Why the Breeders' Cup looks the smart target for Gosden's Derby hero

Golden Horn’s trainer, John Gosden, is renowned as a smart man. But what do the numbers say is the smart thing to do with the G1 QIPCO Irish Champion Stakes winner?

To evaluate the choice of G1 targets, we turn to Racing Post Ratings (RPR), a widely accepted measure of international racehorse merit, which has stood the test of time over the years.

Following his controversial win at Leopardstown, the most obvious targets for Golden Horn are the Arc at Longchamp (October 4), QIPCO Champion Stakes at Ascot (October 17) and Breeders’ Cup Turf at Keeneland (October 31). The scheduling of these races means he could also run in the first and the third. We will assume for these purposes that he won’t be retired to stud just yet.

Using the data provided by the RPRs of the last 10 renewals of these races, we can construct various metrics that capture how good they tend to be. These metrics tend to be referred to as standards by the various bodies who evaluate racehorse form. For comparison, we will also look at how the Irish Champion Stakes stacks up with them.

Three measures are considered:

1) The RPR of the winner;

2) The average RPR of the first three finishers;

3) A weighted average of the first three finishers, with coefficients 3, 2 and 1.

It is desirable to add 3lb to the ratings of fillies who have populated the results of the major middle-distance races in question, for it requires a 3lb better performance by a colt to run a dead-heat with them. (Personally, I am strongly against fillies and mares receiving a sex allowance in races against males.)

RPR of the winners only

Average RPR of the first three finishers

Weighted average of the first three finishers

The most important conclusions to draw from this data are:

1) The QIPCO Irish Champion Stakes is firmly established among the best races in the world. This is related, of course, to the co-development of Coolmore/Ballydoyle. However, many non-Aidan O’Brien horses have bolstered its reputation too.

2) The Irish Champion was a better race than the old Champion Stakes at Newmarket, but the latter has not surprisingly surpassed it at Ascot due to its greater prize money and higher profile as part of the British Champion Series, which was instigated in 2011.

In an average year, a horse has to run 2lb (a length better) to reach the same position at Ascot than at Leopardstown. Note that, if we restrict our attention to all horses not named Frankel (the 2012 QIPCO Champion Stakes winner) and Sea The Stars (the 2009 Irish Champion winner), the average of the first three finishers is reduced a little: the Ascot Champion drops from 127.0 to 125.4 and the Irish version from 123.7 to 122.9.

3) The Arc, as might be anticipated, is the hardest of the top races to win. The winning RPR has a mean of 130.3 and a sample standard deviation of 2.5. This latter statistic means that, if the population of Arc-winning ratings was normally distributed and the sample representative of the quality of the race in future, it could be expected from the mathematics that two-thirds of all Arc winners would run an RPR between 128 and 133.

4) While most experienced racing fans could anticipate that the G1 Breeders’ Cup Turf is of lesser quality than the European middle-distance races that precede it, there is a seriously wide gap in the standard required to win on either side of the Atlantic.

While Breeders’ Cup Turf winners have a mean of 123.3, the sample standard deviation is similar at 2.8. We can infer – with the same caveats as apply to the Arc – that about 66% of Turf winners should fall between RPRs of 121 and 126.

And, very much the point here is that the gap may be widening, subject to random effects. There is a correlation coefficient of -0.79 between the RPR of the Turf winner and the number of the year. So, higher-numbered years like 2015 are likely to be associated with lower ratings of Breeders’ Cup Turf winners.

This last conjecture isn’t safe to assume just from the data, but it is fair to have a strong prior belief that it might be the case, given that similar downward trends are affecting competition in many other divisions of U.S. racing, not just middle-distance races on the turf.

Given that Gosden graduated from Cambridge with a degree in Economics, it is certain he knows far more than I do about the concept of Expected Value. However, I will still attempt an amateurish look at evaluating the best decision to make for Golden Horn, using this important prism.

According to RPR, the average winner of the BC Turf is nearly three Arc standard deviations lower than is required to win in Paris, but the rub is that trying to win the first may have a significantly negative impact on a horse’s much greater chances of winning the second. The Arc prize-money (€5 million) is around 60 percent higher than the Turf ($3 million), if my currency conversion is accurate, but the figures suggest that, all other things being equal, Golden Horn’s chance of winning at Keeneland is significantly greater than 1.6 times his chance of winning at Longchamp.

Indeed, bookmakers quote the Cape Cross colt at 7-1 for the Arc which, considering their in-built profit margin, gives him about a 10 per cent chance of winning. In prize-money terms only then (I am assuming the distribution of place money is similar), Golden Horn should run in the Turf if connections think he has a 16 percent chance (5-1) or better. In fact, Golden Horn would be half that price if bookmakers thought he was going to run.

Irrespective of these considerations, however, Golden Horn’s connections seem minded to put some store in managing his reputation - he was withdrawn from July’s G1 King George at Ascot on the day of the race over fears that he wouldn’t handle yielding going. Consider the effect on his reputation, then, of these two scenarios, which might reasonably transpire if Golden Horn ran to an RPR of 126 (his last four numbers are 124-118-132-127):

1) The ground at Longchamp surprisingly comes up good. Golden Horn gallantly finishes third to Treve and New Bay;

2) On the expected fast surface at Keeneland – with its newly constructed turns – Golden Horn finishes his career in a blaze of glory with victory.

If there really is a difference in the expected earnings of a stallion according to the number of defeats he suffers, rather than the level of his form, as many in the bloodstock world assert, it is a no-brainer to run Golden Horn at the Breeders’ Cup.

Gosden has a great record shipping horses and it would seem to me that Golden Horn would be a certain runner if exactly the same race – on a fast surface - were staged at Newmarket. What are the reasons you would choose a path of lower expected value for the colt?

If the current trajectory of the Turf is meaningful, the expectation for winning future renewals is almost heading towards G2 in European terms. And there are not many of those worth $3m.