Thursday, May 29, 2008

Shot Prevention

We have had a very interesting discussion lately in the comments about whether goalies can reduce shots against. It has been demonstrated, and I agree with the math presented, that if goalies are able to reduce shots against by 2 per game that would be roughly the equivalent of an extra .005 in save percentage. However, it still has not been demonstrated that an effect that large is actually observed among NHL goalies.

I'm going to focus on Martin Brodeur, again, but as he is one of the gold standards of puckhandling as well as rebound control it makes sense for us to do so. I am going to look at the shots against statistics of his backup goalies to see what kind of information that gives us about New Jersey shot prevention.

Here are all of Martin Brodeur's backups that played at least 150 minutes for New Jersey and at least 150 minutes somewhere else in the league between 1993 and 2008, listed with their shots against averages in New Jersey and for the other teams they have played for (1993-94 to date, to coincide with Martin Brodeur's career):

Scott Clemmensen: 26.6 in NJ, 24.2 elsewhere
Corey Schwab: NJ: 21.7 in NJ, 29.0 elsewhere
Mike Dunham: 26.5 in NJ, 29.7 elsewhere
Chris Terreri: 27.0 in NJ, 28.8 elsewhere
John Vanbiesbrouck: 23.3 in NJ, 27.6 elsewhere
Kevin Weekes: 28.0 in NJ, 29.5 elsewhere

Most of these are very small sample sizes, so we need to look at the aggregates. Overall, Brodeur's backups played 10,748 minutes and faced an average of 26.1 shots per game, 0.7 more than Martin Brodeur himself. This average is skewed by Chris Terreri's season in 1993-94 when he split time with Martin Brodeur. Since Brodeur would go on to play so many games per year from 1995 to present, Terreri's minutes from 1993-94 make up nearly a quarter of the total backup sample, and the league was still relatively high-scoring (and high-shooting) that season compared to later in the decade. If you take a year-by-year average of shots against by backups, the result is 25.4 shots per game, exactly the same as Brodeur's career mark. For the group, the average shots per game outside of New Jersey was 28.9, nearly a 3 shot per game difference.

Let's look at some of the other stats. Every single backup goalie listed above had a better winning percentage for the Devils than for the other teams they played for. Except for Kevin Weekes, all of them also had a better save percentage and GAA in New Jersey than elsewhere.

NJ backups: 73-64-19, 2.52, .904, 26.1
NJ backups elsewhere: 461-600-170, 2.77, .904, 28.9

The save percentage numbers are interesting - exactly the same in New Jersey as everywhere else. Is there an explanation for this? Does this invalidate the shot quality measurements that claim New Jersey had the league's best defence? No, it does not, because this is a simple sum total of all the backups' stats, and again the stats are skewed. Chris Terreri makes up about half of the New Jersey minutes but less than 10% of the non-New Jersey minutes, while John Vanbiesbrouck only had 540 minutes in New Jersey compared to over 27,000 minutes on other teams.

To deal with this issue, I weighted the goalie stats based on New Jersey minutes played, and here are the results:

.419 win %, 3.08 GAA, .891 save %, 28.4 shots/gm

These numbers are well below average. It could be that Brodeur's backups played on worse teams than normal - that is somewhat indicated by the low winning percentage. Even so, it is probable that the weighted results are closer to reality than the simple sum totals, for the reasons described above. The shot effect isn't as large, but the shot quality effect certainly shows up.

So, looking at these backup stats, can we make any conclusions about Brodeur? There could be other factors at play here, but the data appear to show a clear New Jersey shot prevention effect, with the goalies facing 2-3 fewer shots per game playing in New Jersey compared to everywhere else. They also stopped a higher percentage of shots, and won games at a much higher rate. The difference in shots against per game between Brodeur and his teammates was less than 1 shot per game in total, and almost identical when comparing the seasonal averages. That doesn't appear to provide any evidence for the theory that Brodeur can decrease shots against, but we don't know the typical difference in shots between a starter and his backups. The numbers given could be still be evidence of a slight effect if backups generally give up fewer shots than the starter, which may be a reasonable assumption based on a weaker strength of opposition and the team playing tighter defensively with a weaker goalie in the net. So for now, the results are still somewhat inconclusive, and further study is required to help identify and measure any goaltender shot prevention effects.


Anonymous said...

Where's Bruce? haha

Bruce said...

I'm right here. I read this post with interest, but I couldn't think of anything to add to it. When you start fooling around with weighted averages on guys like John Vanbiesbrouck, who played the last 9 of his 882 games crazyglued to the Devils bench in the manner of all Brodeur's other backups, or Scott Clemmensen, who played 3 games with the horrible Leafs and posted an .839 Sv% which looks bad compared to those he compiled over four seasons with the Devils, I'm not sure it tells us very much, but I commend the attempt.

I agree with CG's conclusion:

the results are still somewhat inconclusive, and further study is required to help identify and measure any goaltender shot prevention effects.