Class Performance Index relative to other ratings
In my previous post I discussed how advantage at a yearling or two year old sale can be found by looking specifically at offspring out of mares that have a CPI of 2.62 or better. Foals out of these mares represent an advantage over mares that have lesser race performance, all other things being equal.
I also discussed how through the concept of missing heritability by specifically selecting for mares that could run themselves, we are hoping to diminish this effect that can influence prediction of outcomes.
The Class Performance Index is a rating that has been developed by the Jockey Club Information Systems. It is not a widely use statistic as far as I can see, most likely because unlike the AEI (through Joe Estes) or APEX ratings (Bill Oppenheim), there hasn't been someone writing about its use. Because of this it is not really an understandable statistic so I decided to take a look at it and see how it relates to common figures namely the Equibase Speed figure (USA), the Racing Post Rating (U.K) and the Racing & Sports Rating (Australia).
I gathered 100 horses with a CPI distribution in the population in keeping with my previous studies not the CPI and also obtained their peak Equibase figures. You can see from the above chart that the relationship between the variables of CPI and Peak Equibase figure was a logarithmic one, or more specifically a log value best described it. The correlation and subsequent equation to explain the relationship between the CPI and Peak Equibase figure was:
y = (10.596*LN(x)) + 82.256 R² = 0.79755
The natural log of the CPI was the closest fit that I could generate for the data and with an R² of 0.79 it is a reasonably good fit (possibly would be better if we were able to use the peak CPI of a horse rather than the average CPI as it is calculated now). Ideally you might also use another tool such as Nutonian's Eureqa software, but for the purposes of this study it worked.
The Racing Post Rating was an even closer fit to the CPI. The correlation and subsequent equation to explain the relationship between the CPI and Peak Racing Post figure was:
y = (9.9994*LN(x)) + 85.806 R² = 0.87491
This was the data where the variables of CPI and the rating correlated the least. The correlation and subsequent equation to explain the relationship between the CPI and Peak Racing and Sports figure was:
y = (4.7213*LN(x)) + 48.799 R² = 0.66987
While each of their 'best fit' varied a little, the equations generated did go towards explaining the relationship between the variables. From there I was able to use the three predictive equations to see what value would be predicted for each of the equations on the basis of using our 'elite runner' CPI of 2.62.
So what this is now saying is that the CPI is 'roughly' equivalent to a Peak Equibase of 92 and a Peak Racing Post Rating of 95. If you keep in mind that mares that have a CPI of 2.62 or better offer a statistical advantage in the breeding shed (that is buying yearlings out of mares who have a CPI of 2.62 or better is a more profitable play long term), it is roughly equivalent to say that it is more profitable to keep your yearling buying to those out of mares with these performance ratings.
Interestingly, over on his own web site Chefderace.com, Dr Steve Roman had put together a chart of equivalent ratings between Equibase, Racing Post and Timeform Ratings. By Steve's Calculations a Racing Post Rating of 97 is roughly equivalent to an Equibase rating of 93 (and a Timeform of 96 and a Beyer Speed Figure of 81). Our calculations by correlation with the CPI of 2.62 being our 'elite rating' results in a Racing Post Rating of 95 and an Equibase of 92 so I'd be pretty confident that the CPI of 2.62 or better is a good proxy for an elite runner and the CPI is a fairly accurate metric to use to measure performance.
But here is the more interesting part. A CPI of 2.62 doesn't mean that the mare has to be a stakes winner. She doesn't even have to be stakes placed. A mare that wins a good handicap at Newmarket, a couple of metropolitan races in Sydney or a maiden special weight and allowance at Saratoga over a normal career will most likely get a CPI of 2.62 or better. The yearling market does invariably expect to get paid for yearlings out of stakes winning mares, but they need not be stakes winners for the advantage to be found which represents another advantage to be had (that is, you probably underpay for the yearling out of the mare that won a couple at Saratoga and wasn't a stakes winner relative to a yearling out of a listed winner at Parx or some other lesser track, even though statistically there isn't much between them).
I'm a fan of what mega-horsehandicapper Dana Parham once famously said…”I strike where I have significant advantage, where I have advantage I have winners.” Finding winners is firstly about finding advantage and if you are able to cumulate a lot of small advantages results in a more consistent selection process.