Cory and Marty's Records: The Answer

The Question

For those who haven't already done so, I suggest reading the post I made two weeks ago of which this is a sequel of sorts. In that article I posed the question of the Devils scored so much more for Marty than Schneider and why, consequently, they had a better record for Marty despite poorer performance in both Save Percentage and Goals Against Average. I took some early stabs at an explanation before asking for advice from our readers on where to go next. In combining the feedback from the poll and the comments, I've decided to look at 3 things:

1. Seasonal Scoring Trends (who was starting when we were hot and who was starting when we were not)
2. Shooting Analysis (Results of Shots in certain situations) 
3. With or Without You Analysis (Which goalie played with which skaters most)

Seasonal Scoring Trends

EXCEL FILE: Devils Goalie Log

Using consolidated data from Hockey-Reference, NHL.com, and Extra Skater, I created a master sheet for the game log of the team and then separated it into game logs for the two separate goalies. The first thing I did was look at the amount of consecutive starts. The two spikes were both from Schneider, the first being a string of 6 starts from Jan. 28 to Feb. 8 and the second a 5-game streak from the 1st to the 10th of April. In those 11 games, the Devils scored 16 goals for an average of 1.45 goals per game. If these runs were stripped, Schneiders GAA would have gone from 1.95 to 2.13. Still a far cry from Brodeur's 2.81 but it's progress. This identifies an element of randomness that contributed to Schneiders low goal support. Namely, his highest density of starts came in games that the Devils just weren't scoring goals even by Schneider's standards.

The second thing I looked for was when the Devils were on a hot scoring streak. I took the 6-game running total of goals scored and looked for peaks. I found two of note. From December 7th to 21st the Devils played 7 games and scored 25 goals (over 3.5 goals per game). Brodeur had 5 of those starts. Taking away those starts brings Marty's goal support average down to 2.59 and barely changes Schneider's. The other streak was the 6 games after the olympic break where we put up 26 goals. Those starts were evenly split. If you take away those starts Schneider drops back down to 1.85 (1.96 if you pick up where we left off last paragraph), but Marty's plummets to a 2.35.

This was not a statistical analysis. This was an experiment where I took away some of the extraordinary portions of the season. If we take out the games I mentioned in this paragraph, we get a sample of 28 Schneider games and 32 Brodeur games in which their goal support was 1.96 and 2.35 respectively.

Shooting Analysis

EXCEL FILE: Devils Goalie Shots For Analysis

Using the data through March 28th offered at Hockey Prospectus, I made a sample of situations the Devils were shooting in and the success rates. Now when you make a big chart like this it is not always obvious what you're looking for until you find it. So what I did was convert the raw data into percentages of the total shots taken with that goalie on the ice and then compared the percentages and searched for the highest differences. I found 3 numbers of interest, but one was merely higher shooting percentage for Brodeur which we already knew about so I will only mention the other two.

The first was that the biggest difference in situational shooting was the powerplay. 12.9% of Marty's shots came on the 5v4 powerplay and only 12.2 of Schneider's did and that represented the biggest difference in the investigation. A brief further analysis reveals the Devils scored 22 powerplay goals with Marty and 19 with Schneider (remember it's only through March 28th). Not a huge difference, but considering Schneider had played more games even at that point it's interesting. The were on the ice for the same amount of PP shots but Schneider was on for 100 more even strength shots.

The other difference is the missed shot percentage. The biggest difference on the whole chart was that the Devils missed 2.74% more of their shots when Schneider was on the ice. Furthermore, the Devils in 33 attempts did not score an empty net goal for Schneider, but got 4 in 27 attempts for Brodeur. Now the numbers here say the Devils were credited with 0 missed shots when Brodeur was off for the extra attacker, but 11 when Schneider was off. If the readers would like to enlighten me as to what a SOG vs missed shot looks like in an empty net situation I would like to know since these number seem suspicious. Regardless, the Devils missed far less shots, especially with the goalie pulled in Brodeur games.

With or without you analysis

EXCEL FILE: Devils Goalie WOWYs

This analysis was done to reveal if there were any players who played appreciably more or better with certain goalies. It used data from stats.hockeyanalysis.com. I found a lot of data that was almost interesting, but in the interest of keeping an already long article as brief as possible I will only mention the highlight of my investigation.

Some players did experience higher Corsi or higher GF/20 with one goalie than the other, but what was by far the most interesting to me was the TOI differences for defenseman in particular. Below is an excerpt from my excel sheet.

The final column is the one of interest as it shows the difference in percentage of total goalie ice time spent with the defender. It is ranks in descending order of highest percentage of time spent with Brodeur. I draw your attention to the two top and two bottom defender names. The 5 middle defenders were all pretty close in percentage (within 2%) but the biggest difference in percentage of ice time spent with one goalie were Larsson and Gelinas who spent a total of 113 more minutes with Brodeur despite Schneider logging over 300 more minutes on the season. Scheider, in return, spent more of his time with Salvador and Fayne.

I should say that in my opinion, the only one that matters here is Gelinas as his stats pop off the charts. Though we like to think of Fayne as a boring defender, he was actually 2nd only to Gelinas in on-ice scoring (measured in goals for per 20 minutes) and 3rd in Corsi For per 20. Furthermore, Larsson was actually lowest in on-ice scoring. In the interest of full disclosure, this chart may not be as enlightening as it appears off the bat because of these two facts. However, the difference between Gelinas and Salvador cannot be understated and I do feel as though that at least partially contributed to the difference in offense -- and possibly some of the difference in GAA and SV% as well.

Randomness?

EXCEL FILE: Devils Gsupp Randomness

A lot of what I've been looking for has not been a denial of the random nature of this past year, but rather a further explanation for what caused this variation. We like to toss around the word "random" like it means that these stats are a flip of a coin, but hockey games are far more complex than that. What I tried to do was isolate which individual highly-randomized events were particularly instrumental in creating this goal-scoring disparity. If we ignored that and wrote it off as random like some of the commenters suggested, I was curious where we would fall on a league-wide scale. I used Robert Vollman's statistics to identify the variation or goalies goal support within a team. In the past 2 years, there have been 3 teams with higher deviations. The 2013 Capitals refused to score for Neuvirth, the 2014 Flyers seemed to prefer Mason, and the 2013 Leafs couldn't light the lamp when Scrivens was in net. In other words, this was not unprecedented. Oddly enough it does seem to be the first time though that the better goalie has gotten less goal support.

Conclusion

It was random. However, like i said, the randomness can be attributed to certain things. In this article I've shown that Schneider played his two longest streaks when the Devils were cold. Furthermore, Brodeur reaped the benefits of playing through our 2 most prolific scoring runs. Also the Devils seem to shot more on the powerplay for Brodeur, miss less shots, and capitalize on empty netters. And lastly, the young defenders got more time with Brodeur while Schneider got the defensive-minded vets.

What are your thoughts? Does any of this information make you want to see different action next year with regards to who plays, what we do, or how much to rest Schneider? Do you agree it is largely random what happened this past year? Do you agree that the things I mentioned were at least partially the reason? Please leave comments voicing your opinions.