Wednesday, 11 January 2012

Comparing Point Allocation to GVT

As you may know, Point Allocation, which we've been discussing lately, is not the only comprehensive player valuation system available. Perhaps the best-known system is Tom Awad's Goals Versus Threshold (GVT), used extensively at Hockey Prospectus. If you're familiar with the scale and range of GVT values for modern players, you might be interested in how these figures would translate into their equivalent Point Allocation (PA) values. So here's a quick and dirty system you can use to convert between the two systems.

First, you have to realize that GVT is normalized to 82 games and does not normalize for minutes played, whereas PA used an 80-game basis and normalizes a player's minutes to 20 per game for forwards and 25 for defencemen. That's the first adjustment that needs to be made.

When that's done, you need to multiply the resulting GVT by 0.4 for forwards, 0.7 for defencemen or 0.8 for goalies. (If you're going from PA to GVT then the multipliers are 2.5, 1.4 and 1.25 respectively). The fact that there is not one constant multiplier that works for each position indicates that GVT and PA value positions differently. Specifically, from PA's perspective it seems that forwards are overvalued by a significant degree. But by using these differing multipliers, you can account for some of the systematic differences between the results.

To illustrate, let's convert Martin St. Louis' 2003/04 season from GVT to PA. We know he played 82 games and 1,686 minutes and recorded 27.3 GVT. First we adjust for the games and minutes, giving 80 games, and 1,600 minutes, and 25.9 GVT. We then multiply this by 0.4 (since he's a forward), which results in 10.4 Total Points Allocated (TPA). This is a rate of 6.50 TPAK (TPA per thousand minutes), which as we already know is certainly the level of a superstar performance, indeed at the upper range of that.

No comments:

Post a Comment

Hostgator promo codes