Why there is a SI cover jinx, and why the wild will make the playoffs
I posted this on the wild/kings thread but I think it deserves its own thread. I want to explain a simple concept called regression to the mean. I start out with the assumption that winning hockey games is due to both talent and luck. Lets apply it to a team's winning percentage during two months of a season (Team A and Team B in February and March). To keep things simple, assume that in both months the average winning percentage of both teams was 69%. We focus on the Team that did well in February, Team A with a winning percentage of 85%. What can we infer from this great winning percentage? 1) That this team is more talented than the average team. However, we can also infer that 2) Team A also enjoyed better than average luck in February, because wins are determined by both talent and luck.
For Team B that had a winning percentage of 50% in February, you can infer that they are both less talented and less lucky during February. Of course none of these inferences you know for certain (it is entirely possible that Team B is exceptionally talented and had an extremely unlucky month). However, the inferences we have made above are much more likely to be true.
Now what if you had to predict the two teams winning percentages for the month of March (given that they will average to 69% again)? You expect Team A and B to retain the same level of talent in the second month, so your best guess will be "above average" for the Team A and "below average" for the Team B. Luck is a different matter however. You cannot predict luck, so your best guess is that it will be average, neither good nor bad. Thus, your best guess of Team A and B's performance in March should not be a repeat of their performance in February.
Team A will be less successful in March than in February, because of the unusual amount of luck they enjoyed in February is unlikely to hold.
Team B will be more successful in March than in February, because of the unusual amount of bad luck they had in February is unlikely to hold.
The more extreme the winning percentage in February, the more regression we expect, because an extremely good winning percentage suggests a very lucky month. The regressive prediction is reasonable, but its accuracy is not guaranteed.
Btw the statistics I used were: Team A (Blackhawks in Feb), Team B (Wild in Feb).
Team A (Bhawks in March) winning percentage of 64%
Team B (Wild in March) winning percentage of 75%
The Sports Illustrated jinx is a perfect example of regression to the mean at work. The SI jinx is the idea that an athlete whose picture appears on the cover is doomed to perform poorly the following season (btw there was alot of talk of the jinx when the bhawks finally didn't get a point in a game because they had just been on the cover). Overconfidence and the pressure of meeting high expectations are often offered as explanations, but there is a much simpler account of the jinx: an athlete that performs exceptionally well in the proceeding season, also had the benefit of alot of luck--and luck is fickle (see also sophomore slump).
Regression to the mean is a mathematically inevitable consequence of the fact that luck plays a role in the outcome of hockey games. There is no causal explanation here. It is not very satisfactory as we want to explain why teams are suddenly hot or cold. We want the causal account. But it is what it is.
Thus, there is variation from month to month with teams. However, I think the mean point for the Wild is in the top 8. So yes, there might be some unlucky games here or there but it is still very likely that the wild will make the playoffs. Downturns happen; sometimes they are just mathematical inevitabilities. Cheer up chicken littles.