A funny thing happened on the way to our Kentucky Derby Pedigree Ratings

640x79_derby

Some background: We adore taking reasonably complex concepts and developing algorithms to try to distill those concepts into single numbers that make it easier to understand horse racing. It’s fun! This is the kind of stuff that can help the most-experienced players grasp a race faster and newer players start to understand the game as well. We generally apply this strategy to pedigree ratings; this page explains the general concepts at work here.

In short, what we offer in TimeformUS Pedigree Ratings is a snapshot of the last decade of performances from a horse’s extended family. We get this snapshot by weighing the TimeformUS Speed Figures horses (in each runner’s family) earned at relevant distances and surfaces. No surprise: We think our ratings are an excellent way to speed up the process and allow our customers to quickly understand if a horse is bred to handle today’s race.

The challenge with using this methodology as it relates to the Kentucky Derby is just how few races are run at distances of a mile and a quarter or longer on dirt.

It’s obscenely few.

So stop right there, those of you looking for neat and easy answers about who will relish the stretchout to the Derby distance.

We believe you do not have enough recent data to be confident. How little data is there to go on, over the last decade? Consider the following (all stats are Jan 1, 2004-April 23, 2014 inclusive, a sample size of 544,000 North American Thoroughbred races, not including Steeplechase etc):

65% of the races run in the US in the last 10 years have been run at 7 furlongs or less.

Looking for races that are 10 furlongs or longer, and run only on dirt (you know, our preferred surface here), only .2% fit this criteria. That’s right, more than 1/2 a million races on all surfaces over the last decade+, but only 1106 were 10 furlongs or longer on dirt.

Using a purse size of $50,000 as a proxy for “class,” the number of 10-furlong dirt races of course gets even smaller–only 285 out of a half million+ races. We feel like we saw all of them.

So, what to do when trying to zero in on a 10-furlong pedigree rating for each Derby starter?

Conclusion:

We’re not publishing 10-furlong pedigree ratings in TimeformUS PPs.

Not enough recent data.

We’ll publish findings here, but we won’t put them in our PPs.

OK, the chart with our results first, and then the explanations of all the columns below:

Screen Shot 2014-04-28 at 8.33.52 PM

To build 10-furlong pedigree ratings that are worthy of your consideration even in this format, we considered the following factors:

Column A–Likely Derby contender OR horse of interest whom we thought we’d throw in here.

Column B–This was our first run at 10-furlong pedigree ratings (we do all pedigree ratings on a 100-point scale). It upgraded, on the aforementioned small data sets, pedigrees associated with longer “Classic” distance races. Ride On Curlin, for example, earned some points simply because his sire was Curlin, and Curlin himself was quite successful at longer “Classic” distances. The most interesting thing about column B is its relationship with Column C. SPOILER ALERT: We put the heaviest weighting on this column and really spent the rest of the study being compensatory for its small data set.

Column C–This is our “traditional” dirt route ratings, which consider these horses’ families’ speed figures in ANY dirt race of a mile or longer. As you might expect, there are quite a few horses who rate worse at 10 furlongs than they do at routes in general. The interesting ones are the ones who don’t look so hot in Column C (a general sample of all route races) but look better in Column B (zeroing in on Classic distance racing). Both Danza and Samraat would have been farther down our list if they hadn’t been upgraded for their grandsires–Street Cry and Giant’s Causeway.

Column D–“Purists” aren’t going to like this one bit. We lightly used each Derby candidate’s own top TimeformUS Speed Figure at 9 furlongs, and converted it to our 100-pt scale. Yes, yes, purist, we get why you don’t like it. But we thought it was interesting for these purposes, largely as a way of penalizing any horses who may have impressed in sprints but hadn’t at least matched their top in the longer races.

Column E–Each Derby candidate’s Dosage Profile!!! Oh, it’s not fashionable. And yes, years ago it was presented by some in the media as…a magic singular number. But for our purposes it is useful because it goes well beyond the last 10 years to dig deeper into a horse’s pedigree–you know, back to when they actually ran long-distance dirt races in this country. So while our pedigree ratings zoom in up close, dosage does a better job of looking deeper into the past, and so we use it here.

Columns F through J–This involved taking liberties with dosage. At this point no purists are left reading, so we’ll speak freely. We weighed the “Classic” component of Dosage Profiles more heavily (in an effort to better differentiate candidates), and we converted all of the dosage-related fields onto our 100 pt pedigree rating scale. We’re not even displaying columns G and H–they’re just conversions of dosage to move it onto the 100 pt scale. In the end, column J gives us a version of dosage that attempts to better separate the field. We expect someone to call this “dangerous,” but it’s a dangerous game.

Column K is the lay-up algorithm, the grand finale, the payoff, the ranking. Viva Tapiture! It weighs our 10F pedigree ratings 6x, our dirt route pedigree rating 4x, the re-imagined dosage profile 2x and the top figure earned 1x.  

Voilà.

640x79_derby

This entry was posted in Data Studies, Product features, Race Previews and tagged , , . Bookmark the permalink.

1 Response to A funny thing happened on the way to our Kentucky Derby Pedigree Ratings

  1. Jim meeks says:

    Jim says great! Pay attention there are nuggets here!

    Like

Thoughts?

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s