Fooled By Randomness

We as horseplayers, industry watchers, and well, humans, immediately want to look for reasons for something when it happens. As someone much smarter than I wrote, we look for causality in just about everything. The problem with that is, much of the time there is none.

Today it was reported that Belmont viewership was down over 15%.

That was because, of course, there wasn't a Derby runner in the field, that the field wasn't deep enough, the card wasn't as good, and maybe people didn't really like Andy Grammer. 

Maybe one of those are correct, or (more likely) none of them are.

The fact is, this is a one-off horse race, with little long term data to compare it to that makes any sense. The result could be completely and utterly random.

This is a characteristic of small data, and we fall for it often. If a trainer is 8 for 11 over three years with a move, we should bet him each and every time thereafter. What often happens, is the trainer then goes 0 for 14 and we lose our shirts. We then say "I guess he's no good with that move anymore," when he was never good with "that move" in the first place.

Data only gets actionable, if it's actionable. A 72 hole golf tournament is pretty good at identifying the best golfer, because that's plenty of holes to smooth the random. But even then it only identifies the best golfer that week. A cricket match, lasting days, probably does a good job; certainly much better than a 7 game series in other sports. But it still signifies something within an event only.

The bottom line with the Belmont results - and before that the Preakness - is that many millions of people watched a horse race on television this spring. When we look at that in the context of the "big day" handles and interest, buttressed with the apples to apples long term trend, this is a positive number. There are a lot of issues in horse racing to fix, but it's doubtful this is one of them.