Hockey Pool Predictions Consumer's Guide

Because I'm so nice, I thought I'd share a bit of info with you guys. That way you can be Wicked Smaht like me.

There has been a lot of talk about my mighty database, which I'll admit was quite mighty.

The database was used last year to garner a 3rd place win in last year's edition of Superdraft. The way I see it, it can only be fluke that I didn't win! After all, the database was used to successfully win Barry's Charity Draft. It was also used by Leaker to win the prestigous Goldcup Draft. (Leaker won a hockey draft, you say? Well yes, by using the SuperDataBase.) It was also used by Broker Bill to solidify a middle of the pack placement in his own office pool. (A success, since he usually finished in the bottom 5%???) OK, so I guess that's enough trash talk for now...

Anyways, I'll take this opportunity to unravel the mystery of the SuperDataBase. The database was built from the so-called-expert hockey pool magazines. I used McKeen's, The Hockey News, and Forecaster. I spent a significant amount of time entering the G-A-Pts data from each of these sources into a big spreadsheet, combining it with Salary information, and completing some fancy-schmancy data manipulations. In the end, I used it to pick my team, and also to predict the outcome from Superdraft2005. Perhaps I'll discuss that in another post.

Now a year later, we know how many points each player actually produced, and I figure I owe it to myself to look back and see how things turned out. So, just how accurate were the predictions? Shown below are the separate charts for the top 50 forwards and defensemen. The blue lines are the actual points, the pink are predictions from McKeens, the yellow are from The Hockey News, and the aqua are from Forecaster.

For the forwards, it appears that the experts underestimated the effects of the new rule changes. Enforcement of the rules lasted the whole year, and scoring was up. For defensemen, the predictions were OK in general, with the exception of the top 5 which were all bad! In examining the charts, it looks like all publications rated players in the same fashion. If one predicted wrong, they all did.

If you are interested in getting into the data, please click on the title of this post, which will bring you to the Excel spreadsheet.

For those of you more into the actual statistics, please refer to the following table, which shows the average and standard deviation of the absolute deviation of total points for various categories of players:

How you read this is, for the top 175 forwards in my database, McKeen's predicted the total points to be 15.3 different from actual, on average. The lower the number the better! The standard deviation indicates the spread in the data. We have 95% confidence that the number is within +/- 2 standard deviations. The lower the standard deviation, the more consistent the prediction deviation to the reported average. There is a whole lot of spread in the data!

Again, in looking at the data, it appears that each publication is equally bad. It also shows that as you look closer at the top players (top 25), the predictability gets worse.

My conclusions:

1. Even Leaker could write for these publications (if he could fix his grammar).

2. Hockey predictions are not guaranteed.

3. I will probably continue to buy these stupid prediction magazines, no matter what the data tells me.

4. I will be altering my approach this year. Tune in next year at this time to find out how I did it!

Iffy