Last year, our regression to the mean prognostications doomed Henrik Sedin and the Washington Capitals. This year, as commenter cajuncook noted, it’s Dallas who’s taking a beating.
On christmas day, Dallas had 46 points and was 2nd in the NHL standings; New Jersey was dead last with 20. I noted that Dallas had a ridiculously bad shot differential, while New Jersey's was very good. Clearly score effects come into play to some extent, but there's no question that Dallas is much worse in the possession department than New Jersey.
So I asked for input – how many more points would Dallas get over the rest of season than New Jersey? I made a bit of a top-coding error…I assumed that no one would think that Dallas was 10 wins better than NJ, but 20% of voters thought Dallas would put up 20+ more points in the standings. The average estimate was probably 6 or 7 points in Dallas' favor.
So how have things turned out since that day? New Jersey has had 24 points in 20 games, while Dallas has 19 in 18 games. In order for Dallas to hit what 20% of voters thought they would, they'll need 25 more points than New Jersey the rest of the way (29 GP for Dallas, 28 for NJ). If New Jersey puts up a pedestrian 28 points, Dallas would need 53 points over 29 games, which isn't going to happen. What's more likely is that New Jersey puts up something like 34 points and Dallas very much fails to hit the 40 or 41 projected by the community.
Does the mere mention of a team being a candidate for regression to the mean cause them to regress to the mean?
