Wednesday, August 4, 2010

A Crude Way of Rating Coaches

In the shower a few days ago I was thinking about the post I'd finished the night before on Tomas Fleischmann and goals-versus-threshold, and how I'd added up the GVTs of the Capitals players. That got me thinking, what would the Capitals look like if the GVTs are removed? Since GVT is calculated in goals, if we take away all these GVT values, we take away a certain amount of goals (a combination of adding goals allowed or subtracting goals for), to give an estimation of what a team with average players might look like. Then I thought, what if I repeated for each team in the league? I'd probably find different results, though I don't know how much the results between teams might vary. I guess once we remove the players then we get to the core--system, and through the system, coaching.

I will subtract the total offensive GVT of the players from the goals for amount to get a rough goals for of the average team, and add the total defensive GVT to the goals against number to get a rough goals against. Using Pythagorean Winning Percentage I should be able to find which team's system is best. That team's coach could have been the best in the league in 2009-2010.

Some expectations I have before I begin:
  • Bruce Boudreau won't show up very highly, but not at the bottom either.
  • Barry Trotz, Joel Quenneville, Jacques Lemaire, Dave Tippett, and Mike Babcock will show up near the top.
  • Jacques Martin and Rick Tocchet will be near the bottom.
Without further ado, the results (program and results):
1    NJD    29.977%    170.3 GF    281.7 GA   
2    CHI    29.879%    170.1 GF    282.0 GA   
3    WSH    29.622%    170.6 GF    284.5 GA   
4    DET    29.616%    170.4 GF    284.2 GA   
5    LAK    29.535%    170.5 GF    284.9 GA   
6    BOS    29.489%    170.5 GF    285.2 GA   
7    CGY    29.403%    170.7 GF    286.1 GA   
8    PHI    29.357%    170.4 GF    285.9 GA   
9    CBJ    29.347%    170.3 GF    285.8 GA   
10    NSH    29.201%    170.5 GF    287.1 GA   
11    OTT    29.191%    170.4 GF    287.0 GA   
12    NYI    29.176%    170.4 GF    287.1GA   
13    SJS    29.161%    170.4 GF    287.2 GA   
14    VAN    29.071%    170.4 GF    287.8 GA   
15    MTL    29.071%    170.4 GF    287.8 GA   
16    PHX    29.006%    170.2 GF    287.9 GA   
17    TBL    28.931%    170.2 GF    288.4 GA   
18    BUF    28.938%    170.7 GF    289.2 GA   
19    MIN    28.863%    170.4 GF    289.2 GA   
20    PIT    28.812%    170.2 GF    289.2 GA   
21    NYR    28.736%    170.6 GF    290.4 GA   
22    DAL    28.686%    170.4 GF    290.4 GA   
23    EDM    28.656%    170.4 GF    290.6 GA   
24    STL    28.631%    170.3 GF    290.6 GA   
25    TOR    28.612%    170.4 GF    290.9 GA   
26    ATL    28.587%    170.3 GF    290.9 GA   
27    COL    28.491%    170.5 GF    291.9 GA   
28    ANA    28.404%    170.5 GF    292.5 GA   
29    FLA    28.0240%    170.3 GF    294.8 GA   
30    CAR    28.006%    170.4 GF    295.1 GA   

Note: I did not incorporate shootout statistics and used goalie GVTs as purely defensive

Not all too surprising I guess--the good teams finished near the top, the worse teams near the bottom. The spread, however, is pretty small (in only one season, it should be noted). Philadelphia and Columbus changed coaches (among other teams), so that may have lead to improved numbers later in the season, but since GVT is dependent partially on games played, the players' GVTs may not have had time to catch up. The differences lie in goals against numbers, as goals for are almost even across the board. It looks like there is no one best system, but rather, a coach succeeds when he has good players to work with.

My next project will be to use players' historical GVTs over the past three seasons or so to better judge what sort of talent a coach truly has (specifically I have Philly in mind). This exercise over one season though looks inconclusive.

No comments:

Post a Comment