Big Data, Small Data (II)

We wrote about using big data a few months ago in the context of remembering to use the qualitative as well. I won't rehash it (as I tend to do with this 3,000 post blog) but I will offer an offshoot post from something I read today. 

Jeremy Balan tweeted a neat thread about Zillow and its home pricing algorithm. This "big data" algo worked super-well with back data and in real time for awhile, but when it came to buying houses in real time it cost the company a whole pile of money. 

It's a really good thread as to why, and I think it's important, for us who play racing. 

Big horse racing data is used by the teams, and it can see what we do not. Many #theyknew horses are not #theyknew as in inside money, but more the AI algo knows. As well, with rebate, many horses are bet down where you and I aren't going to play them even if we like them; but the algo says it's a play. 

One area, as the thread author alludes, where big data can fail is the qualitative as we talked about in the last post. 

But another, in my view, is that it simply isn't as nimble with short term trends - track changes, and trainer trends are two of them. I personally love playing dead front end bias tracks, because I look at where the teams are, and they seem to be still betting speed. If a move-up trainer is ice cold because there's a rumor someone raided a barn, they're still, again, in my view, too well bet. 

I know the teams and their algo updates and is aware of these things. I know they dot i's and cross t's when things change. But if we're good handicappers, I can't help to think we'd be better at it. 

I took an interesting seminar today with something along these exact same lines, and we do see it even with multi-million (billion) dollar algorithms for marketing and business. They can be slow. There are holes and there are blind spots. There are things we can do better than they do, and they are recognizable. 

In horse racing, from Beyer to Crist to Cramer, we've often heard how we need to be aware when we have an edge. This was in the context of the person sitting next to us. Today it's much harder, but recognizing when we do have an edge on the machines and acting upon it is the same thing. It's just a different century. 

Have a nice evening everyone.