Ramblings, and What a View The American, Race 8 at Santa Anita, April 2018


Had a chance recently to read some blog posts on machine learning exercises related to horse race handicapping.* It's a natural thing for grad students to do - search for a problem that may win friends and influence people.

The connection between horse racing and machine learning is natural. There is plenty of data to work with, a clear interest in prediction, and races - even the worst of races - have certifiable winners.

The student write-ups cite the usual machine learning painpoint - how much time and effort is required to normalize that data and the difficulties in correlating the different  variables.

If in the matters to be examined we come to a step in the series of which our understanding is not sufficiently well able to have an intuitive cognition, we must stop short there. We must make no attempt to examine what follows; thus we shall spare ourselves superfluous labour. - Descartes

What the researchers quickly alight upon is how good the crowd is at finding the winner - and how that mere fact makes the undertaking unprofitable. Your algorithm has to do  than more than pick the likely winners - it has to pick the unlikely winners. The machine learners that succeed in handicapping still lose money overall - because the price on their winning picks fail to offset their losers.

There already are statistical methods out there that quantize much of the data on horses. It might be useful for the students to take those for givens, and focus on one especially vexome area for their machine learning test. Pace might be such a place.

                  -*-*-*-                                                -*-*-*-                                -*-*-*-

Epitomime has focused on four areas of handicapping: class, speed, form and pace. The last two seem particularly hard to master. For the analysis exercise for The American Stakes, a $100,000 Grade 3, 1 mile event on the turf at Santa Anita (Race 8 on Apr. 21, 2018), the focus was on the pace. I think I have some familiarity with some of these horses - having seen them both win and lose does enforce some small amount of objectivity. So I will give shorter shrift to class and speed and form, or look at these with a special emphasis on pace.



Proven 7-year-old campaigner and old-fave Manilla line progeny What a View (2) has post position and speed to get a jump on these. If back class surfaces he could steal this mile race. Black Jack Cat (5) and Sawyer Hill (6) should track, as with others. Again, as with What a View, it is a case of trusting back class to pick Black Jack Cat over Sawyer Hill (stepping up) or the others. A particular possibility is Kenjistorm (9) who has raced with several of these, who may be said to be formful, but who may not have proved himself at this level as compared to WaV and BJC.  BJC had a great string of victories in 2018, topped by a fantastic third on Breeders Cup Day. The lone late runner in the group has some prospect in that it's expected to be a fast (and possibly contested) pace up front. The lone latter is Colonist (3) who has merely beat the likes of Milton Freewater but who has some miraculously improved numbers of late,  has a late kick and gains the services of Drayden Van Dyke.

The wager is: 2, 5 over 2, 3, 5. Win, 5; Place 2. (These latter bets serving as place markers to force bread crumb evidence of selectivity and establish which horse we thought most likely to win. The crowd gave the nod to Black Jack Cat too, sending him off at even money.)



The result: 2-3-5-9. Paying $40.20 for a $1 exacta. And $7.80 to show (and missing the Win payoff of $14.20.)

The Kenneth Black-trained What a View took the early lead, the others pressed, but not too menacingly. What a View ran so nicely on the turns! And, tho Colonist made up ground and came on well, What a View won by a neck in 1:33 and 4 for the mile. What a View's jockey Steward Elliot - he of the New England accent that harkens back to his and our early Suffolk Downs days (when Dan Bertucci[?] would call him 'Stuey" -- rode him masterfully. Setting the trainer's preferred half-mile fraction of 46 (actually 46 and 2). Black Jack Cat "weakened some" along the way, and did not seem to have his 2018 fire -- yet. Clearly, it wasn’t just ''pace'' that led to What a View's inclusion among selections by this "machine learner," or his win in the real world.. but it did run on the ground as it seemed to look on the paper. -- Racetrack Romeo

What a View increases his lead on the far turn
in The America, April 21 at Santa Anita.

The aftermath of the race saw winning trainer Kenny Black just about in tears during post-race interview with Britany Eurton. Tears of joy, one guesses, and pride in a horse that endeavors.

"Where are you going next?" she asked, referring to the horse's next outing of course. He couldnt really answer such a base question, then comically shrugged.

"To the bank, I guess," said Black. I thought him sporting and, as I said, truly tearfully appreciative of the horse and the jockey, too.

All particularly poignant on a day marred by two equine fatalities in stakes - old faithful Bullard's Alley at Keenland and young turk Ten Blessings at Santa Anita. Mis-steps always lurk.

                  -*-*-*-                                                -*-*-*-                                -*-*-*-

Memorable was the pre-race interview between Eurton and Black, where she asked the question always asked, "what did you tell the jockey?"

While he can talk to them for days about the horse, Blacker said, he never gives strategy instructions to the jockeys. Why?

"The good ones don't need them, and the bad ones can't follow them."


                  -*-*-*-                                                -*-*-*-                                -*-*-*-

* I'd have liked to have used an analogy here based around the individual generally described as the first programmer -- that is, Ada Lovelace -- but a Wired story has convinced me that, while Ada played the horses, the idea that she used Differential Machine match to predict outcomes is likely conjecture.

Racing!



Watch the video while you follow Chart: http://www.drf.com/race-results/tracks/SA/country/USA/date/04-21-2018



Comments