Wednesday, August 26, 2009

Hockey is Coming

I can smell it in the air. I can feel the days getting shorter, the weather slightly cooler, the leaves are starting to fall from the trees, and in a few short weeks NHL hockey players will be reporting for duty! I spent this past weekend nestled up to my computer creating my NHL Forecast 2009 spreadsheet. It is a vain attempt to predict the future based on past performance data and player age.

The first step is a review of how players evolved in 2008 from 2007 and establishing a rate of improvement or decline based on age. I am sad to inform those over the age of 27 that it is likely that you have passed your peak in physical performance capability. 28 and 29 are relatively stagnant, and 30+ are generally in decline. There are exceptions to every rule, but on average this is what the data tells us. Once you have your parameter values for rates of change, it is time to gaze into your crystal ball and start predicting outcomes.

The rate of change data has a relatively normal distribution, and as such we know that approximately 2/3rds of the players will be within one standard deviation of the expected performance and 95% will be within two standard deviations. Then go through rosters team by team and assign a unitless Z-Score to each player based on whether a player should improve or decline due to ice time, line mates, and power play time. Do a little reverse algebra where you multiply it by the standard deviation, add it to the mean, and out pops a prediction! I suppose you have to be more than just a mathematician to arbitrarily assign a parameter value based on imperfect information; fortunately I am able to do so based on my life-long addiction to hockey statistics and information, and my childhood education from my father who played professional hockey. Oh, and my Degree in Mathematics also helps...

To anyone competing in a hockey pool this season, you too can do this at home! All you need is 20 hours to compile a database, set up a series of equations based on the instructions provided above, meticulously research every scrap of information in the public domain, and voila, your own homemade forecast! It is so easy even a child could do it (I know because I began forecasting as a child). If you actually have a life and don't have the time, the Fantasy Hockey magazines have hit the newsstands. I have purchased two, the Hockey News and The Score. I personally recommend the Score Forecaster. It is evident upon detailed analysis that they employ mathematicians who play a role in establishing their predictions. All they need to do is ask their hockey "insiders" to give them a number between 1 and -1 for how good a season will be for any given player, where zero means exactly the same. They wouldn't even need to account for age, because the mathematician can do that in the regression analysis.

The Hockey News magazine on the other hand seems to swing for the fences more often. The Score, like me, seldom makes a prediction outside of one standard deviation, but the Hockey News does it several times. 30 point swings will occur, but they are relatively rare and predicting them is very difficult. If Zach Bogosian scores 50 points, THN will be able to boast about its forecasting acumen, but in reality there is only a 15% chance or less that a player of his age and his career arch see that big of a jump in production. It happens and they might be right, I just would not count on it. When one swings for the fences, their strikeout rate also climbs. Plus the score evaluates more players and writes a more detailed analysis of each player.

The Iceman will be shifting his creative focus from politics to sports in the build up to NHL training camps. For those looking for a fun place to play fantasy hockey for free online, the best in the business is Yahoo. Registration has not yet begun for the upcoming season, but perhaps I will start a league for Blogging Tories?

1 comment: