wet and dirt and stale runs — and pay special attention to the runs that happened in the wrong conditions, because those are the ones that taught you something.

Every run on our racecards carries a rating on one comparable scale, which is what makes this arithmetic possible in the first place. The method behind the ratings is described here.