The Devil is in the Details

April 03, 2014


The New Jersey Devils will play in back-to-back games this weekend and it would be a reasonable guess to assume that the starts will be split between Martin Brodeur and Cory Schneider. What we know for sure is that Cory Schneider is starting in Friday's game. To date, the goaltenders have split the starts 51% to 49% with a slight edge to Schneider. This post examines that split, the thinking behind the split, and the consequences of that thinking (both for the Devils this season and the futures of Martin Brodeur and Cory Schneider).

Stopping the Puck

With league average goaltending coming in at about .913, the days of Martin Brodeur as a starting goalie have been over for quite some time, but nobody has told the Devils. Below is his recent four-season history as an NHL goalie:

SeasonGames PlayedSave Percentage

We can construct a similar table for Cory Schneider:

SeasonGames PlayedSave Percentage

It's pretty clear that, at this stage in their careers, Schneider is the superior goalie. If you're a coach asking yourself which goalie gives you the better chance at stopping the puck tonight, the answer is obvious. If goalie selection were made based on save percentage, no one would think you were crazy to give Schneider at least 60 starts. By extension, people would think you were crazy to split the starts 51% - 49%.

So just why is Coach Pete Deboer ignoring the data?

Winning Games

To date, the Devils have wins in 42% of their games this season. And if you break that data down on an individual basis, you find something surprising: the Devils' win percentage with Brodeur is 48% and only 35% with Schneider. That's right, the goalie with the .900 SV% is winning games with greater frequency than the one with a SV% of .916. Early in the season (through December 1), the Devils' win percentage with Schneider in net was an alarming 25%. At the time, Brodeur was receiving 56% of the starts.

Considering the end result (did your team win or lose), you can understand the pressure a coach might feel to start Martin Brodeur more often than his save percentage would ordinarily warrant.1 But why were the Devils winning more often with their weaker goalie?

Shooting the Puck

You might be inclined to think that teams behave differently when they have different goalies in net. A rough proxy for team behavior would be their ability to generate shots. For games in which Brodeur was the starting goalie, the Devils have produced 26.5 shots on goal per game. With Schneider starting, that number comes in at 26.3 shots on goal per game. From a shooting the puck perspective,2 there is no difference in team behavior that would explain why the Devils win more frequently with Brodeur in net.

Scoring Goals

The Devils sit in the 27th spot for goal production in the NHL this season with 2.4 goals per game. If we split the scoring rates by starting goalie, we find that the Devils produce 2.9 goals per game with Brodeur in net (good for 8th best in the league) and only 2.0 goals per game when Schneider starts (good for 29th in the league). For the season, the team is shooting with 9.2% success. If we split this number between the goalies, we find that the Devils shoot with 11% success when Brodeur is in net and 7.6% with Schneider. For reference, with a league-wide save percentage of .913, about 8.7% of shots end up as goals. Now we've found the key reason for Brodeur's winning success this season: his team has found a way to score on 11% of their shots when he is in net and this scoring rate is masking his poor performance in net.

Earlier, I noted that there was no measurable difference in team behavior with the two goalies on the ice. And here I'm telling you that the Devils score at a significantly higher rate with Brodeur in net. The two statements are not contradictions; the Devils aren't behaving differently with Brodeur on the ice, they are just getting lucky.

The Impact of Goal Support on the Devils' Record

In an earlier post (Dec 2, 2013), we outlined a plan for the Devils to make the playoffs that required Cory Schneider to start the bulk of the Devils' games. While the Devils did start to lean in that direction (Schneider's share of the starts has jumped from 44% to 51%), it appears the changes are going to come up short. The Devils currently sit three points out of a wildcard position and their rivals hold a games played advantage.

What About Shootouts?

It would be unfair to ignore the impact of the shootout on the Devils' performance this season. The Devils have appeared (some might argue with the use of that term) in 11 shootouts this season. The problem with shootout results is that study after study has shown them to be a product of luck (that is, being 'good' or 'bad' at the shootout is not a repeatable skill). Thus, when you see that the Devils have gone 0-11 in shootouts this season, you realize that bad luck has cost them a few points in the standings in this season (Schneider has been in net for seven of his team's shootout appearances).

Putting it All Together

The 2013-2014 New Jersey Devils are on pace to score just shy of 200 goals this season. If they were to yield 200 goals against this season, they'd be a playoff bubble team. To secure a playoff spot, they would probably need to reduce that number of goals against to 195. And if you include some bad luck from the shootout, your team's finish is even more susceptible to the total number of goals against. You can see how a team's goals for and goals against impact their position in the standings below:

As it stands, the Devils are on pace to give up at least 210 goals. So, how does a team that faces 26.3 shots per game keep their season-long goals against number at about 195? They start goalies who can provide them with a save percentage of at least .909 (calculation example).

And now we've come full circle. Martin Brodeur provides below average goaltending for the New Jersey Devils, but has been the starting goalie for half of their games. He's been the starting goalie for 19 of the team's losses. If we use the .909 SV% as our guide, 74% of the losses incurred with Brodeur as the starting goalie can be attributed to his poor play (for comparison, Schneider comes in at 52%).

When the Devils fall short of a playoff berth this Spring, you can dig around for slumping forwards and shootout results to try to account for the failure. But there's a much simpler reason staring us all right in the face: Pete Deboer elected to give Martin Brodeur far too many starts. How many is too many? Well, we don't know Deboer's plan for the final six games, but our estimates suggest that Brodeur will have been started at least 15-20 games too many.

Looking Ahead

Despite what the data tells us should happen, Martin Brodeur will almost certainly have a job in the NHL next season (if he wants one). And this will almost certainly be a mistake by whichever GM pulls the trigger. Five teams let goalies with a SV% at or below .900 start 25+ games this season; none of these five teams will make the playoffs.

As for Cory Schneider, some might view this season as a failure for him given his 14 wins and failure to make the playoffs. These people would be wrong. With even modest goal support next season, Schneider will find himself in the top quarter of all NHL starting goalies again. It's safe to say that when your worst season in the NHL is better than the career numbers of guys like Ryan Miller and Marc-Andre Fleury, you're probably doing something right.

1. This is known as 'riding the hot hand' and has been dismissed as an effective goalie selection strategy. Source
2. Effects on shot suppression have also been found to be minimal. Source


Outstanding article. Really enjoyed reading this, puts a couple of myths to rest and shows just how NJ management handled their team this season.

Great job as usual, Mike!


You're counting the 1 GA from a shootout loss in your total goals against, NJ has given up 193 goals in 81 games so far. It's pretty dumb for the NHL to include shootout winning goals as goals for/goals against in their standings.