Dave Nonis isn’t into Corsi, but why should he be?

It isn’t difficult to understand why statistical analysis hasn’t really caught on in Toronto

Nonis also looks back on a time when the Leafs had a better Corsi rating than they do today – and were losing a lot more games. Back in 2009-10, Toronto generated a lot of shots on net but was let down by its own goaltenders.

That period seemed to create skepticism for the Leafs GM about stats that he believes have grown so much in popularity because of fans and media.

“We were outshooting teams on a nightly basis and losing every night,” Nonis recalled. “Our so-called Corsi stat was probably pretty good and right now our Corsi stat sucks. But we’re winning hockey games.”

Credit to Nonis for actually getting at the crux of the issue here. While I’d rather spend my days researching hockey statistics than defending the validity of statistics, ultimately widespread acceptance of the numbers that myself and many others use is going to come down to how well we predict things.

Just to point to what Nonis is talking about, here is the Leafs’ so-called Corsi (shot attempts for divided by shot attempts for plus shot attempts against) for the last six seasons, along with winning percentage:

  Corsi % Win %
2008 51.6% (10th) 43.9% (24th)
2009 50.3% (13th) 41.5% (26th)
2010 52.8% (4th) 36.6% (29th)
2011 47.8% (25th) 45.1% (23rd)
2012 48.9% (18th) 42.7% (24th)
2013 44.1% (30th) 54.2% (9th)

They don’t line up at all.

Without getting too deep into it or by quoting correlation co-efficients at you, I can say that for the most part, you can predict the future performance of NHL teams by looking at its Corsi number. Edmonton, for instance, started 9-3-2, Minnesota was first in the league in mid-December, and the Los Angeles Kings slumped to a 15-14-4 start, good for the 10th in the West, and fired their coach.

We all know what happened to those teams in the end. The Oilers wound up in the draft lottery and the Wild missed out on that by a single point, and the Kings recovered and went on a tear down the stretch and won the Stanley Cup.

Over the last three seasons, since I’ve been really into this stuff, I’ve watched collapses and rises from teams predicted to do well… just by looking at something as simple as shot differential. For the Leafs, I can understand the hesitation, because they just don’t do what we think they’ll do, and this has gone on for years. When they team was out-shooting opponents, they were losing. Now that they’re being out-shot by historical margins, they’re winning. You take the winning over the losing any day, and why would Dave Nonis mess with success? The bets he’s taken on players have worked out extremely well thus far. David Bolland had recovered back to his 2011 self before he got hurt, Mason Raymond has been a dear so far and David Clarkson has, uh, well, it was a real valuable assist. Point is, the team is winning and there’s no reason to change it.

Funny, Wild GM Chuck Fletcher was quoted in both that Chris Johnston piece, and the coach Mike Yeo was quoted in an Elliotte Friedman 30 Thoughts yesterday at CBC:

“We do some advanced analysis,” Yeo said. “But there wasn’t one or two stats that told us to do this.

“It was recognizing we weren’t going to get better without doing it … We use what the defence gives us. If there is opportunity to carry the puck, [to] allow our players to use their creativity … wait and find the open spaces, they can do it.”

It helps that the Wild are now a much faster team. Up front, there’s been a veteran makeover in Zach Parise and Jason Pominville and an infusion of youth with Charlie Coyle, Mikael Granlund and Nino Niederreiter. These five have joined captain Mikko Koivu to form two strong lines.

That philosophy seems to be opposite of the style of game the team was running with in 2012, when the Wild got off to a hot start despite a dump-and-chase style that didn’t put up good Corsis, but put up good win totals early on. The Wild allow bloggers in the press box, and were probably tuned to the massive debate that was surrounding the validity of their record going on during the early part of the 2011 season. I wouldn’t be too surprised if during the lockout, Fletcher took some time to read back on some of the major points during the large debate and decide to change the organizational philosophy a little and focus a little more on possession. He had a great opportunity, since there were two teams with public zone entry data published by bloggers during the fall of 2012, and one of them happened to be Minnesota.

You can only care about something if you have a reason to care, and as long as the Leafs are winning, why should Nonis give this a second thought? At this point the only concern should be that the Leafs have trouble against some top teams like Boston and Chicago, and the focus should be on banking points during this easy part of the schedule before December when they face all of San Jose, Dallas, Ottawa, Boston, Los Angeles, St. Louis, Chicago and Pittsburgh in a 14-day period. That’s a stretch of the schedule to keep your eye on since it will really test the mettle of this team, especially if they’re still down a couple of horses.

And yes, both you, Mr. Nonis, or you, Mr. Snarky TheLeafsNation Commenter, are perfectly in the right to ignore the mortars being fired between traditional analysts and bloggers, until there’s kind of a reason to, you know, pay attention. Right now, if you’re in Leafs management or a fan… there really isn’t.

Also in the meantime, please refrain from calling much of this stuff “advanced”. Corsi is literally addition and subtraction, and sometimes a little division. It’s nothing a third grader couldn’t do on her cheap, solar-powered calculator.

  • It seems that every year there are a couple of teams that perform better than or worse than their poseession stats indicate they should, and it’s *always* tied to goaltending. When Toronto had solid possession, they had league-worst goaltending. Now they have terrible possession and league-best goaltending. Look at New Jersey and Calgary last year – until Feaster blew the whole thing up, Calgary was a solid possession team with league-worst goaltending. New Jersey was one of best puck-possession teams in the league and had the second-worst goaltending behind Calgary.

    Corsi + League-Average Goaltending = predictable record. Unusually spectacular goaltending or unusually terrible goaltending throws the whole thing out the window, but that only applies to four or five teams every year.

    • Yeah, and that’s why looking at Corsi to evaluate the past drives you insane. There are going to be outliers every year, and the Leafs seemingly transitioned from “awful goaltending” to “amazing goaltending” with nothing in between.

    • This brings up the question of quality vs. Quantity. Is the goaltending that bad or that good, or are certain goalies facing a lot of low quality shots? Also, some teams tend to take higher quality shots with more talented players. So many variables at play here to really out much value in shots based metrics alone.

  • STAN

    Pythagoras would be proud. Mathematics help explain the universe, not to mention to prove that what we just witnessed in a hockey game actually transpired. I mean, just look at those amazing numbers.

    In the Leafs case, however, brute facts are these: as a team they are just not hungry enough for the puck. You don’t have to be Descartes to understand how this works. Pure eye witnessing tells us that most Leafs would rather NOT have the puck, for long, and are far more comfortable trying to defend.

    This is one reason why a guy like Mason Raymond stands out like a sore thumb. When he gets the puck his first instinct isn’t to get rid of it nilly-willy, it’s to have a look around, find some open ice, use his excellent wheels and wait for teammates to get into a position to take a pass. And he has soft hands combined with a brain – if it’s time to get off the ice on a line change and he has the puck, he’ll gently toss it into a corner, not bang it around the boards or slap it straight to an opposing player.

    Kadri has that same instinct, without Raymond’s speed, length and acceleration.

    Gardiner too. Same with Rielly. Kessel occasionally.

    But as a team, there’s no cohesion and the Leafs’ ‘system’ is so predictable that opponents have a laughably easy time keeping them hemmed in their own zone and breaking out of theirs. There’s no doubt that the Leafs have lots of talent, but it’s the same broken, old-school system that goes back to the Pat Quinn days.

    How often do opposing D-men pinch at the blueline along the boards, knowing full well the Leafs ALWAYS fire the puck around the boards as hard as they can, hoping to sneak it across the blueline to relieve pressure? 40 or 50 times a game? And if the D-man missed it, it’s usually an icing call. Go figure.

    So I blame coaching, or the lack thereof. There’s been no improvement, nothing learned.

    How hard is it to study successful teams? To see HOW they maintain possession, how they forecheck and pressure the puck. How they space themselves. How they stay onside, how they don’t carelessly and needlessly ice the puck. Boston is probably the best example.

    The Bruins have a plan, a system and a purpose. It almost seems like they’ve been coached and are all on the same page.

    Advice? STEAL THAT PAGE (and better, more sustainable numbers will follow)

  • Bertly83

    Part of the corsi attraction is it’s simpleness. That’s the same reason it isn’t taken seriously by a lot of “hockey people.” Alone it doesn’t tell you anything. It begs questions, but requires context. Personally, I’m not surprised at their corsi numbers. They’ve been consistently winning and ahead in games and their preferred defensive style certainly isn’t aimed at improving their corsi.

  • Bertly83

    “Without getting too deep into it or by quoting correlation co-efficients at you…”

    Why not? It always seems like the advanced stats folks throw lots of number around but resort to the vaguest generalizations and shy away from statistics when the subject matter becomes really import and the numbers could be really illuminating.

    The correlation co-efficient between team corsi and winning is about 0.62 ( For those unfamiliar, the values of the correlation coefficient range from -1 to 1. >.7 strong relationship, >.5 and

  • Bertly83

    Sorry some wierd code in the comment system cut me off.

    “Without getting too deep into it or by quoting correlation co-efficients at you…”

    Why not? It always seems like the advanced stats folks throw lots of number around but resort to the vaguest generalizations and shy away from statistics when the subject matter becomes really import and the numbers could be really illuminating.
    The correlation co-efficient between team corsi and winning is about 0.62 For those unfamiliar, the values of the correlation coefficient range from -1 to 1. >.7 strong relationship, >.5 and

  • Leaf Fan in Mexico

    One more try:

    “Without getting too deep into it or by quoting correlation co-efficients at you…”

    Why not? It always seems like the advanced stats folks throw lots of number around but resort to the vaguest generalizations and shy away from statistics when the subject matter becomes really import and the numbers could be really illuminating.

    The correlation co-efficient between team corsi and winning is about 0.62 ( For those unfamiliar, the values of the correlation coefficient range from -1 to 1. Less than .7 strong relationship, Between .5 and .7 moderate relationship and less than .5 is a weak relationship)

    So at .62, that’s moderate and somewhat useful, but far from infallible. There is lots of variance and outliers.

    I wish every Corsi quoting person that cuts and pastes from http://www.extaskater.com knew that co-efficient before they typed “unsustainable” or “regression to the mean.”

    “Corsi is literally addition and subtraction, and sometimes a little division. It’s nothing a third grader couldn’t do on her cheap, solar-powered calculator.”

    It’s not quite that bad. It feels like much of the advanced stats discourse seems to be a high school level.

    “Mr. Snarky TheLeafsNation Commenter”

    Me, I just want people to raise their game. There are lots of answers in statistics. We are probably on the right track, but not nearly sophisticated enough, yet. Develop a better mouse trap. Good quantitative work should reveal and explain things so well that it’s almost uplifting to a curious mind.

    • I’m going to do something more in-depth about the coefficients and the predictive nature of the beast, but I didn’t think the spot to do it was here. Those numbers aren’t going to change anybody’s opinion right now.

  • If the team/coaches/fans are looking at shot for and against data rather then shot attempts for and against (corsi) then the incremental value that Corsi or Fenwick provides is minimal. I would conclude that they are aware that the leafs are being handily outshot and perhaps have a different interpretation of what that means about the team. Namely, they do not believe the leafs are being outplayedto the same degree that Corsi imply.

    Further, from what management and coaches have communicated, they prefer and see more value in looking at quality scoring chances for and scoring chances against which follows the “spirit” of shot attempt differentials is trying to capture and is perhaps even superior in some ways. That said, they do not make scoring chances (or whatever data they are tracking) available to the public nor have provided regression data of its predictability.

    So in reality, we don’t have sufficient information to form a definitive opinion. We can assert that leaf management is out to lunch and assume this Corsi analysis is superior to whatever the leafs are doing internally. Or assume that leafs scoring chances data (or whatever “better” stats they track) is superior and Corsi is insufficient for the team. Alternatively, we can also reserve judgement either way and wait until the leafs struggle which validates Corsi (or wait to some point throughout this season where the leafs sustain their strong play and then we may have to question Corsi’s appropriateness to the leafs).

    Basically, one side or the other will have their position proven correct. Either the leafs fail or Corsi will fail to be preditive. That said, whatever occurs, I expect some hardened supporters in both parties will point to events such as injuries or puck luck or other externalties to continue to support their position which will right themselves in the following season.

  • You can “mine” any data to find correlations that work in the past. Will those correlations be predictive in the future?

    The way to prove your case on the predictive value of your case is to predict … I’ve yet to see a blogger post predictions for games around the league … of course maybe the good ones are just betting their edge and raking.

    Ive seen a few PHD`s come into the financial markets and think they are going to slay the markets with intellect, only to find the dynamic and changing nature of human endeavor makes it not as easy peasy as static math formulas.

    But yes stats can have predictive value. Three adages

    1. It will never be as simple as if A occurs then B results

    2. simple as possible, never simpler

    3. if its obvious, its obviously wrong

    If you are here you are making some dollars betting your stats edge not defending historical data mining on blogs.

    • I’m taking an issue with this because I’m doing a lot of finance reading, but there’s a clear difference between economics and hockey.

      Hockey is a closed environment. It has a limited flow of data and the data is directly related to the environment itself. For something like finance, you’re looking at all of these outside factors that impact the market that you can’t predict. Hockey has strict rules, a time limit, and speculators don’t make big shifts that change the value of a goal or a shot from day to day.

      It’s difficult to quantify a lot of things about hockey for sure, but it’s easier to predict future Corsi (it’s always very similar to past Corsi) than it is to predict future shooting % or save %. That’s sort of where the value comes from.

  • Leaf Fan in Mexico

    The Leafs beat the teams they should be able to beat (aka not great teams) and have trouble with or, at best, steal a game against the elite teams. Does Corsi (or any other stat) predict what talent can do or does talent predict what Corsi will be. Either way, we have a couple years on this team to find out….

  • Leaf Fan in Mexico

    Nonis just recognizes Corsi for what it is, just another stat. This doesn’t mean that it’s useless but it’s no miracle stat that’s any more important than all the other ones.

  • Leaf Fan in Mexico

    Not to mention that the biggest issue here is the fact that relying on stats to make predictions for hockey is somewhat laughable. In the game of hockey with so many intangibles and unpredictable moments, things that have happened in the last few years seem pretty irrelevant to me big picture-wise. You should be able to tell me why you think a player is good without quoting stats once. If you need to back up your points about the player, then perhaps could use stats to do so but talking about a player’s worth based on his old stats is just crazy.

      • Leaf Fan in Mexico

        Easy, I’ll tell you why SIDNEY* is good.

        He’s an explosive skater, fast yes but no one is close to his one-two-three-go stride. He’s got an amazing shot, both accurate and hard which is a combination few have. His hockey IQ has got to be the highest in the league (his ability to see the ice and read the play) this allows him to think two or three steps ahead of the play which opens up space and opportunity for him to show off his playmaking/sniping ability. Not to mention the guy has amazing hands and is in such good shape phyiscally that he can withstand a shoulder to shoulder hit with almost anyone in the league… Which is a weakness for most skilled players. I could write paragraphs more if I felt like it, seeing as how this is Crosby we’re talking about.

        All this from someone who isn’t a fan of Crosby.

  • Leaf Fan in Mexico

    He should care about corsi just like he should care about other metrics: they aren’t playing battleship. You shouldn’t be hurling things at the board, seeing what hits, and making it up from there. “This isn’t working, so let’s blindly try something else without identifying why it wasn’t working” is never, ever, ever a strategy in ANYTHING.

    Hockey people understand the game, as they tell us constantly, and if they understood the game and what makes a winning team they’d have clearly defined goals and requirements. So either go on record saying that getting outshot, outworked, and never having the puck but rocking .940 SV% is what wins cups, admit that corsi is in the ballpark of identifying that the Leafs have a bit of a problem, or admit they have no idea what they’re doing.

    • Actually, trial and error as you described it is used everyday in various experiments… Perhaps not with hockey but it’s a decently common method when starting from scratch.

      A person who has been around the sport (playing, coaching, watching, etc) their entire lives would have personal experience to go with their opinion, experience that someone who hasn’t played the game wouldn’t have. Seem easy enough to understand right?

      Now to suggest that just because we have this experience, we should be able to come up with some algorithm or fail-safe plan that guarantees a Stanley Cup win or else every point we make is moot, is ridiculous. You saying that just highlights your ignorance. Everyone knows that the Leafs have a problem with getting out-shot but I don’t need corsi to tell me so. Seeing as how they keep winning despite many setbacks, I would say it’s much less of a problem then you’re convinced it is.