Note: A good friend and loyal reader asked me on Twitter what the GVT stat means in my previous post. I don't profess to have invented it; that credit goes to Tom Awad and Hockey Propectus. However, I have researched it and value it as a great all-around metric for determining value.
GVT is to hockey what VORP is to baseball. It strips away the winning and losing aspect of hockey and equates player statistics to goals. Regardless of a team's performance, all a player is trying to do is increase goal differential in his team's favor, whether offensively or defensively. He can't control what happens when he isn't on the ice, even though that might make the complete difference towards the outcome of the game. GVT takes a player's individual contribution and scales it relative to replacement value. We can then use that to predict how a team will do.
For example, let's take my "perfect small market team" from the last post. Let's look at Jonathan Toews. His GVT last year was 20.8. That means he contributed, through offensive, defensive, shootout, and goaltending totals, +20.8 more to the Blackhawks' goal differential than a replacement-level player would. His breakdowns are as follows:
OGVT: 13.7
DGVT: 5.7
SGVT: 1.4
GGVT: 0 (obviously, Toews is not a goaltender)
Total: 20.8
These totals isolate Toews' performance to the various areas of the game. What stands out to me: 5.7 DGVT is very good for a forward. Selke winner Ryan Kesler also posted a 5.7. Noted defensive guru Pavel Datsyuk posted a 2.3. Bottom line, everyone knows Toews is a great scorer by looking at his points numbers, but GVT can show us a real boost in his value because of his defense. This is why Toews ranked as highly as any skater in the league last season despite lagging some 15-20 points behind players like Martin St. Louis, Corey Perry, and the Sedin twins.
We can also use GVT to predict team performance. I emailed Mr. Awad about the calculation this morning, and he was gracious enough to respond to me. The first step is to sum the player GVT contributions of all players. Again, reference my perfect small-market team. The total player GVT was 258.7. However, this doesn't represent team GVT, since we have to account for replacement level. Replacement level is defined as 1.5 goals per game, or 123 goals for the season. We subtract this out from the player summation to find that my perfect small market team would have a goal differential of 135.7. As mentioned, this would be the highest goal differential since the lockout by a wide margin.
But that's not all. We can then scale this predicted goal differential to get a predicted point output. Mr. Awad was extremely helpful in sharing the formula with me, so credit to him here. 3 GVT = 1 point in the standings. So our imaginary team of 258.7 GVT would have a predicted point output of (258.7 divided by 3) 86.2333 points above replacement level. We then add in replacement level of 50 points to get a predicted output of about 136 points for my imaginary team.
How was that 50 points derived? Remember that the replacement level GVT is 123 goals for the season. Divide that by 3, and we get 41 points for a replacement level, 123 goal team. The exact average point total of a team since the lockout is 91. Subtract 41 from 91 and we get 50.
Important: A team full of 0 GVT players would not produce 0 goals! They would just produce 0 goals above replacement level. A team with nothing but replacement level players would produce a goal differential of 1.5 per game, or 123 for the season. Anything above that results in a positive GVT. Anything below that results in a negative GVT. That's why, just like VORP in baseball, we can see players with a negative GVT, and why it's possible for a team with a positive GVT player summation to have a negative team GVT: if a team's player GVT summation doesn't exceed 123, we would expect that team to have a negative goal differential for the season.
A great story appeared in the New York Post today about the hockey sabermetric movement. It's also worth a read.
No comments:
Post a Comment