Handicapping with the Anecdotal (Hint - the CAW's Don't Do it)

I saw one of those "trending stories" things this morning on twitter, about a mattress firm. We've all seen these stores in strip malls. 

What fuels the conspiracy theories (and my favorite comment - "Who owns these stores, Walter White?") is the anecdotal. Namely, I don't know anyone who ever bought a mattress there, therefore something must be up.  According to the firm's financials, the stores do fine, and with high margins, only need to sell a handful of mattresses a week to be profitable. 

We do this very often in horse racing. 

A rider we don't like is replaced by a one of our favorite riders and the horse wins. The reaction, because of this visibility bias, is that this angle will work every single time. However, as Crunk often points out on twitter, over a large sample, the move is often lateral.

Last night I heard an angle that I've heard over the years at (rainy) Mohawk about driver James MacDonald: "He hates driving on off tracks and is ineffective on them".  

I decided I'd check that angle over the last few years, and:

UDRS on off tracks: 0.339

UDRS overall: 0.338

If we're really diving in, his ROI does decrease a little on bad tracks, but it could be noise or sample size. Call me in three years when we have enough. 

For a generation we heard blinkers on was a great angle, because it kind of makes sense as the horse will be sent, right? And again, for years when we examined global samples, blinkers on actually underperformed blinkers off as a betting angle.

I told a story on Chris's pod that back in the day I bought a book at the Gambler's Book Shop in Vegas titled (something like) "20 Foolproof Horse Racing Angles". Not long after, I had a sizeable database and tested them the best I could (some were not totally testable). Not only were the '20 foolproof angles' not close to profitable, some of them weren't even angles that had a positive impact value. 

The CAW's and other players have, over the years, exploited many of the anecdotal angles we've all used and heard. They generally don't deal with the one, they deal with the many. 

You and I are not CAW's, and most of us think Python is a snake not a programming language, so we'll never be able to verify every single angle or thought we hear from horsepeople, public handicappers, or our buddies at the OTB. But if we want to beat the game, I believe we have to "trust but verify" as much as we possibly can; at the very least, in the words of Barry Meadow, be a "Skeptical Handicapper".

In the end, sometimes mattress stores are in business simply because they sell mattresses. 

Have a great Monday everyone.