In November, I wrote an article that took a closer look at Jimmy Howard’s performance to that point of the season.
For this article, I wanted to follow up on some of the things I wrote about in that article as well as exploring more about what High Danger Chances Against can tell us about Detroit’s performance this year.
While this article will involve advanced stats, I want to make it relatively easy to follow for those who don’t know too much about that area of analysis. Different sites have different ways of evaluating scoring chances, although they all have the same basic idea.
For this article, I am using stats from Natural Stat Trick. Their scoring chance statistics are based on the criteria used by the now-defunct War on Ice. I will be focusing on high danger scoring chances for this article
(From Natural Stat Trick) Attempts from the yellow areas are assigned a value of 1, attempts from the red areas are assigned a value of 2, and attempts in the green area are assigned a value of 3.
Add 1 to this value if the attempt is considered a rush shot or a rebound. A rebound is any attempt made within 3 seconds of another blocked, missed or saved attempt without a stoppage in play in between. A rush shot is any attempt within 4 seconds of any event in the neutral or defensive zone without a stoppage in play in between (originally defined by David Johnson on the now-offline Hockey Analysis, and modified to 4 seconds by War-on-Ice).
Decrease this value by 1 if it was a blocked shot.
A High Danger Scoring Chance is defined as a chance that has a value of 3 or higher.
During last season especially, many games I watched seemed to be filled with these types of chances happening against Detroit, so I became interested in evaluating if we had more than average and what that could tell us.
Why High Danger Chances?
When doing research for the original article, I learned how important High Danger Save Percentage (HDSV%) is considered to be for goalies.
In an article for Sportsnet, Steven Burtch wrote:
The difference between the average elite starter and the average replacement-level goaltender is almost entirely based on HD SV%. There is literally no distinction in terms of average ability for each group on other shots.
Additionally, Emmanuel Perry, the founder of Corsica.hockey, wrote:
...it appears the skill-driven component of Sv% is almost entirely contained in a goalie’s ability to stop shots of the High-Danger variety.
If that’s the case, how much can we tell about a team’s performance by just looking at high danger statistics? That’s what I was interested in.
So Let’s Take a Look
In my original article, I tracked a rolling five game average of Detroit’s High Danger Chances Against per 60 Minutes (HDCA60). All stats for this article are 5v5 (and score-adjusted if applicable). I went back and finished the chart for the entire season:
The league average was 10.73. Detroit’s average for the year was 10.57. I was very surprised that Detroit was below the NHL average, but that’s why basing analysis on only data or only the eye test doesn’t give you a full picture. If you’re not familiar with a rolling average, the chart above shows the average of the previous 5 games for each data point.
The worst five game stretch was games 11-15. I looked for patterns that might explain why certain stretches were particularly bad. The most obvious reason would be the opponent we played. During this stretch, Detroit gave up 18.8 HDCA60 against Tampa Bay, which would fit with this hypothesis, but it doesn’t seem that simple, since the next three games were against Florida, Arizona, and Ottawa, in which Detroit gave up 14.4, 12.6, and 14.8 HDCA60 respectively. (The first game of the five game stretch was against Buffalo, which had 8 HDCA60, otherwise the five game average would have been even higher.) The team went 2-3 during this stretch.
The other stretch in which the five game average was the highest was games 61-65. This stretch, however, is marked by an even larger disparity in chances from game to game. Detroit gave up 23.5 HDCA60 to the Rangers, 13.7 to the Wild, but also 4.03 to Carolina, 6.3 to St. Louis, and 6.85 to Winnipeg. Detroit had 2 wins and 3 losses during this period.
For the sake of length, I’ll just say that the low points (meaning good) on the graph look to be the result of one or two particularly good games in preventing HDCA as opposed to an actual stretch of low HDCA.
How Much Impact Does HDCA Have?
Like I said earlier in the article, I was interested in looking at how much HDCA can tell us about a team’s performance.
If HDSV% is most of what separates goaltenders from each other, then it would make sense that HDCA and HDSV% would tell us a great deal about what to expect from a team in terms of goals against.
I started off by looking at the correlation between High Danger Goals Against per 60 Minutes and Goals Against per 60 Minutes for NHL teams from the 2017-18 season. As you would imagine, there was a strong correlation.
If we look at the correlation between HDCA60 and GA60, the correlation is decent, but not as strong. You can see there are more outliers.
I thought this should at least be partially explained by a team’s HDSV%. I looked at the four largest outliers on both sides of the trend line, and that seems to be the case.
NHL Average was 87.44, and the colored boxes indicate a HDSV% at least one standard deviation above or below the average.
What About on the Individual Level?
Lastly, I wanted to take a look at the rate of HDCA when players are on the ice, both skaters and goalies.
I used HDCA60 since HDCA is cumulative, and the results would be skewed by ice time. For this chart, I added a gradient using Time On Ice / Game Played. The darker the color, the more that player played per game.
Let’s look at the defense first.
A lot of this chart is not surprising, but there are a few things that stick out.
First of all, Danny DeKeyser is at least a full chance lower per 60 minutes than Jonathan Ericsson, Trevor Daley, and Mike Green. The average for NHL defenseman is 10.83, so it might surprise people to see him almost a chance per 60 below the average, but here he is.
The player that sticks out negatively is Xavier Ouellet, which unfortunately is not surprising. While I am a Joe Hicketts supporter, the sample size is too small to get excited. We’ll have to see how he does with a more regular role, hopefully next season.
Daley and Ericsson played most of their minutes together, so it’s not surprising to see them close to each other here. Mike Green, however, played most of his minutes with either Niklas Kronwall or Danny DeKeyser, who are both substantially lower on the chart than he, although Green did play the most 5v5 minutes per game on the team.
Let’s move onto the forwards:
On the forward side, the NHL average was 10.64. Dominic Turgeon is included for completeness, but his sample size is so small it doesn’t tell us anything.
Outliers on the negative side are Dylan Larkin, Martin Frk, and Luke Witkowski. Larkin comes as a bit of a surprise if you just base your evaluation of him on the eye test. His Expected Goals Against / 60 (xGA60), which takes into account shot quality, was the worst for forwards on the team last season. He also had the highest xGF60, so he partially makes up for this deficiency with high danger chances for Detroit when he was on the ice. (His xGF% was 47.85, so the net effect was still negative.)
Yes, Larkin played against very good competition, but so did Henrik Zetterberg, who played more 5v5 minutes per game and averaged 3.5 HDCA less per 60 minutes than Larkin did.
Larkin took a major step forward this season, and I am expecting him to continue to improve. This is one area in which he can do so.
On the positive side, the major standouts are Gustav Nyquist, Henrik Zetterberg, Frans Nielsen, Anthony Mantha, and Tyler Bertuzzi.
Zetterberg and Nielsen being where they are is not surprising, but the other three may be a little more so.
For someone who can be criticized for not always providing consistent offense, Nyquist can be seen here as adding value by limiting offense against Detroit.
What about the goalies?
Howard played the lion’s share of minutes for Detroit this year, and he kept the team in a large amount of games. When you look at the data, however, something interesting emerges. Howard faced considerably fewer high danger chances than Mrazek or Coreau, but allowed more goals on those chances than Mrazek. Coreau didn’t play many games this season, but even so there is a stark difference in HDGA60 for him compared to the other two.
Basically, Howard faced fewer HDCA60 than NHL average and more of those shots were goals than NHL average. This is something I was completely not expecting, and is completely different from what I found in November.
I want to make sure that I’m not underselling this. Considering how important high danger chances are, Howard giving up more than league average while facing fewer than league average HDCA60 could have had a substantial impact on Detroit’s performance in the standings this year.
Since it’s been noted for a while now, I wanted to include goal support for each goalie. Howard continued to get much less goal support at 5v5 than did Mrazek.
- Limiting High Danger Chances Against is important.
- While during the games it often seemed otherwise, Detroit allowed fewer of these types of chances per 60 minutes than the league average over the course of the season. There is a reason that simply relying on the eye test can lead to incorrect conclusions (as can simply relying on data).
- While Dylan Larkin had a very good season, limiting these types of chances against will allow him to be even more valuable to the team.
- While it’s not surprising that Mike Green did not come out looking well when looking at a defensive metric, his most frequent partners, Danny DeKeyser, and Niklas Kronwall, were noticeably better, even accounting for the disparity in ice time.
- Even though it often seemed otherwise, Jimmy Howard did not perform as well against high danger chance as Petr Mrazek did for Detroit in this area this season as a whole (although he certainly did in some individual games.)
If you would like to look at interactive versions of the player charts I made, you can find them here