[keenrodent]

Without rewriting the article, the point is that there's an exploitable gap between the output of modern prediction systems and athlete performance that is explainable by information not accounted for in prediction systems.

The writer suggests that it wouldn't be too hard or expensive for someone to gather this "extra" information, and with enough of it and enough study of it, it's likely they could come up with a winning system over automated prediction systems.

From my own experience around MLB and MiLB teams I know that teams do this themselves all the time. They look for pitch patterns, pitch tips, changes in mechanics, changes in vision, changes in performance within games and within seasons due to tiring or wear-and-tear. A bettor with some resources could collect this same information and gain a meaningful advantage.

For a famous example to build on the other examples in the article, in 2006 everyone in baseball except Brad Lidge knew he was tipping his slider [0]. This kind of thing, if your net is spread wide enough, gets discovered all the time. A savvy bettor or syndicate or bettors could come to know this. Teams certainly did.

Anyway, my $.02 cents.

[0] http://www.chron.com/sports/justice/article/Lidge-was-telegr...