By Ed Grimshaw · Cutting Comment · Data: Proform Racing / RaceMetrics

Two questions have been nagging at me for a while — the kind that get settled with total confidence over a pint in the members' bar and absolutely no evidence whatsoever. Is the average racehorse running less over its lifetime than it did twenty years ago? And does betting turnover fall away steadily as fields shrink, or does it go off a cliff at some point nobody's bothered to find? I've long thought the BHA's Race Planning function could do with someone who has actually had to price a race for a living rather than only ever watch one, and these are exactly the two questions that function ought to have settled years ago, on its own initiative, without anyone outside having to ask.

So I asked Simon Walton at Proform Racing to check both properly. Both turned out to be more interesting than I expected. One had an answer nobody in the sport appears to have actually checked. The other had no answer at all — not because the data doesn't exist, but because the people holding it have decided you don't get to see it. Between the two sits a fairly complete diagnosis of how British racing manages information: supremely confident about things it has never measured, and militantly vague about the one thing that would tell you whether the whole enterprise still adds up.

The Season That Never Moved

Ask around the weighing room on the first question and you'll get a fast, confident yes — offered as proof of everything from softer training regimes to a risk-averse veterinary culture to owners who won't run a horse unless the ground, the mark and the moon are all correctly aligned.

Simon Walton at Proform Racing pulled every run recorded on a British track since 1997 — starters only, each horse filed against the season of its debut and the code it ran under most. The first table back contradicts the premise of the question.

Flat horses have run between 5.2 and 5.6 times a season for thirty straight years. Jumpers have sat between 3.4 and 3.75. There's a Covid-shaped dent in 2020, when 4.47 replaced what would ordinarily have been another 5.2 or 5.3, and then the figure bounces straight back to 5.41 by 2025 as if 2020 were a rumour somebody else believed. Whatever else has changed about this sport in three decades — the prize money, the fixture list, the shape of the Flat programme itself — how often an active horse turns up has not moved at all. This is a horse that is, in any given year, exactly as busy as it has always been.

Which makes the next table considerably more interesting, because it says something close to the opposite.

The Career That Did

A horse that debuted between 1999 and 2003 ran, on average, 15.8 times on the Flat across its whole career. A horse that debuted between 2014 and 2018 ran 13.2 times — sixteen per cent fewer, measured against the turn of the century. Take the more conservative, complete-cohort comparison twenty years apart, 2004–08 against 2014–18, and it's still down ten per cent. Either way, a meaningful slice of career has simply stopped happening. Jumpers tell a gentler version of the same story: their decline, from 15.3 to 14.0, happened almost entirely before 2008 and has barely moved since.

Put the two tables side by side and the summary writes itself: the horse is not running less. It is leaving sooner. Something is ending careers earlier — retirement to stud, injury, an owner doing the sums and finding they no longer work — while doing nothing whatsoever to how often the horse runs while it's still active. That is a genuinely different problem from the one the weighing room usually describes. It doesn't point at training caution or veterinary conservatism, which would show up as fewer runs per season. It points at whatever is deciding when a career ends.

One caveat, and it arrived with the data rather than being dug out of it: the count only tracks British runs, so a horse exported to Australia, Hong Kong or the United States vanishes from this data looking exactly like a horse that has simply stopped racing. Exports have grown across the period measured, so some — not all, but some — of the Flat decline is horses leaving the country rather than leaving the sport. Leaving that caveat out would have made the headline better and the analysis considerably worse.

What hasn't changed, and this is the detail that actually matters, is the bottom of the distribution. The share of Flat debutants running three times or fewer has sat at roughly one in five for every cohort back to 1999 — 18.9 per cent then, 20.4 per cent now, essentially flat. The disposable, one-run horse the industry frets about in public has not become more common. What has genuinely thinned out is the top: horses reaching thirty runs or more have fallen from 15.1 per cent of the 1999–2003 cohort to 11 per cent of the 2014–2018 one, and the jumps equivalent from 16.0 per cent to 12 per cent. The 2019–2021 cohorts read lower still, at 10.9 and 8.3 per cent, but those horses are still racing and both figures will rise. The durable horse — the one that gives a stable, a punter and a racecourse eight or nine years of form to actually get to know — is the one quietly going missing. Nobody had written that headline, because nobody had counted that column.

The Second Question Nobody Can Fully Answer

The second question was sharper, and it's the one that ought to worry the BHA rather more than the first. Turnover, I suspected, is heavily driven by the price of the favourite, and the price of the favourite shortens as the field gets smaller — fine so far, every punter knows that instinctively. The real question was whether turnover falls at a steady rate as fields shrink, or whether it accelerates. If it accelerates, nobody planning the fixture list has factored that into anything.

Here the trail runs into a locked door. Nobody outside the bookmakers and the Horserace Betting Levy Board and a few chosen individuals holds genuine turnover data, and the Levy Board publishes totals, not a single per-race or per-fixture figure. I know this from the inside: I've spent several months this year asking the HBLB's own chief executive, Alan Delmonte, for exactly this — fixture-level levy revenue, nothing that would touch an individual account. I was told, informally, that bookmaker confidentiality prevents it. Racing's own funding body cannot tell racing which of its own fixtures are worth running, and calls the resulting silence “confidentiality” rather than what it actually is: a gap nobody with the power to close it has any commercial incentive to close.

What Proform can show is the mechanism itself, built from three years of Betfair starting prices across every clear-favourite race run in Britain.

Two things stand out. First, the market is honest across the range that carries almost all the volume: from two runners up to fourteen, implied chance and actual win rate agree to within a point, band after band, for three years running — that's tens of thousands of races pricing themselves correctly. It loosens only at the very top, where the sample thins to a few hundred races a band: eighteen-plus-runner fields show the favourite winning noticeably less often than its price implies (16.8 per cent actual against 19.3 per cent implied), and the fifteen-to-seventeen band shows the opposite wobble. That's small-sample noise, not a calibration failure — the busiest band on the table, eight to nine runners, is out by three-tenths of a point.

Second, the price does not fall in a straight line as the field shrinks. It moves gently, about three points a band, all the way down to eight runners — and then it jumps. In fields of five runners or fewer, 36 per cent of favourites go off odds-on; in fields of eighteen-plus, none do. Eight runners isn't a threshold picked to fit the table — it's the same point at which each-way terms drop from three places to two, a rule that governed 42.5 per cent of British races run in 2026 up to 8 September. The suspicion that started this, that the damage to turnover accelerates rather than tracking steadily downward, is exactly the shape you'd expect from a curve like this one. Proving it in actual pounds would need the one number the Levy Board won't release.

A Concession, Reluctantly Made

I have spent a fair amount of this publication's column space arguing that bookmaking, as currently structured, extracts value from racing rather than investing in it, and I stand by every word of it. But credit where it's due: the pricing engine that industry runs behaves like a Swiss watch next to the institution nominally planning the sport around it, which behaves rather more like a shed clock that's fine as long as nobody actually asks it the time. A market that recalibrates itself correctly for eighteen-runner sprints and five-runner novice chases alike, using nothing but the pooled judgement of everyone willing to back an opinion with money, is not the broken part of this business. The fixture list, built with no visibility of what that market is telling it, is.

What Would Actually Fix This

None of this needs a summit, a working group or another consultation document. It needs three things that already exist and are simply being withheld or ignored.

Publish fixture-level turnover — or at minimum publish it in aggregated bands that protect individual bookmaker positions while telling racing which fixtures are commercially viable and which are being run to fill a slot. The technology required is a spreadsheet, not a breakthrough.

Route fixture-planning decisions through someone who has actually had to price a race, not just watch one. Putting a current or former betting-industry professional inside Race Planning is not radical. It's the minimum competence you'd expect of any business deciding what to sell.

And measure the thing that's actually declining. The industry keeps a nervous public eye on the one-run horse, whose share of the population hasn't moved in a quarter of a century. It has not noticed that the durable horse — the one that races for eight seasons and gives everyone something to actually follow — has been quietly vanishing from the population the whole time nobody was looking at the right column.

These two questions took a handful of pulled tables and three years of Betfair data to answer properly. Nobody in racing seems to have thought it worth the afternoon.

Proform Racing Ltd and RaceMetrics is a data partner of Cutting Comment. Code CUTTINGCOMMENT for 20% off for 3 months