Tuesday, February 20, 2007

The Value of a Shutout

Goalies love shutouts. They guarantee victory and usually result in first star billing. They mean that the goalie has done his job perfectly through 60 full minutes of play. Goalies who post lots of shutouts attract attention; one reason that Martin Brodeur's current season is being considered very special is that he has posted 11 shutouts, nearly twice as many as anyone else.

But what is the value of a shutout? First of all, shutouts are very team dependent. They become more difficult with each additional shot, so goalies that face more shots will generally record fewer shutouts. Shot difficulty is also a factor - the number of odd-man rush chances, power plays against, and close-in shots have a strong impact on the likelihood of a shutout. These problems are of course common to all goalie stats, so I will just point them out as a caveat.

The main issue with shutouts is that very few games end up 1-0, so most of the time the goalie could have given up another goal without changing the game in any way. For example, Ryan Miller of Buffalo has been infamous this season for conceding late, meaningless goals with his team well in front, costing him the shutout but not impacting the result in any way.

How much does a goalie lose by giving up 1 goal? I took Brodeur, Luongo and Kiprusoff, considered by many to be the best goalies in the game today, and broke down their performances based on the number of goals allowed in each game. The results indicated that the difference between a shutout and a one-goal game is slight. Brodeur won 1-0 twice, Luongo and Kiprusoff once each. That means in 16 of the combined 20 shutouts they recorded, they could have given up another goal in the last minute of play (or "pulled a Miller") without changing the game's result.

Brodeur's winning percentage when giving up 1 goal was 91%, Luongo's was 90%, and Kiprusoff's was 95%. These are all close to 100%, and indicate how unfair it is that shutouts are counted and celebrated, while solid, one-goal-against performances are overlooked and forgotten. The winning percentages were similarly high with 2 goals against, Brodeur at 81%, Luongo at 82%, and Kiprusoff at 81%. The break point appears to be with the third goal, as all the goalies had a losing record when giving up 3 goals or more. Kiprusoff was particularly pronounced: when giving up 3 goals he was 4-12-0 for a .250 winning percentage. Luongo was at .423, Brodeur .375. All of them were very unlikely to win when giving up more than that.

This indicates that rather than crediting goalies only if they get a shutout, a better alternative would be to count how many times they give up 2 goals or less. The results indicate that goalies who give up 2 goals or less will win 85-90% of the time. I will check this stat for the entire NHL, but the assumption seems valid, especially since all three of the goalies sampled play for relatively low-scoring teams.

Therefore, rather than awarding 1 point for a shutout and 0 for everything else, I propose creating a stat called a "quality game" (inspired by baseball's quality start statistic), which refers to any game where the goalie played the entire game and allowed 2 goals or less. Of the three, Martin Brodeur ranks first in quality games with 35, Luongo second with 32, and Kiprusoff third with 29. All three had about a 90% winning percentage when they played a quality game. I feel this stat gives a better reflection of their relative levels of play than using shutouts alone (Brodeur 11, Kiprusoff 6, Luongo 3).

I plan on doing more research with the quality games stat, to find out which goalies consistently give up 1 or 2 goals per game, which goalies are often bailed out by their teammates, which goalies are having bad luck in close games, and which goalies have a few abnormally good or bad performances that are skewing their overall records.

No comments: