|
Yesterday some friends and I went to a horse race held nearby mainly as a way of socializing and spending a Sunday afternoon, as none of us are really into horses or betting. I don't think it was a particularly significant race either, like Kentucky Derby that I somehow heard about despite not even living in the US. We tried betting some small sums for fun (like 5$), I lost everything and my friend won enough to buy himself a cheesburger. What struck me though was that most people, including the friend who won and most others I've seen around, made their decisions based on completely useless factors like the nationality of the horse, the jockey, or its name. And to be fair how could you not bet on Private Ryan (he was left behind). So it seems like it should be trivial to do better than that. For example by estimating the probability based on recent results of the horse, plus factors like rest time, different jockeys, weather, altitude and what not. Models aren't perfect but have to be better than using the color of the horse to determine the winner. Still, I can't be the first one with this idea, so what's the catch?
|
# ¿ May 15, 2017 09:19 |
|
|
# ¿ May 2, 2024 14:59 |
|
Thanks for sharing, that's very interesting! Have you noticed some people winning consistently? What I'm seeing here is that it's pretty much like betting on the roulette, there's a lot of variability in every horse's performance, making it difficult to come up with any sensible guess. Still, I wanted to see just how (in)consistent the performance is but apparently this information doesn't exist? The best I could find had finishing positions and winning times, which is next to useless. Like for F1, you can get exact individual lap times for all drivers in each race, for example, so I was hoping to see something similar. Is this like a trade secret or something?
|
# ¿ May 16, 2017 13:41 |
|
^^^ A one-eyed horse named Patch would definitely get my bet, which is also how I would lose my money This stuff is really interesting once you start digging into all the various details! My initial idea seems to be dead in the water due to unavailability of data, but it's fascinating to learn about it nevertheless. I had no idea different countries did it so differently, or about the crazy mechanized totalisators. The thinking with the model was that since the odds are at least partially influenced by other (mostly clueless) betters, it should be sufficient, and not too difficult, for the model to be just slightly better then random to result in positive expected value in the long term. You don't need to predict the winner perfectly every single time, if you can reliably average 110% returns over the whole race day. Just a clear history of how the horse performed at a given distance previously would make a big difference, and the weather and surface conditions could narrow the guess down further. Plus all the other factors Pikestaff mentioned. This information might be possible to get with some upfront investment, but I probably wouldn't bother just for shits and giggles and instead keep enjoying mainly as a nice opportunity for socializing.
|
# ¿ Jul 4, 2017 20:09 |