(3/11/15: I have to issue a pretty massive correction to this article, as I was using the incorrect definition of what the NHL considers at Takeaway. It does change the outcome of many of my charts/graphs, however the conclusion as a whole remains unchanged – the Stars offense is not being as defensively responsible as needed. All data is from the same time period, so will not include any updates from recent games.)
After this weekend’s disastrous games, I think it’s time to dive back into defensive metrics and why we seem to be failing miserably at holding onto a lead.
One of the terms commonly thrown around when discussing defense is “Shot Suppression”. The stat that best represents this is Fenwick Against. Fenwick is just Missed Shot & Shots on Goal (it leaves out blocked shots, which are included in Corsi). Why leave out blocked shots? Well, over large sample sizes, Fenwick has a slightly better correlation to possession. Also, there is some disagreement as to whether shot blocking is luck or skill. Since we’re looking at the full season of data, Fenwick is an appropriate measure here.
All numbers from War-on-ice, 5v5, and are score adjusted (because lord knows, we’ve played from behind a lot lately), unless otherwise noted.
Now, everyone knows we have a defensive problem, but it’s one thing to know there’s a chink in the armor, and another to see the giant gaping hole. Dallas is currently 22nd in the league in Fenwick Against Per 60 (FA/60). Last season, they were 15th.
This chart shows the 10 game rolling averages of Dallas, Detroit (who is #1 in FA/60) and Boston (who I chose because they split the difference at #11). That nice big dip is December. Coincidentally, the Stars earned 1.3 points per game that month. In February they’re only racking up 0.91 points per game.
No wonder the playoffs seem to be slipping further away.
But where exactly is this going wrong? Our D-Core seems to be taking most of the blame, so let’s look at them first.
There’s a lot going on in this chart so here’s a quick glossary:
- CF% = Corsi For %
- SCF% = Scoring Chances For %
- SCA60 = Scoring Chances Against Per 60
- FA60 = Fenwick Against Per 60
- TK/60 = Takeaways Per 60 (the player strips an opponent of the puck, pulled from hockey-reference.com, still at 5v5, but unadjusted)
- GV/60 = Giveaways Per 60 (the player gives away the puck to an opponent, also from hockey-reference.com, still at 5v5 & unadjusted)
- TO Delta = Takeaways Per 60 – Giveaways Per 60 (a measure of how defensively responsible a player is)
- ZSO% Rel = Offensive Zone Start % Relative to the team without the Skater on the ice
- TOIC% = Total Ice Time % of the Skater’s Competition
- CorC% = Corsi For% of the Skater’s Competition
A couple things should jump out at you here, first that the Daley pairing is having a really tough time in FA/60 – they’re the only D-men who are worse than the Stars’ average, and all are worse than the League average, too. (Schlemko is grey because his numbers were not included in these averages because of his very low TOI).
Another very odd thing to me was just how many giveaways Goligoski and Klingberg have. On the 5v5, they log the most ice time of any pairing, so that’s very concerning, especially since they play the most in the defensive zone of any pairing and against the hardest competition (as shown by TOIC% and CorC%).
On the other hand, our defense is doing an excellent job of pressuring the other team for the puck, with a Takeaway/60 average of 1.03, lead by Rookie Sensation John Klingberg. This is indicative of the aggressive, offense-led style the Dallas defensemen are expected to play.
Still, if we’re considering Detroit as the Gold Standard for Defensive Metrics this season, Dallas has a long way to go – Detroit’s average Giveaway/60 is 1.86 to Dallas’s 2.39 (though, as a homer, I have to point out that the Stars’ TO Delta is actually better than Detroit’s at -1.36 vs -1.41, the higher the number the better) .
But the forwards aren’t blameless in this season’s defensive breakdowns either.
I tried to color code this the way I do in Game Recaps. It’s not unusual that you see a slight uptick in SCA/60 and FA/60 with forwards because defense is their secondary responsibility.
What’s mostly concerning on this chart is that there are 5 out of the Top 6 forwards at higher than team average FA/60. Detroit only has half their Top 6* below team average in FA/60.
While it is more likely to have higher giveaways when you have the puck more, Dallas’s Top line is posting an average TO Delta of 0.27. Detroit’s Top Line? 0.79. Even our “shut down line” (aka the Pitbulls) have some work to do; their average TO Delta is 0.14 compared to Detroit’s defensive zone line’s 0.23.
Yes, there’s a huge disparity between the Stars’ turnovers and the Red Wings’, but is it significant, especially since we’re still above league average in possession & scoring chances?
While there’s no actual correlation to these numbers, it’s pretty visually significant. Detroit’s skaters are mostly grouped tightly, which says to me that they are being coached to all have a similar mentality towards defensive play. Dallas, on the other hand, is all over the place, both in terms of SCA and Giveaways.
Dallas plays a very aggressive game offensively, and it might be backfiring.
These charts give an idea of where the “best” players are deployed for both teams. Detroit has very clear pairings and lines (Detroit’s D-man correlation between ZSO% and TOIC% is 0.93; that’s about as good as you can ask for).
On defense, you have Kronwall & Ericsson facing off against their toughest competition, usually in a defensive zone draw. For forwards, Babcock goes with match ups that shelter his 4th line. Yes, they’re defensive zone starts, but it’s typically against other 4th line players.
Dallas has literally no correlation between Zone Starts and Competition level for forwards, and a mild correlation for the defense.
These deployment decisions have to be conscious coaching decisions, because, like everything Dallas does, these next set of charts fly in the face of conventional wisdom.
This is where you start to see Dallas’ aggression bite them in the butt. Detroit’s numbers fit convention – the more time you spend in the offensive zone, the more likely you are to have the puck, therefore you are less likely to see shots against.
Dallas…does not do that. In fact, Dallas does the opposite. Both forwards and defenders who spend more time in the offensive zone have higher Fenwick Against rates (WTF% strikes again).
Most of our issues seem to be starting in the Offensive Zone, when a shift doesn’t go according to plan. In fact, there’s actually a mild correlation between offensive zone starts and Giveaways/60 for both teams.
So what can we conclude from all this? Well, most of our defensive issues are coming from our poor shot suppression numbers, this is a fact. What’s not being taken into account is that the Dallas Forwards are turning the puck over in the offensive zone far more than the defense is turning it over in the defensive zone.
These turnovers could be what are leading to our high SCA/60 and FA/60.
Detroit’s Top 6 have an average ZSO% Rel of 6.53%, a TO Delta of 0.64, and a FA/60 of 34.03. The Stars RTop 6 have a similar ZSO% Rel of 6.57%, but have instead an average of TO Delta of 0.01 and an FA/60 of 40.27.
It’s the Daley/Jokipakka pair usually in the offensive zone, with a ZSO% Rel of 6.35%. They’re averaging the lowest TO Delta of all the defense, but an FA/60 of 44.75. Is it Daley’s fault that Spezza is turning the puck over twice as much as he is? In fact, they’re the only defensive pairing that is over league average in FA/60, whereas only three of the Stars’ 15 forwards are under league average.
I’m not saying our defense is good – Detroit’s defense is 0.53 GV/60 better with 6 less FA/60, but the Stars need to attack the problem that will give them the best return in the short term.
In short, if you want to fix the Stars’ defense, you have to first fix the Stars’ offense.
*Detroit’s Top 6 was determined based on TOI/game since I don’t track lines for that team and includes Abdelkader, Tatar, Datsyuk, Zetterberg, Helm and Nyquist.
**I am aware that Takeaways and Giveaways as recorded are even more subjective than SOG. But these are the official NHL numbers, and all came from the same data source in an effort to minimize this effect.