How good is the coaches’ poll? (revisited)

Back in 2008, as the excitement over the impending release of the WCHA polls was building, I found myself wondering how meaningful the polls really were. We all get excited about the predictions, as if they will somehow influence the outcomes, but then forget about them by the end of the year when PWR is all that really matters. So, I decided to look back a few years and compare the coaches’ predictions to the actual outcomes, which I wrote up in the post, How good is the coaches’ poll.

That analysis determined that the coaches were actually pretty good at picking WCHA outcomes. They beat throwing darts at a dartboard by a huge margin, and even consistently beat recycling last year’s finish as a prediction for this year. Since a couple years have passed, I decided to take another look and wasn’t surprised to see that the coaches have done even better in the last two years (here’s the raw data).  In 2008-09, the coaches picked every team within one spot of their actual finish except CC (picked #1, finished T-3).  In 2009-10, the coaches picked every team within two spots of their actual finish except Minnesota (picked #3, finished #7).

However, the “previous year’s finish” predictions also did quite a bit better in the last couple years.  That means that the good teams are staying good and the bad teams staying bad, making things a lot easier to predict.  Here, fully updated, is how far off the predictions were from the outcomes (lower numbers are better):

Year Coaches
Poll*
Previous
Year*
2009** 28 43
2008** 12 42
2007 65 84
2006 52 43
2005 21 48
2004 86 89
2003 62 146

* – calculated as the sum of the squares of the differences of each prediction with the actual outcome
** – new since original post

So, it seems that even in the easy years (the “previous years” outcomes of 42 and 43 were better than the coaches outcomes of the previous two years!), the coaches add significant insight beyond how good the teams were last year.

Once again, here are the coaches’ picks for 2010-11:

1. North Dakota (10), 120
2. St. Cloud State (2), 105
3. Minnesota Duluth, 100
4. Denver, 88
5. Minnesota, 82
6. Colorado College, 67
7. Wisconsin, 64
8. Nebraska Omaha, 55
9. Bemidji State 46
10. Minnesota State, 41
11. Alaska Anchorage, 19
12. Michigan Tech

Leave a Reply