E=MC Corsi: A look at some Panthers advanced statistics
The Panthers' recent run of victories has led to some widespread criticism of the team's play based on advanced metrics. Let's take a look at the numbers.
We are Florida Panthers fans. We are more or less the hockey equivalent of the Statue of Liberty's welcome- the tired, the poor....the huddled masses? Well, you get the point. So we could be forgiven if we have a large chip on our collective shoulder. Said chip has gone to DEFCON 1 during the Panthers recent winning streak, as the barrels of many NHL writer's pens were leveled against....yes, you read that correctly, against the team with the longest win streak in the league this season. What was their weapon of choice? Advanced metrics.
Lets be blunt here- the metrics do measure a very valuable piece of information: possession. In the NFL they call this time of possession and time of drive. The dreaded New England Patriots have made a science of long drives that deprive opponents of more than 4 plays of offense in a quarter. You can't win if you don't have the ball. Soccer (football for our LBC reader(s) in the United Kingdom) has long considered time of possession the key to playing the game- even if it results in an outcome referred to as "nil-nil."
The NHL is a newer entrant to the measure of possession- and in a game where- due to the nature of constant play and changes in possession- the very measure of possession is difficult. Some very intelligent hockey folks came up with a means to measure possession we now refer to as Corsi and Fenwick. Both of these metrics use the number of shots-for and against to determine who had more possession of the puck during the game. This is a sound philosophy. If a team has the puck they are likely to shoot it, and the other team is likely to do the same, but you can't shoot it without possessing it. This is also valid.
But... the metrics folks started noticing some things that were altering the truth behind the tale: as games progressed, teams with the lead were often times sitting back and not taking offensive chances, resulting in opponents getting more shots off. To deal with this issue they developed "score-adjusted Corsi" concepts. Some very smart hockey folks have taken this even a step further- charting where all those shots are coming from. That information has become more available to the general public in the last two years. This was important because much like the NBA had seen, the NHL saw some coaches develop systems that allowed large numbers of shots from what they considered less dangerous areas of the ice. At the same time, the theory was that these teams would put more emphasis into guarding the more dangerous scoring areas of the ice. Thus, the shot stats might be heavily skewed, but in ways that did not accurately reflect the game's possession characteristics.
Shot charts show the location of shots-for and against so that a person can look more closely at what sort of possession a team was gaining during a game. Shots from the outside and bad angles will be reflected in Corsi as good possession, but they really do not indicate scoring chances. Fenwick goes a bit farther- including blocked and missed shots, this being especially important as some NHL coaches turned their teams into shot-blocking machines while others told players not to risk injury in that fashion. Once again however- even bad shots are included in the possession calculation.
Turning to the Panthers- and their critics in the metrics community- there have been some bad games that resulted in victories during the win streak. The Panthers acknowledged as much after the Ottawa victory- with Gerard Gallant stating they played tired and were outplayed, but got the win. Erik Gudbranson said the same. The same was true in the win over Edmonton. But that's what good teams do- they win ugly at times. As my fellow hockey fan Middlebury-Doug said after an ugly win by the Cats over the Wild, "that's a game veteran players scratch out a win for their team." Let me assure you, it is a fact- that the team is 100% committed to using advanced metrics all the way from ownership to the team's legal counsel and they are firm believers in the necessity of statistics. This is no Patrick Roy situation where they simply thumb their nose at the numbers. The team is well aware of where the problematic stats are, and is working to correct what needs to be corrected. In the meantime however, it's not all as bad as it seems.
Let's start with this- the major concern the analytics folks have with the Panthers is that the differential in shots-for and shots-against is large. As of this writing, the Panthers are 15th in the league in shots-against and 28th in the league in shots-for. That results in terrible Corsi and Fenwick ratings (which at their root are shot differential measurements). Stanley Cup contenders normally have very good Corsi and Fenwick ratings.
Analytics folks, or at least those in the media, also have a concern that the Panthers winning is merely the result of stellar goaltending bailing out a poor team. The Cats did boast the league's top save percentage after the Vancouver game, with a .928 shared between Roberto Luongo and Al Montoya. That high save percentage leads to an above average PDO (102.6), good for 2nd best in the league. High PDO's can not normally be maintained and fall back to the average. PDO measures the team's shooting percentage and save percentage. The team's shooting percentage sits at 9.50% which is in the top-10 in the league.
I am not going to even discuss the Calgary game in this article, as the hope is that it was a one-off at the end of a road trip and is itself a statistical outlier. Whether this is true or not we cannot determine until the Tampa Bay game on Sunday, but for now it is not worth including as it was so odd and an obvious result of its circumstances. If that type of play continues in Tampa and during the homestand, than we'll address it at that time.
Finally, the concern is that the Panthers are a team deriving all possession and offensive play from the top-3 forwards and top pairing defense of Aaron Ekblad and Brian Campbell. The first thing to note in response is that Ekblad and Campbell have not been paired for the majority of the winning streak. Ekblad has been paired with Dmitry Kulikov and Campbell has been paired with Gudbranson. But more importantly, the Panthers have 8 players (7 forwards and 1 defenseman) who are on pace to have 40 point seasons.
Should all this be a cause for concern? Probably. Except there is a lot more going on here than meets the eye. And there is one other statistic that is being ignored: goal differential. The Panthers are 3rd in the league in goal differential, having scored 113 goals, while only giving up 91 (+22 differential for those weak in math). The company they are keeping in that category is excellent, as they are surrounded by Dallas, Washington, Los Angeles and Chicago.
What it looks like on a few stats is that the Panthers give up a league average number of shots-against, have tops in the league goaltending, barely ever possess the puck on offense- which results in no shots-for, and get lucky on the few shots they do take to score their goals. The reality is a lot more complicated than that.
The fluid nature of statistics:
Some of those ugly wins during the streak moved the Panthers down the league-wide stats list in shots-against per game, and shots-for per game. At the time of the Yahoo Sports Puck Daddy article, that was more-or-less the spark that ignited this fire, the criticism was that the Panthers were 16th in the league and therefore keeping poor company with some bad teams in the shots-against per game metric. But the Panthers would move between 16th and 13th in the league around that time period, and after the Vancouver game would find themselves once again in 15th place in the league with 29.4 shots-against per game.
You had best understand that these numbers are very fluid. Some of that fluidness is based on what other teams are doing: if the league's other teams have a stretch of games where they give up more, or less, shots per game, or play high-shooting teams, or low shooting teams, it will adjust the Panthers position relative to them as well. As you will see later in this article, the Panthers have not really changed the way they are game playing, although execution will shift from bad to good or visa versa. The numbers and their place on the charts vary by game and by what other teams are doing. A week ago the Panthers were 11th in the league in shooting percentage, at roughly 9.15% and after the Edmonton game, were in 8th place with a 9.50%, and after the Vancouver game they moved into 7th place with.....the same 9.50%. Say it with me people: fluidness
If this article is to indeed be a response to those who are roasting the Cats for being 15th in the league in shots-against per-game and 28th in shots-for per game (down from 24th just a few games ago)- we should start with this: virtually every season there is a statistical outlier that gets into the playoffs. Some of those outliers are then used as adjectives to describe any team that posts poor Corsi and Fenwick numbers, as in: "the Panthers are this year's 2013-14 Avalanche," or something of that ilk. A recent CBS Sports article about the Panthers winning streak mentioned-- of course-- the 2013-14 Avalanche, as well as some of the other teams I am about to discuss. The problem is that the badge does not always fit, as there are many different factors and numbers in play than simply shot differentials.
To be sure, the 2013-14 Avalanche had some crazy results: 25th in the league's regular season in shots-against per game (32.7) and 20th in shots-for per game (29.5). That's bottom-10 in the league in both categories. In effect, the Avalanche rode a soft division and exceptional goaltending- they were 5th in the league in save percentage (.919), and Semyon Varlamov posted a .927 save percentage on the season, to the playoffs. Varlamov was 3rd in the league in goals-against-average (2.41) while seeing an average of 31.95 shots-against-per-game. Clearly, excellent goalie play propelled the Avs into the playoffs that season. To move up even one place on the shots-against chart they would have had to have a full 1.6 fewer shots per game-- just to get into 24th place. Clearly, the Avalanche were not tightly grouped with other competitive teams in that category- and to be sure- shots against is typically a very important metric to look at, as thats how opponents score.
Last season, the Avalanche crashed back down to the non-playoff regions of the earth. But that is not to say another team did not take their place. The Ottawa Senators are the team analytics folks ended up hating- as they blew up the stats. 9th in the league in shots for-per game with 31.0, but 25th in the league in shots-against with an average 32.1 per game. Just like the Avalanche, last season's Senators rode excellent goaltending into the playoffs (along with some fantastic goal scoring). Craig Anderson posted a .923 save percentage and 2.30 goals-against average while seeing an average of 32.4 shots per game (1134 shots against in 35 games played). His fellow goaltender Andrew Hammond posted a .941 save percentage and 1.79 goals against average through 24 games. Because of the importance of defense to success- the Senators numbers were backwards of where a team would want them. It is more important to be good at preventing shots-against than it is to be great in shots-for.
But Ottawa was not alone. The metrics folks roasted the Calgary Flames ride to the playoffs as well. This one is a bit closer to the Panthers- and on the right side of things, as it is a better result on the defensive side of the game.The Flames were 12th in the league in shots-against per game with 29.2, but..... let me put some emphasis here-- because nobody really pointed this out:
The Flames were one shot-against per-game away from being 6th best in the league. Let me say that another way- if you need: if the Flames gave up one fewer shot-against per game, we would have been talking about them as a top-10, almost top-5 team in the league last year on the defensive side of the puck. Thats how close the numbers were. Ahh, you say, but they were 28th in the league in shots-for per game with 27.5. To which I would point you to goals scored last season: the Flames scored 241, good for 8th in the league. Their team PDO was on the high side- at 101, 5th best in the league, meaning they did get some puck luck. Nonetheless, what wins games? Defense, of course, and the Flames did a good job of protecting Jonas Hiller, who finished the season with a .918 save percentage and 2.36 goals against average. And despite few shots on net on the offensive side, they were in the top-10 in goals scored.
What other teams were giving up 29 or more shots per game last season that may surprise you? Well Carey Price saw 29.59 shots against per game and still marched Montreal to the playoffs with his 1.96 goals against average and .933 save percentage. Steve Mason saw 29.21 shots against per game in Philly. Lets go to the surprise factor though:
How about this one: Corey Crawford of the world champion Chicago Blackhawks faced 29.14 shots against per game last season. Did you see that one coming? Likely the answer was "no," in part because you were not concentrated by the NHL media on that one. But the Toronto-centric NHL media definitely kept you far more informed than anyone wanted to be about this one:
The Maple Leafs fired coach Randy Carlyle last season, and Jonathan Bernier saw an average of 29.91 shots against per game, about .7 more than Crawford. More on that shortly.
The defensive oriented New Jersey Devils let Cory Schneider face an average of 28.72 shots against per game last season. Is Washington better this year than last? Seems it to me, yet Braden Holtby faced only 28.0 shots-against last season per game, on average, but is facing 28.11 this year, as the Caps give up an average of 28.7 shots against per game. The Panthers give up an average of 29.4 per game, a difference of .7 between these two Eastern Conference leaders.
Parity and its effect on the pack:
So what is our situation here in Florida? The Panthers got roasted by some analytics folks in part because they were reputed to be in the middle of the pack in the league in shots-against, and in the 10 worst in shots-for per game and were riding a hot goaltender, and that can't be sustained. Except it's not that simple. Hold on tight, as this could get complicated before we finish.
The critics pointed out that the Panthers are 15th in the league in shots-against-per game (and were 16th at the time the roasting reached its high point), and they say that's not good. But, what is the difference between the 15th place team and the top 10 in the league in shots-against? Less than half a shot per game, .4 shots. That is what separates the Cats from a top-10 spot as I write this. 1 fewer shot per game would put the Panthers in 8th place and on the heels of the Los Angeles Kings. Roberto Luongo is facing an average of 29.69 shots against per game, and has a 2.06 goals against average and .931 save percentage. Those are great numbers. But he is also in the company of Carey Price and Corey Crawford from last season in the number of shots he is facing.
Thats not all though, as he is also facing a similar number of shots THIS year to Corey Crawford (29.26), Petr Mrazek (29.00), Tuukka Rask (29.34) and better than Henrik Lundqvist (30.17) and Craig Anderson (33.00).
Location, location, location: Shot location is highly favorable, and that is the key:
Would it be better if the team did not give up as many shots against? Of course. But to look at where shots are being taken from, and how many are getting through to Luongo and Montoya, it becomes evident that the Cats are not giving up a large number of chances in the high percentage areas of the ice ("the house"). That has a positive impact on save percentage. Are the goalies propping the Panthers up, or are the Panthers making great goalies look even better? Probably a bit of both.
When you talk about shot location, you have to be careful, because remember- Randy Carlyle was run out of Toronto in part for utilizing a system that allowed too many shots against. To be sure there were other reasons as well, most notably that the Leafs were just a lousy team that played poorly, and he lost the locker room. They were last in the league in shots-against per game (35.9) and 25th in shots for per game (27.9), and Bernier was not quite up to the task of bailing the team out even though he posted a 2.69 GAA and .923 save percentage. The team's save percentage of .914 was 9th best in the league- so the goaltending was a positive point for the team. Nonetheless, Toronto was giving up too many shots and not taking enough. So they hired possession guru Mike Babcock to coach the team.
The media tends to coin any team that allows shots in large numbers from less dangerous areas of the ice, a Carlyle style team that plays poor hockey. Other times teams that play that way are compared to Patrick Roy and the Colorado Avalanche. In some cases that can be accurate. But in others, where the team plays the system better, or the system is simply better designed, with some modifications, it is a legitimate style of play. More on this to come....
What exactly a good number of average shots-against per game is seems to vary as well. Babcock himself has led teams that allow what analytics folks are referring to (in the Panthers case) as too many shots against. The guru coached a 2013-14 Detroit team that gave up an average of 29.3 shots against per game (13th in the league), and 28.3 in 2014-15 (8th in the league). Not bad, when you consider some of the best teams in the league last season posted the following in the shots-against per game department: Montreal 30.1; Chicago 30.2; New York Rangers 29.5; Ottawa 32.1; Washington 29.3; Nashville 28.3; Tampa 27.9. Somebody needs to talk to Joel Quenneville....
The more important aspect of shots, whether for or against- is location. There is a reason that only one NHL defenseman playing regular numbers of games is in the top-100 players in the league in shooting percentage. That reason is: where they typically shoot from. That defenseman in the top-100 is South Florida native and Philadelphia wunderkid Shayne Gostisbehere, who sports a 15.6% shooting percentage at this moment. Even with respect to the human highlight reel Gostisbehere is though, there are reasons he is so high: 1) that he is often carrying the puck deep into the offensive zone, or jumping into the rush like a forward and therefore shooting from higher percentage areas of the ice; and 2) that he is getting lots of power-play goals (4 of his 7 goals scored this season), which opens lanes from the point that are not typically available at 5-on-5.
Aside from the "ghost-bear," most defensemen are shooting from outside the "house," from lower percentage areas of the ice. As a result, the defensemen who have higher shooting percentages than the average D-man are typically getting their goals on the power play. Eight of Shea Weber's 11 goals were power-play markers; 4 of Duncan Keith's 7 goals were on the power play; 6 of Drew Doughty's 7 goals were power-play goals; 4 of Kevin Shattenkirk's 8 goals were PP goals. Other defensemen who have scored enough non-power play goals to matter offensively have lower shooting percentages. Obviously, shots from the point, or outside the house, in today's NHL of packed middles and shot blockers have a small chance of becoming goals.
When you look to the shot location charts for the Panthers last several games, the shots-for are coming from high percentage areas of the ice, so they cannot be simply dismissed as riding a wave of puck luck. Against Edmonton on January 10th, the Panthers only put 14 shots on goal- fewer even than Minnesota and New Jersey did in their games that same night (yes- many teams are playing possession and prevention systems this season). But the Panthers also got 18 shot attempts in Edmonton's "house" area of the ice, compared to 24 shots by Edmonton in Florida's "house." The Panthers scored 2 goals in the house and caused Cam Talbot to make 5 saves on shots from that area. That is the same number of saves Al Montoya had to make on shots in Florida's house.
Against Ottawa, on January 7th, Roberto Luongo had to make only 7 saves on shots in the house. The Panthers attempted 20 shots within Ottawa's house, 3 of which became goals, and 7 more of which Craig Anderson had to save. Looking at the game report, the Panthers were badly outshot- yet the goalies made the same number of saves on shots from high percentage areas of the ice.
It was a little worse against Buffalo on January 5th, as Lu had to make 12 saves on shots in the house. But Buffalo took 32 shots on net in that game. Skip forward to the Vancouver game on January 11th: the Panthers were outshot by Vancouver 30 to 28 in that game. Vancouver got 21 shots off in Florida's house, with 2 goals, 6 blocked, 2 missed, and Roberto having to make 11 saves. Not a great game by any stretch of the imagination. Florida fired off 25 shots in Vancouver's house, resulting in 2 goals, 3 missed, 6 blocked, and Jacob Markstrom having to make 14 saves. Markstrom had to make more high quality saves in that loss than Roberto Luongo did.
Before we beat a dead horse, we will reach the obvious conclusion here: the Panthers are doing two things that a simple shot count and/or shot differential count is not picking up on: they are forcing opponents to shoot from less productive areas of the ice; and they are (themselves)- not shooting from unproductive areas of the ice. That is a recipe for analytical disaster- as the metrics are reading a massive per game shot differential that the statistics watchers detest.
After the Edmonton game, the Panthers sat near the bottom of the league in Corsi-for and Fenwick-for, both of which are shot differential measurements. But what the numbers are missing is that the Panthers are doing a good job of quality shot suppression. More important is that the numbers crunchers are missing where the majority of shots-for and against are being taken from. It's not full "Carlyle," --as the team is not taking as many shots or giving up as many per game-- but it is definitely a shot quality system being played.
To get a true understanding for how the Panthers are stacking up, it is dangerous to use simply the shots-for and against- or the metrics that measure shot differential. Instead, the tedious leg work of going to the shot location charts must be done. My nerdy secret is that I have cut out a 3x5 card to place over the computer screen to show me only the shot events within the house so that I get an accurate number from game to game, but it is nothing short of tiresome to manually count- and I am not going to go back and look for trends for Calgary or Ottawa or Colorado from the past, or even the Panthers from earlier this season. Yes- it is a lot of work to figure out what system a coach may be actually utilizing, and to be sure, this is very much a coached system of play.
Bill Waters- a guest analyst on XM Home Ice, the NHL's XM radio station, stated after the Edmonton game, that he had no idea just how good a coach Gerard Gallant was until recently. In retrospect, he indicated he should have known, especially after the work he did in Montreal, but he unequivocally stated that the Panthers most obvious trait this season is that they are very well coached. Indeed- looking at the numbers we have pointed out here, the endless circle comes to mind- the yin and the yang- as the system lets our goalies stop lower percentage shots to boost their save percentages and GAA, but the goalies stellar play also circles back around and saves the team as they do give up shots. Some type of ancient proverb at work there, eh?
At the other end of the ice, think of how many times the Panthers have gotten a fantastic cycle going but got no shots on goal out of it, or few shots. That is possession- and in hockey terms- lengthy possession- that is not being noted on analytics charts. But the location charts tell a very different story. Some pundits are in fact paying closer attention to some more important numbers, such as goal differential:
Overshadowed by Ovechkin but Florida Panthers win streak now at 12. They've opened a five-point lead in Atlantic. Plus-27 goal differential.
— Craig Custance (@CraigCustance) January 11, 2016
That, my LBC friends, is one heck of a good system being coached by Gerard Gallant, and a lot of folks are just plain missing it by only looking at shot totals and differentials. As in real estate, its all in the location.