The Bounce Theory, Tested on 2.3 Million Runs

Career-best runs and what actually happens next

The bounce is one of the oldest ideas in form study: a horse that just ran the race of its life has emptied the tank, and next time it regresses. Generations of handicappers have crossed off last-time-out career-bests on that logic.

We put the theory to our full database — around 2.3 million runners across Britain, Ireland, Australia, Hong Kong, Japan, France and five other jurisdictions — and the general version does not survive.

What the ladder actually shows

Label every run by how it compares to the horse's own best recent figure and you get a ladder: new career top, matched top, modest decline, large decline, collapse. Then look at what each group does next time.

Next-race win rate falls in one clean direction down that ladder, and the direction is identical in every jurisdiction we tested. Horses coming off a career top or a matched top win two to four percentage points more than the local baseline next time out. Horses coming off a collapse win two to four points less — and a third to a half of them collapse again.

If the bounce were a general rule, the top rung of that ladder would sag. It does not. A career-best is, on average, good news for the next start.

Where the bounce is real

Split the career-bests by the size of the improvement, and the picture sharpens into something more useful than either the old theory or a blanket rejection of it.

Modest improvers — new tops by three to seven points — are the strongest next-race cohort in the data. A horse that just edged past its old best is typically a horse still on the way up, and the next run tends to confirm it.

Big jumpers — improvements of eight points or more — do bounce. They are the weakest of the improving group, and the regression is universal: every cohort coming off a large jump gives back six to eight rating points on average next time.

So "career best" is not one thing. A horse that improved a little and a horse that improved enormously are different animals, and the racecard prints them identically.

The part the market gets wrong

Mean reversion after a big jump is real, but it is also widely understood — so the interesting question is how the market prices it. The answer depends on where the horse is in its preparation.

The fresh big-jumper — early in a campaign, obvious and exciting — tends to be over-backed. The deep-campaign big-jumper — five or more runs into a preparation, written off as exposed — tends to be underbet, because a fresh career-best that deep into a campaign says the horse is still thriving. In Britain the fresh end of this pattern inverts, which is a good reminder that pricing patterns need local calibration even when the underlying form pattern is universal.

How to use it

When you see a last-time-out career-best, ask two questions before you react. How big was the improvement — a few points, or a leap? And where did it come in the campaign? A modest step forward is the most reliable positive signal in form reading. A huge leap is a warning to expect regression, and the only question is whether the price already says so.

On Mr. World Pool racecards every run carries a performance rating on the same scale, so the size of an improvement is a number you can read rather than an impression you have to form. The rating itself is explained here.