2010 Predictions - By The Numbers: Part I

Lets face it: the off-season is no fun at all. Sure, there's the draft (draft schmaft) and the free-agent frenzy, plus the occasional contract veto by the NHL, but really it's mostly a whole lot of waiting around for the puck to drop again. This is especially true for us Leafs fans, as we await anxiously to see whether next year will be the year we make our glorious return to the playoffs.

Well, I got tired of waiting, and decided to do an in-depth analysis of the off-season moves in the Eastern Conference, and try to make a few predictions for next year.

So, I present to you my 2010 Predictions, in 3 parts. In Part I, I will explain my logic, and how I've assessed the moves made to each team in the Eastern Conference. In Part II, I will examine all of the off-season moves for the 14 other teams in the conference (the competition). Finally, in Part III I will examine the Leafs' off-season moves, and compare these moves to the others in the conference.  After the jump, Part I:

First, a note on statistics:

In order to compare the changes to teams, I wanted to use a statistic that was a meaningful way to both compare different additions/subtractions, independent of position, but that would also translate into a meaningful team statistic. After very little thought I decided to use Tom Awad's GVT from www.behindthenet.ca for many of the same reasons that The 67 'Sound used it in his great series of posts (which was also somewhat of an inspiration for my own analysis). You can see 67's summary of the statistic here, but the primary benefits for my purposes are:

• It compares well across positions. This means that it's roughly possible to compare the addition of Colby Armstrong to the subtraction of Vesa (D.S.) Toskala. 
• It's a meaningful statistic that's likely to translate into team success. Offense is measured in goals. Defense is measured in preventing shots. And goal-tending is measured in the ability to stop shots (go figure). It also includes statistics for shootout performance. 

The final thing that I wanted to verify was that this statistic is actually a useful way to assess how  a team performs over the course of a season. I assumed it would, but if this analysis was going to mean anything I needed to be able to make sure that the sum GVT of a team actually contributed to success or failure in the league.

To check this, I regressed the sum GVT scores from every player on every team in the East against each team's 2009-2010 regular season points total. Here is the result:

via farm5.static.flickr.com

As you can see, this formula works really well (90% of variance explained). See that point way out on the right? That's Washington. See the low one at the bottom. That's the Leafs.

Because this is the sum GVT, this gives me some confidence that I can examine the addition or subtraction of a given player and examine it's theoretical effect on a team's point total. That is, simply by adding and subtracting a player's GVT to the team's sum GVT, and plugging the new sum into the regression equation, I can predict how well a team will do next season.

One interesting thing to note is that there are a few teams that fall a little further from the line than others. Specifically, I believe Montreal has a much higher point total than their GVT would predict, and New York and Philadelphia have much lower point totals would predict. This gives me further confidence in the metric (if not the method) as it confirms my belief that Montreal was not as good a team as their record indicated, and that Philly and New York were better teams than their records indicated.

Assumptions

Now, although this seems like it will be a useful way of assessing how a team might do next year, there are a few assumptions I should go over for the sake of transparency:

• First, this isn't thorough. Though I have tried to get all the free-agent acquisitions and current rosters from capgeek.com, I tried to limit my analysis to "impact" players only. This meant that if a player had a GVT of less than 1 or greater than -1, or they were acquired for less than $1,000,000, I probably didn't include them. My assumption was that these players may or may not have a huge impact by being added or removed, but their impact will probably be not greater or less than any AHL replacement or waiver-wire pick-up. Any method will always over or under-extend. I thought I might as well try to not make it too precise and it might be more valid.
• I also didn't do any correcting for whether impact players only played a few games with a team (e.g., if they were a trade-deadline acquisition). I believe GVT adjusts for the number of games played (and icetime), but in some cases this will still under or over-represent a player's impact on their current team depending on how they played for that short period. For example, Dion Phaneuf's GVT with the Leafs was only 2.1, but with Calgary it was 5.6. However, it's hard to say which of these represents the real GVT for the player on a particular team. Furthermore, just because Atlanta lost Kovulchuk before the end of last season, doesn't mean they're not going to feel his loss this year. Thus, if he was an impact player and played more than a few games with a team, I took that GVT to either be added or subtracted form next year's team GVT.
• Finally, I made a huge assumption about the reliability of the GVT, both for teams and for individuals. If a player played really well last year, I can only assume his impact will be similar next season. This is a huge assumption, especially when players change teams. Although the GVT stat accounts for icetime and role, it doesn't account for quality of teammates or competition from what I can tell. So, these predictions will be only as good as the GVT is reliable (or in other words, how consistent a player is from year-to-year and team-to-team). 
• One more. I've obviously only included transactions to date. When things change, I may update my predictions, but I don't want to wait until September to post this.

So, in Part II I will take a look at the 14 teams in the East to see how their off-season changes affected their roster and their likely point totals for next season. I will then compare this with the Leafs in Part III, and try to make some overall comments about what I think this means for next season. I will post Parts II and III later in the day.