Regression and (arguably) the gambler's fallacy
Reader Tony asked a good question this week, and after writing a lengthy response in the comment section, I decided it was worth posting here for others to see. Tony's question: I know we always say that things even out over the season but do you think it takes that long? I wonder if you measured a player's expected points over a 1 or 2 month period would it match up to his actual points? We could therefore start to predict when a player is 'due a big win'. Being 'due' is a tricky proposition. Although some fans don't like to reduce their beloved heroes to a statistical probability, that is all we're really doing on this site and once you make that leap, the idea of being 'due' kinda falls into the gambler's fallacy (the notion that after rolling 'black' five straight times, the roulette wheel is bound to come up red). I say kinda because it depends whether you believe that (a) each player has little or no control over how his sho...