With racing in the UK and most of the world currently suspended due to the Covid-19 outbreak, a lot of us have an unexpected amount of time on our well-washed hands.
There is only so much daytime telly, putting out the bins and cleaning of the kitchen that any of us can stand. Desperate times call for desperate measures.
Which is why I am inviting you to turn off that telly, and leave the bins and kitchen until later, to join me instead in an exploration of the world of handicapping. It will involve some theory but also some practical examples.
At the end of it, you may decide that rating races is not for you: hopefully, no great harm will have been done. Or you may decide that it is for you, as I did a long time ago. Either way, the aim is that you will understand better how others handicap and some of the principles involved.
What is handicapping?
“Handicapping” means different things in different parts of the world and in different contexts. The handicapping referred to here is the British/Irish version of assessing racing performance in a numerical way: that is, with ratings.
Comparisons are at the very heart of horseracing, from the early “my horse is better than your horse: let’s have a match” to the more recent “this year’s champion is better than last year’s champion” and “my country’s champion is better than your country’s champion”. Various ways of measuring these things have been developed, but the most enduring and most effective are ratings.
Timeform ratings in Britain first appeared in the 1940s and assessed horses on a universal scale, up to 140 (and ultimately a bit more) on the flat and, later on, up to 175 (and again beyond) over jumps.
Ratings are poundage-based, in that they represent the theoretical amount of weight required to equalise the chances of horses of differing abilities. Weight is used in reality in the races themselves in an attempt to achieve such ends.
The alternatives, such as to start off horses at time intervals or from staggered starts, are impractical. Weight matters – don’t let anyone try to kid you otherwise – for all that its precise effect in a given circumstance may be disputed. Weight is, as it were, the lingua franca of handicapping.
The theory is that a horse rated, say, 100 is 10 lb better than one rated 90, all other things being equal, though what that 10 lb means in terms of lengths or time will vary according to circumstance.
In order to rate horses, we need to rate the races in which they run. One of the first steps in doing this is to establish how much better, or worse, each performance within a race is than the others.
Calculating a race
Since 1997, the official margins between horses have been conversions of the time lapses between them at the finish. The precise conversion depends on circumstance, but the good news is that the British Horseracing Authority (BHA) has recently taken to publishing the actual times each horse recorded in Britain (Irish racing had been doing so for some time), if on something of a delay.
These may be found attached to each horse’s performance in the Results Section of the BHA site, as well as in the sectional details of races covered by Total Performance Data here on attheraces.com.
This means that any calculation of superiority/inferiority between horses in the same race should start from a time-based calculation. Form and time are inextricably linked.
Let us see how this might be tackled in a real-life example, such as the Betway Winter Derby at Lingfield on 22 February. The seven runners all carried the same weight, and for the purpose of this exercise we will ignore any considerations of weight-for-age.
Again, for the purpose of this exercise, I will use the time conversion described in The Timeform Knowledge of 1500 * (actual time minus standard time) / (actual time), but with “standard time” replaced by winner’s time.
In line with Excel conventions, * means “multiply”, / means “divide”, and instructions in brackets need to be completed before being reintroduced into the equation.
This works well enough for ordinary circumstances on the flat, and for most racing on Polytrack as it happens, but less well in others.
On the day, and at the finish, Dubai Warrior was nearly 6lb better than Court House, who was in turn about 1 lb better than Bangkok, who was 10lb better than Dalgarno, and so on.
The next question is how to tackle an assessment of this race. Contrary to what used to be maintained in some quarters, this is not simply a question of guessing which horse “ran to form” and then basing an entire assessment of a race around such an assumption.
Besides anything else, horses do not magically run to the precise same level from one race to the next for the convenience of those who make figures.
All athletic performance is subject to a degree of variation, and the dynamic world of horseracing – in which horses are running at different distances, on different surfaces, at different courses, and at different stages of fitness and maturity – perhaps more so than most.
This is where the concept of “race standardisation” helps. We know, from countless evidence over many decades, that some races tend to attract horses of a similar standard year to year, and that others do not. Either way, the degree to which they do, or do not, and what that standard tends to be, may be established by analysis.
Every horse has a history which can be useful in ascertaining what it has achieved in the present (and may achieve in the future), but most races do, also. Information theory dictates that it would be foolhardy to ignore this.
As an example, a race like The Derby at Epsom attracts the same sorts of horses each year: they are those that connections believe are the cream of the middle-distance classic crop, and in many cases they will have indicated as much by their form in the various trials.
This may not be entirely apparent by those horses’ pre-race ratings, but it is very strongly implied by their ability to come to the fore in a race with such prestige and value. If you treat it “just like any other race” you will come unstuck, time and again.
In another example, if the precise same set of unraced two-year-olds pitch up at Ascot you should treat them differently to if they pitched up at, say, Catterick.
Good horses do sometimes debut at the latter track, but not often, and they are likely to make their presence known by beating their rivals handsomely when they do. Good horses often debut at Ascot and sometimes get beaten by even better horses into the bargain.
So, how might race standardisation apply to the Betway Winter Derby?
What we need is an assessment of recent editions of the same race (or, failing that, similar races or a modelled assessment of the same). In particular, we want to know what it usually takes to win, finish second, finish third, and so on, in this race.
You can use established ratings services as your starting point here – the figures will quickly become your own if you tinker with them enough – or you could try to generate your own ratings from scratch.
The latter is a major undertaking, but can be a fruitful one. However, in the first instance, let’s piggy-back our initial figures on those of the BHA handicappers. Their “Performance Figures” have been attached to results on the BHA site for a while now and cover the last five years of the Winter Derby, which is sufficient for standardisation purposes.
An average winner of the Winter Derby would be rated 110.4 on the BHA level, a second would be 107.4, a third would be 104.0 on average, and so on.
From that, we can deduce that this year’s winner would be rated 110.4 if we used information only from recent winners, and if this year’s winner was “typical”, but 112.9 (the historical average plus the 5.5 lb that Court House was in arrears of Dubai Warrior) if using information only from second-placed horses, and 110.6 if doing the same with only third-placed horses.
The fourth and fifth horses were beaten sufficiently far by Dubai Warrior that they push those figures up to 121.0 and 122.6.
The value of these various figures becomes less the further down the field the horse finished. You would not, for instance, place the same predictive value on the figures derived from a horse finishing last as from one that won. The weighting I use is 1/N, where N is the finishing position of the horse.
So: that winner’s figure of 110.4 gets multiplied by 1; the second’s figure of 112.9 by 1 over 2, or 0.5; the third’s figure of 110.6 by 1 over 3, or 0.33; the fourth’s figure of 121.0 by 1 over 4, or 0.25; and the fifth’s figure of 122.6 by 1 over 5, or 0.2.
You add the results together – they come to 258.5 – and divide that by the sum of the weightings – 1.0 plus 0.5 plus 0.33 plus 0.25 plus 0.2 equals 2.28 – to get the standard figure for the winner: 113.4.
As it happens, the BHA handicapper’s assessment of Dubai Warrior’s win in this year’s Winter Derby is currently 114.
The figure goes up or down depending on how much, or how little, the winner had strung out his rivals, especially those who finished closest to him, and the figure would be higher if those margins had been achieved at 5f than if they had been achieved at 16f
As you may have realised already, all of this becomes much easier if you put the calculations into a spreadsheet or similar!
It is worth mentioning that there should be an adjustment for field size – the bigger the field, the higher the expected rating, both empirically and mathematically – and that ideally you should work out each historical race individually so that you get a range and distribution, not just an average, but the above serves well as an illustration of the overall approach.
Precedent suggests that the winner of this year’s Winter Derby “should” be rated 113 or 114, with the range (you can take my word for this, or, even better, try to work it out for yourself) of 110 to 116.
That range is narrow enough to inspire confidence that precedent is a good guide in this instance. If that range was, say, 100 to 120, then it would be telling you that the race strength varies a lot and that precedent may not be much of a guide.
In the absence of any other information, you have a very workable starting point for your assessment of the race. Indeed, you should be questioning yourself if you come up with a figure by separate means that is significantly different to those quoted. Exceptions do occur, but exceptions are what they are.
Race standardisation has been used for decades in the assessment of two-year-old races, where little other information is known, and it has been used more widely in the computer age for races of all shapes, manner and sizes. You do not have to be a slave to the process, but you should find that race standardisation is an invaluable aid in giving you a ballpark figure and in ensuring that you do not stray too far from an objective measure of performance without good reason.
Next week, we'll look at calculating ratings in races where horses have carried differing weights.