3 ms·
The Case for Human Intelligence
- sharemywin 10y agoDid someone actually predict something with this "human intelligence" if not than what's the point from a prediction perspective.
- keenrodent 10y agoWithout 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-telegraphing-his-pitches-1640733.php http://www.chron.com/sports/justice/article/Lidge-was-telegr...
- chrisherd 10y agoClassic case of always doing the same thing because it's the way things have always been done. Only takes one person trying something differenct to start a 'revolution'...
- theobon 10y agoThe underlying issue appears to be that prediction systems do not take into account recent changes. Which is common sense, a prediction system is going to lag. It makes a cute fluff article about the value of beat reporters when dealing with baseball stats. It can destroy lives when it comes to predicting if a person will default on a loan or is likely to re-offend.