Understanding the true performance of your horse

Anyone who has studied any form of data science will be familiar with George Box's aphorism that all models are wrong; some are useful. The same can broadly be said of charts. PhotoFinish.Live is full of charts that provide us with clues to performance, but rarely tell us the whole truth. Where would the fun be in that?

Speed kills

Earlier today I was looking at the current crop of juveniles (two year olds) in season 14. I wanted to see who was the fastest S- graded horse this season, with a view to trying to discern whether there is a pattern to the breeding of fast horses, and where my S- 2yo colt, Enodoc, stacks up against them.

In order to create a fair comparison, I organised the finishing time of every run by every S- this season (14) and mapped those times to the overall list of times for each combination of distance (4-12 furlongs) & conditions (fast to sloppy) - so 9 distances x 5 conditions = 45 separate lists of times. I broke the distribution of times into a 1,000 equal bands. I tried 100 bands first (percentiles) but the higher grades of horses all run in the top 1% (or percentile) of time - so we need 1,000 bands to help better break out their performance. We will call this measure quantile, where a quantile of 50.1 means the horse ran a time that was in the top 50.1% of times for that combination of distance and conditions. Clear as mud?

The top 20 performances (Enodoc nowhere in sight at 64th) for S- graded juveniles in season 14 so far, relative to all performances ever, sees Missile recording the best relative performance, over 6 furlongs in sloppy conditions, with a time that rates in the top 0.1% of all runs in those conditions at that distance ever run by any grade of horse. That is pretty impressive. What is interesting is that the second place horse in the same race, Super Soaker, has the second best relative performance this season. Both were almost a second clear of third place. What is more interesting is Missile and Super Soaker are owned by the same stable, Gaucho's. What is even more interesting - if that was even possible at this point, is that they were sired by the same stud, who's stud fees are probably going to go up, Biggie Smalls, https://photofinish.live/horses/guThJGbHyjmoZhirjGqrM6W6hGX2Uw7mBxZBgrJoYMM, owned by, you guessed it, Gaucho's.

Stop me if you think that you've heard this one before

We could end the story here, having learned something new, and a new bloodline angle to investigate. Biggie Smalls, doesn't have 1/1 blood, what makes his progeny so potent, and what does Gaucho's have cooking in his breeding stables? Biggie Smalls landed on the podium 60 times in 90 races, his FF ratings suggest a precocious juvenile period followed by a gradual but consistent decline over later seasons, with a preference for longer distances than his two sprinter offspring of Missile and Super Soaker. Maybe they will peak early and not follow up in coming seasons. But they didn't inherent their sire's preference for longer races so maybe they won't peak in the same way (I am not sure there is much analysis of peaks to date).

Instead, I think there is a broader point to this tale. Our earlier table shows that Missile recorded the fastest relative time of any S- juvenile this season, regardless of distance or condition, and a time in the top 0.1% of all times across all grades ever. Surely I can look up his charts tab and see that run shining like a New England lighthouse, way above the levels of mere mortal S- juveniles can muster.

Missile's run appears within the box of the boxplot of his time chart, which means its not in the top 25% of all runs for S- grade horses, despite filtering for Left-Dirt races on the filter at the bottom of the page. We should expect to see amazing performances like this right on the whisker of the boxplot - right at the top.

All charts are wrong, some are useful. This chart can hide a great deal of insight if you are not careful with it. Firstly, you can only filter for surface and direction, or condition - not both together. The result of this is horses that revel in slower conditions, look slower than those that like firm conditions. Yes you can click on soft as a filter, but then you lose the surface of filter on dirt, and dirt races are significantly slower than turf ones. We aren't interested in raw times at different conditions, we want to know how well our horses run relative to the conditions. A top 0.1% of times for 6 furlongs, sloppy, dirt, should be right at the top of the range, not within the box.

How can we use this knowledge?

For new stable owners, I would recommend that you separate the factors when considering where to run your horses optimally. A horse with soft preferences will often still run faster in yielding or even good conditions - but their performance will be relatively worse than when they run soft or sloppy. It's much harder for a sloppy conditions specialist horse to have their dots higher up in the boxplot - because the boxplot is just looking at a bundle of conditions and just measuring time.

One thing you can do is triangulate this information with the FF rating chart. We discussed the curious relationship between time and FF rating in this post https://aipaddock.com/the-difference-between-ff-rating-and-finishing-time. Remember this is a game of partial information, and the fun is in seeing the absolute performance and not being distracted by the noise. The aptly named Missile should have readings off the charts, not in the 2nd quartile of its distribution.

When looking to buy horses on the marketplace or through claimers, take a close look at Dirt-Soft horses. I think they often look slower than they actually are. Turf-Firm horses are always going to look like they post good times - but given the way that boxplot chart is created, we now know that its harder for a Dirt-Soft horse to post times in the upper quartile (above the box). Maybe someone has put a great Dirt-Soft horse up for sale not knowing their true value.

Another tactic you can use is to leverage community tools like https://photofinishedge.com/. Type in the name of your horse and you can see an adjusted boxplot (bottom right) that takes conditions into account when plotting times for each distance. Attention to detail is the difference between winning (and profitable racing) and losing in a game like PhotoFinish.Live.

Join the fun and put these insights into practice at PhotoFinish.Live and if you are considering starting your own stable please consider using my referral code: PADDOCK or just click on this link: https://signup.photofinish.live/?referralCode=8EUMC4P2

Please remember this is a web3 game where your spend your own money. Nothing I write about should be considered financial or investment advice.