A horse's career earnings -- a reliable indicator or a misleading one?

On a runner's form sheet, one line almost always catches the eye: the amount of earnings a horse has accumulated over its career. Several hundred thousand euros for some, a few hundred for others. The temptation is strong to read this figure as a ranking of quality: the more money a horse has won, the better it must be. The reality is more nuanced. Career earnings tell a real story, but they tell it poorly if read on their own.

What career earnings actually measure

Career earnings add up every prize a horse has brought its owners, race after race, since its debut. It's a cumulative sum. And that's precisely where the first trap lies: a cumulative sum depends first and foremost on the number of opportunities it's had to grow.

An eight-year-old horse that has run a hundred times will, all else equal, have accumulated more than a four-year-old that has run twenty times. Not because it's better, but because it's had five times as many tickets to play. The total rewards longevity and regularity of presence just as much as pure talent.

Career earnings measure a cumulative total, not a current level. A large total may belong to a horse in decline that shone three seasons ago, just as a modest total may hide a young horse on the rise.

The biases that make the figure misleading

Several factors distort a straightforward reading of earnings.

Age and race volume. As mentioned, a high total is often the fruit of time spent racing more than of intrinsic superiority. Dividing earnings by the number of races run already gives a fairer picture.

Discipline. Prize money isn't calibrated the same way across different worlds. Comparing the gross earnings of a trotter from Vincennes and a flat-racing galloper doesn't make much sense: they don't follow the same scales, nor the same racing calendars.

The level of the races. A horse that accumulates minor placings in highly endowed races can show a higher total than a horse that wins regularly in more modest events. The first isn't necessarily the better performer on the day, under the exact conditions of the upcoming race.

The era. Earnings are built up over years. A horse's form and context may have changed radically between when it built up its fortune and today's race.

When earnings become a useful signal

That doesn't mean this data should be ignored. Taken in context, it does provide real information.

A horse that has won large sums has, by definition, faced good-quality opponents and managed to hold its own at least some of the time. Under comparable conditions, it's a marker of class. The ratio between earnings and number of races, how recent or old those earnings are, and consistency with the horse's current discipline: these are the cross-checks that turn a raw figure into a usable indicator.

That's exactly the logic behind HIPIKA's analysis tool. Career earnings are never read in isolation there: they're placed alongside many other criteria, put in context by age, discipline, recent form, and the level of the field faced. The same amount doesn't carry the same weight depending on the profile of the horse behind it.

Take a hypothetical example: two horses show identical earnings. The first built theirs over three seasons of steady good results; the second from a handful of big placings two years ago, with no recent success. The figure is the same, the meaning is the opposite.

Why no single indicator is ever enough

Career earnings share a limitation with every indicator taken in isolation: none of them alone can tell you who's going to win.

The favorite's odds are the best illustration of this. The market favorite, that is, the horse most heavily backed and therefore best informed by the collective judgment of all bettors, only wins about one time in three in French flat races. In other words, even the best collective summary the market can produce is wrong two times out of three. A simple earnings figure, inevitably, says even less.

That's why HIPIKA's tool reasons over the whole of the data rather than on a single star criterion. Its estimates are measured out-of-sample, meaning tested on races the tool had never seen, drawn from a history of more than 2.1 million runners. In a recent test covering close to 9,900 races, the horse judged most likely by the tool finished in the top three in about 61% of cases, and won in about 33% of cases. Figures that confirm one thing: we're talking about refined probabilities, never certainties.

What this changes in practice

Faced with a runner's form sheet, career earnings deserve a look, not a conclusion. Before drawing anything from it, it's worth asking: over how many races was this total built, at what age, in which discipline, and in what era? A horse with a big pot isn't automatically the best on the day, and a horse with a modest total isn't necessarily a hopeless outsider.

The indicator is reliable when put in context, misleading when read at face value. Horse race betting remains a game of chance, carrying a risk of loss, reserved for adults. No data, however well cross-referenced, turns a probability into a guarantee.