The Winter Derby was a straightforward race in a number of senses, not least that all the horses were carrying the same weight. Not all races are like that: indeed, the majority are not.
For races in which horses carry different weights, that fact needs to be incorporated into the difference at the weights at the outset. For example, if the second and fourth horses in that race had carried 10 lb penalties, then the difference at the weights would be 0/-4/7/7/17 rather than 0/6/7/17/17, with a minus value meaning the horse in question was better (in theory) than the horse who won.
Calculations need to take that into account. The effect in the fictitious example given would be for Dubai Warrior to be rated 110 on race standards and for Court House – who was beaten by him but carried 10 lb more – to be rated 114. You can take my word for that, or, better still, work it out for yourself.
It is also desirable, though not obligatory, to factor in weight-for-age, so that a three-year-old carrying 9-00 and a four-year-old carrying the same would be treated differently, with the former in effect 9-07, 9-10, or whatever, according to how much weight it should be carrying to compensate it for its lack of maturity.
There is a good reason why ratings organisations have traditionally included weight-for-age in such calculations, as the performance of top-rated horses would suffer otherwise as immature (and yet progressive) horses would likely lag behind. But that argument is not so convincing in the computer/data age in which such improvement could be built in on an individual and ongoing basis.
There is also the question of which weight-for-age scale to use. The various ones converge with more mature horses, but the BHA’s version is clearly wrong with two-year-olds (somewhat academic unless you are compiling timefigures, when it is not) and has a knock-on effect with three-year-olds.
There is much to be said for knowing that a 120-rated three-year-old is equivalent to a 120-rated four-year-old, for all that the former still needs to improve at an average rate for that to be strictly true, so I recommend factoring in weight-for-age and if possible using Timeform’s scales, with which I had some involvement.
Weight-for-age is far more of an issue on the flat than over jumps, due to the ages and maturities of the horses usually involved, and disregarding it in the latter sphere is at least feasible.
There is, of course, a specific race type in which horses usually carry markedly different weights to each other, and that is a handicap. For reasons which should soon become apparent, handicaps need to be treated as a special case.
If you were to use the performance ratings of past contestants of the same race to guide you in your assessment of the principals in today’s handicap, you would find that your ratings would be much too low when lightly-weighted horses won and ran well in the past and the opposite when highly-weighted horses had done likewise.
The important thing with handicaps is not how horses perform in absolute terms but how they perform compared to the task that has been set them.
A horse running off a mark of 90 and returning a performance rating of 98 will finish ahead of one running off a mark of 105 and returning a performance rating of 100, despite its being a slightly inferior horse.
To see this in action, let’s consider the Coral Cup Handicap Hurdle, run recently at Cheltenham.
Those figures are pretty typical for a competitive handicap with many runners: winners have on average exceeded their BHA marks by 7.4 lb in winning, seconds by 4.4 lb in finishing second, and so on.
Assessing jumps races becomes more difficult with smaller numbers of finishers, extreme margins, and, in particular, with significant departures late in the race.
Handicaps are won by horses who are, by definition, usually showing themselves to be well-handicapped (exceptions can be small fields and races in which horses ran from out of the handicap), and that applies to both flat and jumps. To a greater or lesser degree, horses finishing close up are also likely to be well-treated, or at least not badly treated, and so on.
Again, I have used the 1500*(actual time minus winner’s time)/(actual time) mentioned earlier, for simplicity’s sake, though I personally vary that calculation according to the precise circumstances under which a performance has been achieved. A warning: still not all official jumps times can be trusted.
What we need to do is to add the three figures in the right-hand columns together to come up with the standard rating for the winner based on whether we are using the winner, the second, third, fourth or fifth as a guide: in the end, as before, we will use all of the first five finishers but with weightings varying according to their finishing positions.
The standard figures for the winner come out at 147.4 (on winner), 146.7 (on second), 148.4 (on third), 148.8 (on fourth) and 147.2 (on fifth).
After weighting by 1/N (as previously), this results in a race standard figure of 147.5 on the winner: the performance ratings of the beaten horses cascade down from that according to the difference at the weights highlighted in red (but with negative values turned into positive ones and vice versa, so that Black Tears should be 149.2).
Handicap ratings perform in a more or less predictable manner. For instance, the uplift by-position for horses running in handicaps on the flat in Britain in recent years has been: +5.0 for winners, +1.4 for seconds, -1.6 for thirds, -4.5 for fourths, and -7.8 for fifths.
But that is ALL flat handicaps, and the picture will be different according to whether the race is age restricted, won by wide or narrow margins, run at a good or moderate course, and many other factors.
Nonetheless, this is a decent starting point for tackling handicaps which have not been run previously: this is known as general race standardisation, whereas the figures from the Coral Cup above resulted from specific race standardisation.
If you have got this far, you may be wondering when an individual horse’s previous achievements will be considered relevant. It is all well and good using precedent and wider trends to assess races in a formalised way, but what about when you know important information about the individual horses and perhaps little about the races in which they run?
There is another form of standardisation which deals with this, which is known as prior-rating standardisation, and which takes the pre-race ratings of the contestants, converts the race into a quasi-handicap and applies similar rules to those mentioned previously.
The most effective standardisation merges figures at a race level and at an individual horse level so that both horse and race information is taken into account.
It is a complex process, and I do not propose going into it here. Instead, I will introduce a relatively new concept in handicapping which does something similar but in an easier-to-follow way. That is for the next part in this series - to be published next week.