2006-07 Goal Differentials

[Doctor No]

These are from the just completed 2006-07 NHL regular season.

Goal Differential is a metric estimating how many goals better than average a goaltender was over a period of time. Essentially, it's a weighting of save percentage and playing time (because a save percentage of 93% is obviously much more valuable if you can sustain it over a long period of time).

It's described more in depth here: http://hfboards.com/showthread.php?t=253883

Please note that if you don't like save percentage, then you won't like these. Save percentage is not perfect (and I'm not going to defend it here).

At the behest of some people who know what they're talking about, I've removed a goaltender's shots/saves when calculating their league save percentage. In other words, if Dominik Hasek does well, that increases the league save percentage and would penalize him under this metric. Therefore, to calculate Hasek's Goal Differential, I remove his stats from the comparison set.

I've also calculated Goals Above Replacement Level, because it's not really fair to compare goaltenders to the league average. A "league average" goaltender clearly has value - many of them get several million dollars per year just to be average. Goals Above Replacement attempts to estimate how many goals a goaltender is worth compared to someone available for free (e.g., a top AHL goaltender or a free agent from Europe). Based on some conversations on the Hockey Analytics Group, I'm estimating replacement level at a save percentage 1.5% lower than league average (so if the league average save percentage is 91.0%, then replacement level is 89.5%).

Click to read more...

 

[Doctor No]

[font=courier new]
GOALTENDER       TEAM       GD      GAR
Brodeur           NJD    36.87    69.60
Luongo            VAN    35.59    68.13
Kiprusoff         CGY    27.35    60.20
DiPietro          NYI    26.35    55.10
Mason             NSH    25.31    43.97
Backstrom         MIN    24.76    40.18
Lundqvist         NYR    23.22    52.12
Emery             OTT    22.77    48.13
Giguere           ANA    19.59    41.94
Vokoun            NSH    19.44    38.92
Huet              MON    14.55    33.75
Lehtonen          ATL    14.02    45.15
Miller            BUF    11.00    39.29
Nabokov           SJS    10.45    28.85
Hasek             DET    10.22    29.86
Harding           MIN     9.51    12.12
Kolzig            WSH     9.04    35.60
Turco             DAL     8.38    31.84
Fernandez         MIN     6.84    24.21
Roloson           EDM     6.76    36.30
Dubielewicz       NYI     5.78     8.75
Smith             DAL     3.45    11.11
Anderson          FLO     3.00     4.74
Toskala           SJS     2.74    16.46
Legace            STL     2.57    20.23
Thibault          PIT     2.22    10.80
Garon             LOS     1.47    14.20
Bryzgalov         ANA     1.31    11.33
Fleury            PIT     1.19    30.50
Osgood            DET     1.01     8.45
Caron             [COLOR="Silver"]CHI     1.37     1.74
                  ANA    -0.43    -0.34[/COLOR]
                  TOTAL   0.94     1.40
Halak             MON     0.44     7.48
Popperle          CBJ     0.33     0.54
Gerber            OTT     0.29    12.05
Sabourin          VAN     0.23     3.59
Finley            BOS     0.13     0.62
Cassivi           WSH    -0.50     0.37
Schwarz           STL    -0.63    -0.26
Houle             PHI    -0.72    -0.67
Thomas            BOS    -0.93    28.85
Budaj             COL    -0.98    21.51
LeNeveu           PHO    -1.55     0.58
Valiquette        NYR    -1.74    -1.06
Ramo              TBL    -1.82    -1.48
Sauve             BOS    -1.82    -1.48
Biron             [COLOR="silver"]BUF    -3.52     4.48
                  PHI     1.24     8.88[/COLOR]
                  TOTAL  -2.28    13.35
Wall              ANA    -2.33    -1.11
Clemmensen        NJD    -2.36    -0.20
Norrena           CBJ    -2.49    18.81
Fukufuji          LOS    -2.93    -2.28
Burke             LOS    -2.93     7.38
Lalime            CHI    -2.97     1.78
McLennan          CGY    -3.21     1.35
Leighton          [COLOR="silver"]NSH    -1.05    -0.90
                  PHI    -2.34    -0.81[/COLOR]
                  TOTAL  -3.39    -1.71
Hedberg           ATL    -3.65     3.85
Bacashihua        STL    -4.39     2.36
MacDonald         [COLOR="silver"]DET    -7.03    -3.86
                  BOS     2.48     5.41[/COLOR]
                  TOTAL  -4.54     1.55
Aebischer         MON    -5.03     8.90
Khabibulin        CHI    -5.06    19.96
Belfour           FLO    -5.24    18.01
Leclaire          CBJ    -5.44     3.99
Grahame           CAR    -5.53     5.00
Brust             LOS    -6.81    -3.13
Morrison          PHO    -7.13    -6.20
Markkanen         EDM    -8.75    -1.90
Conklin           [COLOR="silver"]CBJ    -7.12    -3.97
                  BUF    -1.63     0.17[/COLOR]
                  TOTAL  -8.75    -3.80
Dunham            NYI    -8.76    -0.48
Weekes            NYR    -9.41    -4.08
Boucher           [COLOR="silver"]CHI    -8.18    -2.35
                  CBJ    -2.65    -1.65[/COLOR]
                  TOTAL -10.84    -4.00
Aubin             TOR   -10.90    -5.34
Sanford           STL   -12.12    -1.52
Theodore          COL   -12.71     0.34
Auld              FLO   -13.05    -2.12
Ward              CAR   -13.32    11.06
Holmqvist         TBL   -13.76     3.25
Johnson           WSH   -14.46    -1.05
Toivonen          BOS   -15.54    -8.01
Esche             PHI   -16.34    -9.10
Tellqvist         [COLOR="silver"]TOR    -0.20     0.09
                  PHO   -16.26    -4.56[/COLOR]
                  TOTAL -16.46    -4.48
Niittymaki        PHI   -17.90     5.61
Joseph            PHO   -19.05     3.16
Raycroft          TOR   -22.62     6.34
Denis             TBL   -24.15    -8.13
Cloutier          LOS   -27.62   -18.50[/font]

 

[Snap Wilson]

You need to come up with a different name for this metric. "Goal differential" is widely considered to be goals for minus goals against.

 

[Doctor No]

And yes, Cloutier was that bad.

 

[Doctor No]

You need to come up with a different name for this metric. "Goal differential" is widely considered to be goals for minus goals against.

It depends on the context, although I'm open to ideas for names.

Goals Above Average (my preferred choice) doesn't work for the obvious abbreviation reasons.

 

[pitseleh]

That's some really interesting work.

It's not surprising that Luongo and Brodeur were a good 10% above Kiprusoff at number 3.

I'm guessing that Luongo would have been closer to Brodeur if they had a similar number of times SH (as it would affect SV%).

 

[Doctor No]

I'm guessing that Luongo would have been closer to Brodeur if they had a similar number of times SH (as it would affect SV%).

That's definitely true. One thing I'd like to do is risk-adjust all shots for degree of difficulty, but it's hard to get the data. (And even harder - meaning impossible - for past seasons).

 

[kruezer]

That's definitely true. One thing I'd like to do is risk-adjust all shots for degree of difficulty, but it's hard to get the data. (And even harder - meaning impossible - for past seasons).

Do you have data for past seasons? That would be interesting to see as well, if you have the time.

 

[Injektilo]

Do you run all the shots for ES situations only? That'd have the effect of evening it out somewhat.... if you include all situations, then a goalie that averages 10 min/game on the PP and 5 min/game on the PK will have better numbers than on facing the opposite situations, all else being equal.

I'd suggest the name Goaltender Impacted Goal Differential myself, not very interesting, but pretty straight forward.

 

[Doctor No]

Do you have data for past seasons? That would be interesting to see as well, if you have the time.

I've got year-by-year leaders on the thread linked from the original post.

Here are the career top ten at this point (1982-83 to present):

[font=courier new]
Roy            441.35      866.65
Hasek          414.27      707.74
Vanbiesbrouck  198.36      568.53
Brodeur        193.58      526.09
Joseph         176.17      569.64
Luongo         173.53      366.41
Belfour        165.27      536.54
Hrudey         128.55      433.47
Moog           114.65      398.97
Hebert         112.33      328.33
[/font]

To segue into the thread on the other subforum, Hasek leads Roy in Goal Differential per 60 Minutes by a large margin (0.61 to 0.44).

And the bottom ten, if you enjoy looking at that sort of thing:

[font=courier new]

Billington    -153.67      -27.85
Chevrier       -98.30       -5.34
McLean         -87.83      165.40
Sidorkiewicz   -83.35       13.70
Brodeur R      -82.79       35.50
Ranford        -72.61      201.42
Cloutier D     -68.33       57.34
Stefan         -66.93       53.86
Wregget        -66.92      183.21
Eliot          -63.95      -25.91
[/font]

Note that most of these goaltenders are still valuable, because they played at an above-replacement level for a long period of time. I'm still sad to see my favourite goaltender (McLean) on this list.

 

[Doctor No]

Do you run all the shots for ES situations only? That'd have the effect of evening it out somewhat.... if you include all situations, then a goalie that averages 10 min/game on the PP and 5 min/game on the PK will have better numbers than on facing the opposite situations, all else being equal.

That's true, although I don't have that information available (particularly for seasons other than the most recent few).

One thing I could do is regress the most recent seasons' results based on ES/PP/PK time and impute the results for goaltenders past (based on their team's number of PP and PK chances). Maybe I'll try that at some point.

 

[Hockey Outsider]

Nice job like always, Dr. No. It's not surprising to see Brodeur & Luongo at the top of the list. I like the idea of calculating GAR; if nothing else, being an average goalie (or an average player at any position) prevents the team from being forced to play a below-average player.

Coming up with some standardized names for these metrics would be a good idea--the accountant in me will forever confuse GAR with GAAR (general anti-avoidance rules).

I remember when I first posted a list like this (goals above average, for lack of a better term), I was roasted because Vanbiesbrouck ranked very high (like he does on your list). Remember, the list isn't saying that Vanbiesbrouck is the third best goalie of all-time. It simply means that Vanbiesbrocuk was consistently above-average for a long period of time (a perfectly reasonable assessment). It's actually quite similar to showing a list of all-time goal-scorers and seeing Mike Gartner in 6th; a combination of moderate skill, consistency, and health will allow any player to rank high on a list of cumulative or career stats.

 

[Doctor No]

I remember when I first posted a list like this (goals above average, for lack of a better term), I was roasted because Vanbiesbrouck ranked very high (like he does on your list). Remember, the list isn't saying that Vanbiesbrouck is the third best goalie of all-time. It simply means that Vanbiesbrocuk was consistently above-average for a long period of time (a perfectly reasonable assessment). It's actually quite similar to showing a list of all-time goal-scorers and seeing Mike Gartner in 6th; a combination of moderate skill, consistency, and health will allow any player to rank high on a list of cumulative or career stats.

Exactly - there's a lot of value in durability and playing at an above-average (or even an average) level for many many games. For just one example, remember how Atlanta's playoff hopes were essentially tanked by Kari Lehtonen's injury in 2005-06.

I'm still not positively happy about the 1.5% adjustment to determine replacement level, although it seems about right when looking at many seasons of data.

And I've finally come around to your opinion that a goaltender's own statistics shouldn't count when determining the "field of play". Thanks!

 

[Doctor No]

And I do still prefer "Goals Above Average" as the name of the metric, but it's got the obvious problem.

Would "GAbA" (or "GAbR" for Goals Above Replacement) be too confusing?

 

[Sensfanman]

Why not Goals Against Above Average? GAAA

 

[arrbez]

Raycroft is even worse than I had originally suspected...

 

[Doctor No]

Raycroft is even worse than I had originally suspected...

Raycroft actually had a very good Calder season (24.88 GD, 48.67 GAR), but he's been essentially replacement-level since. Which is too bad because he seems like a good kid.

 

[arrbez]

Raycroft actually had a very good Calder season (24.88 GD, 48.67 GAR), but he's been essentially replacement-level since. Which is too bad because he seems like a good kid.

I feel bad for the guy for being booed at home...but damn, if only he'd be less crappy...