Red Wings Goaltending Analysis: Team Possession and Jimmy Howard

The last couple weeks, we’ve been going over Jimmy Howard’s play both last season and for his career as a starter to try and see where he stands among his peers in some of the non-obvious areas. Here’s what we’ve learned so far:

• Over the last three years, Jimmy Howard is 10th among goaltenders for highest percent of total starts which meet Hockey Abstract’s criteria for a Quality Start. This is extra-impressive considering how much a subpar performance from last year slowed his numbers down.
• Jimmy Howard tends to bounce around a little bit more from start-to-start than his peers.
• To go along with this bouncing, he tends to slightly over-regress after a very good or bad start. He’ll only follow up a Quality Start with another one a little bit more than half the time, whereas he’s much less likely to drop two stinkers in a row.
• Thanks to this great fanpost from Mike Morris, we’ve learned that last season, Howard could not keep pace with a conceptual machine meant to consistently perform as a top-ten goalie. In the comments, we learned that he still compares favorably over the last five seasons to players like Henrik Lundqvist, Jonathan Quick, and Marc-Andre Fleury.

Additionally, we’ve seen over at D2D43 some good analysis on the types of goals that Howard gave up in relation to his performances during and directly after a Quality Start or a Really Bad Start. By separating the goals allowed into categories, we can see hints that both the defense and Howard play differently after a certain game type (although the author is very careful not to jump to conclusions based on what are pretty small sample sizes).

Today, we’re going to look a bit more at the way the team plays in front of Howard to see if we can build evidence for the skaters’ blame/credit in losses/wins.

Team Possession Metrics

To try and build a separation between Howard’s play and the way the skaters in front of him played, we can look at possession data (shot attempt differential). This won’t create a perfect separation because the mere presence of a goalie changes things, but it’s a good first step at least. One of the issues with possession data is that one goalie doing his job better than another can skew those numbers. For this, I’ve isolated the data that shows up only with the score close (a goal differential not greater than one during the first 40 minutes and a tie game any time after that). To eliminate problems with manpower advantage differentials, I’ve only taken data here with the teams at even strength.

The Good

Using this method eliminates the very real phenomenon of score effects, where both teams see a change in their expected shot attempts created by one team taking more chances trying to catch up while the other plays a bit more conservatively to hold the lead. The Score Close criteria is designed to keep enough data in the sample sizes to be useful while reflecting only information created while both teams are ostensibly trying equally hard to score and prevent scoring.

The Bad

I don’t have a method of separating out or even accounting for certain things which can affect this. For instance, a goalie kicking out a high number of bad rebounds will cause more shot attempts against himself. This would be attributed to the team (although you can argue that even those are partially the defense’s fault). Second, this still treats every shot attempt equally, regardless of its distance from net. This is supposed to even itself out over time, but I can’t confidently tell you either how much it should change things or how much it actually does change things for Howard.

Results

Counting all of these provides us with almost 15,000 combined shot attempts. For this, I went with rate stats per 60 minutes. The smallest sample among these categories comes from the Really Bad Starts, where Howard averages fewer than 30 minutes played (can you guess why?). For this sample, that means 1,130 minutes played with the score close, which is about one-third of a season for a goaltender just outside the top ten in minutes played.

If we’re looking for the defense making Howard’s workload harder or easier depending on game type, we’re looking for differences in the way the Wings control play during different start types.

So the range for shot attempts created is less than 1.7 attempts per 60 minutes wide (reminder: that’s all attempts, not just shots on goal. Misses and blocks are counted here too). The Corsi against range is even smaller at 1.4. The CF% range is also about 1.7%. These aren’t very large effects.

It’s interesting to note that the biggest positive differential for the Wings in possession metrics happen when Howard is having his very worst nights. I think this might actually be a bit of the cart leading the horse for this analysis though. Howard has 24 of his 38 Really Bad Starts over these five seasons where he played 55 or more minutes. 11 of those were games where there were 40 or more minutes of score close time. While we are adjusting for score effects, it’s possible that in games like these, there’s a cumulative “always be catching up” effect where a score close situation isn’t actually being played like one.

One thing we do get a minor hint of here is the idea of “buckling down vs. getting cocky.” I hinted at this in the post about following up bad starts and D2D43 explored it more extensively in his look at the 2013-14 goal against types, but if we’re looking for evidence that the team changes play after a winner versus a loser, then it is notable that the games 1 after a QS and a RBS are on the edges of overall differential.

If we’re looking to assign a narrative to that, it would look like the team plays more carefully after a stinker and keeps it opened up after a confidence-boosting outing. The difference here is that the assumptively cocky skaters give up fewer shot attempts after a good start by Howard and the embarrassed five follow up a stinker by taking it out on the opposing goalie at a lower rate.

However, what’s most important in this analysis is that over decently-sized samples, the difference in the way the team in front of Howard plays varies very little depending on his performances. The differences in Howard’s play after one start, where he seems to over-regress to the opposite extreme looks like it falls more on him than on his team, but not to a degree where we could confidently call this confirmed.

What about Special Teams?

One of the issues regarding Corsi in this analysis is that throwing out penalty differential and shot attempts on special teams can downplay team effects. After all, if we’re looking at how much harder the skaters make it for Howard to win, then factoring back in the 20% of the game played on special teams can shed some light for us.

The Good

Jimmy Howard has 18 total regular season penalty minutes in the last five seasons. While looking for things which make his job easier/harder, penalties do a very good job of this with very little direct interference by Howard himself (although the concept still exists that things like bad rebounds and poor puck handling by Howard can lead to penalties taken by his team).

The Bad

There’s a lot more to penalties than it seems and even using score close numbers can be misleading. You’d think that a team dominating possession more would earn more power play opportunities. By-and-large this is true, but there’s a diminishing effect there because refs tend to lean far more-heavily towards evening out penalty differentials in games than they should and they also tend to favor teams which are losing by one goal more.

The Results

Overall, the Red Wings have a positive score close penalty differential in every kind of game for Howard. There’s a bit more variance in how many penalties and a bit more variance in attempts/attempt differential, but I’m not as confident in drawing conclusions based on this because of the variables in what causes penalty differentials.

Turns out taking a guy off the ice at semi-regular intervals gives the wings about 13 more shot attempts on net per 60 minutes of play while their opponents’ numbers go up by roughly 11. Despite the differential range increasing all the way up to 2(!), the attempt percentages are all eerily similar across the board. When you factor in penalties, it looks like the Wings (and refs) are helping Howard out more by volume, but not by percent of the total.

Conclusion

I don’t know how different I expected the games during different kinds of Jimmy Howard performances to be when I started this, probably more than they turned out to be. If we’re looking to kill the narrative that Jimmy Howard relaxes after a good game and re-focuses after a bad one, then we can’t concretely do that. If anything, the Red Wings seem to make it slightly harder on him in response to a bad game and he still performs much better than his average. However, we haven’t really cemented it either. We know that goalies are variable and we already knew that Howard was more-variable than many. We also know that he’s better than most.

Fortunately, it seems like there are a couple new stats sites out there which can help us look at shot distance. We’ll take a look at that next.

(A big thanks to Nicetimeonice.com for creating a great interface for quickly looking up the data I compiled for this article.)