Archive for Analysis
One of the major drawbacks of using raw puck possession (Corsi/Fenwick) numbers is that it doesn’t account for game and score situations. More accurate measures have been made available by using these numbers at even strength –where most of the game is played– and in score close situations. Score close –where the game is within one goal in the first two periods or tied in the third– is what has been the driving force for the predictive measure of the stats.
Score close situations wasn’t perfect because it limits the number of events recorded, but since it was the most accurate metric that was readily available, we leveraged it*. The most accurate has always been score adjusted puck possession, but it was only available via manual calculations. Luckily, the indispensable war-on-ice added this calculation to their page, and now we can use this to more accurately assess player performance.
*-It’s worth noting that we are limited by not only what is available, but what is available for us to analyze/evaluate. Several intelligent people have manually tracked/adjusted some of these stats, but it’s not something that has been available for all due to the manual nature of that tracking. There will be many more advancements as more data –specifically player movement tracking– is available.
The preference for score adjusted puck possession is still the same as with regular (non-score adjusted) possession: You want to be on the positive side of 50% for CF%/FF% and on the positive side of 0%. I sorted by difference between score adjusted and raw (right column) to show how the numbers were effected. First the forwards:
|Skater||ES rel CF%||ES rel SA CF%||Difference|
The most interesting aspect here is that Jesper Fast has the biggest improvement when taking score into effect. His numbers are weighted down by 25% offensive zone starts though. Tanner Glass is at least a pleasant surprise at the second largest differential, but he’s seeing 45% OZ starts, so it’s less impressive than Fast and Dominic Moore’s improvements.
Now the defense:
|Skater||ES rel CF%||ES rel SA CF%||Difference|
Mike Kostka got a bad rap with those awful turnovers during his short stay. The eye test catches big things like that, but not the small, subtle things done regularly. I’m not saying Kostka should be in the lineup, I’m just saying he wasn’t the huge tire fire everyone made him out to be. Unsurprisingly, Ryan McDonagh takes the biggest leap forward among the regulars (these tables contain players with > 100 minutes played).
My biggest concern, aside from Girardi, is Dan Boyle. Boyle has been underwhelming this season, and his recent demotion to bottom pairing has been evidence of that. There have been rumblings that he’s playing hurt, and I hope that’s the case since his play has been poor this year. The Rangers have been able to overcome Boyle’s ineffectiveness because of the impressive play of Kevin Klein.
Going forward, we are going to use score adjusted numbers to analyze players and teams, since they are a more accurate representation of a player’s effectiveness. However like all possession numbers, deployment –zone starts and quality of competition/teammates– needs to be factored into the equation as well.
In case you missed it yesterday, I was asked why the Rangers kept Dan Girardi over Anton Stralman. After going into qualitative (eye test) and quantitative (#fancystats) analysis, comparing Girardi to Stralman, it was clear to see that Stralman was always the better choice. However, one thing that was not as clear was whether the Rangers actually chose Girardi over Stralman. It’s something I noted in the last line of that post:
It’s worth noting that I think the Rangers chose Boyle over Stralman, and wanted to keep Girardi regardless.
It makes sense that the Rangers didn’t want to deal their captain (Ryan Callahan) and another leader in Girardi that would be given the ‘A’ following the eventual buyout of Brad Richards. So, keeping Girardi was always, in my eyes, in their plan. However, keeping Girardi did not address the obvious need for a defenseman that could run the powerplay. With Dan Boyle on his way out in San Jose, the fit was obvious.
I received one question this week in the mailbag, and it was a doozy, so this entire post is dedicated to it. As always, email me if you have any questions, and I’ll be sure to address each one.
BV, this is a loaded question, to which we can break it down into several parts. Addressing the first part is the easy part: Keeping Girardi over Stralman had a lot to do with perceived value versus market value, and the writing was on the wall when Stralman rejected the Rangers offer of three years, $9 million. The Rangers valued him at $4 million, which is what Larry Brooks reported. That was $500,000 less than what he got from Tampa Bay over five years. The Rangers didn’t want to go that long or that high.
In case you’ve missed it, and if you’ve been watching lately, you haven’t, the New York Rangers are playing some pretty inconsistent hockey. Some point to last year’s start as a sign of hope, but this is an entirely different situation from last year. The Rangers had solid possession numbers last year at this time, but were experiencing a combination of bad luck (bad SV% and SH%) and learning a new defensive zone coverage system.
This year, the Rangers are one of the worst possession teams in the league (20th – 49.5% FF close), but their PDO (combination of SH%+SV%, which usually regresses to 1000) is at 1000, meaning they likely won’t get any better there. Even when Ryan McDonagh comes back to the lineup, he’s replacing Matt Hunwick, who’s actually been one of the better defenders from a possession standpoint.
Long story short: That impressive streak we saw from January through the end of the year last season may not come this year. In fact, it likely won’t come.
The Rangers were absolutely decimated by injuries in October and November, at one point skating without their top line center and four of their top-six defensemen. When you’re dressing options seven, eight, nine, and ten on the blue line, you need better contributions from those in the lineup. Luckily for the Rangers, four guys stepped up their games offensively to fill the void. Rick Nash (12-6-18), Derick Brassard (6-7-13), Martin St. Louis (6-6-12), and Kevin Klein (3-2-5) were major contributors in keeping the Rangers afloat.
Nash has been a force all season, looking like the Rick Nash the Rangers traded for in 2012. MSL has been contributing at his normal career pace despite playing a good number of the games at center this year. Brassard has shown great chemistry with them as well, looking like that $5 million center they re-signed in the offseason. As for Klein, he’s never cracked the five goal mark in a season, and he looks to be well on his way this year.
But these four players have one thing in common – unsustainable SH%, that is surely to regress as the season bears on.
There was a time last season that Henrik Lundqvist was playing so poorly, and Cam Talbot was playing so well, that a very small but very vocal segment of the fan base was calling for a change at the number one spot. Imagine that. Crazy, right? But, it happened. Small sample sizes can do wacky things to people’s perceptions. Talbot had a phenomenal 2013-2014 season, but has struggled so far (relatively speaking) in the new campaign.
Last year, Talbot ended the season with a 1.64 GAA and a .941 save percentage in 21 games played. If he had put up those numbers over a starter’s workload, he would have run away with the Vezina. We all knew (hopefully) that these flawed metrics, although nice to see from our backup, were not reflective of his true talent level. In fairness, they aren’t reflective of anyone’s true talent level.
In 4 games so far this season, Talbot’s GAA has ballooned to 3.48 and his save percentage has slid to .880. Neither of those numbers are particularly pretty. I’ve seen comments on the Twitters and other social media about how hard regression is hitting Talbot, which naturally begs the question: what is the mean he is regressing to?
If you’ve been reading this blog for a while, you know that we like to combine what we see in the goal breakdowns with what we see in the stats. We are a bit limited by time constraints (thus no gifs or pictures of non-goal events), but we point out various areas where the Rangers have defensive breakdowns that lead to goals. While only pointing out goals is limited –If someone wants to pay me to do this full-time, I’ll gladly point out more plays. But I don’t have the time. So those of you complaining, quit it, it’s annoying and rather pointless– it is useful.
The defensive breakdowns are what we see with the eye test. And while the eye test should never be ignored, it does tend to focus on big screw-ups and lend itself to bias. It’s why defense should also be measured by two key metrics: One that we use regularly (relative Corsi) and one that we admittedly don’t use enough (Corsi Against).
Relative Corsi you are familiar with, as it measures possession on the ice relative to the rest of the team when off the ice. The concept here is that the higher the number, the more time you have the puck in the offensive zone, thus limiting opposition opportunities. Corsi Against measures shot suppression, which is a measure of relative Corsi, and shows us the shot attempts against while on the ice.
I’m sure you’ve noticed this by now, but the New York Rangers stink on faceoffs. They are at 47.2% right now as a team. Of the players that have taken at least 50 faceoffs, only Derick Brassard (55.4%) and Dominic Moore (53.3%) are above 50%. Martin St. Louis (43.6%) and Kevin Hates (24.6%, ouch) bring up the rear for players that have taken 50 faceoffs. Derek Stepan isn’t a 50% faceoff guy, so his return won’t really help in that department.
But how much does this affect the on-ice product?
Statistical Sports Consulting printed a study on the effect of faceoffs on goals, and the results are pretty interesting. They first measured the faceoff differential to yield a goal differential, then measured the probability of winning a faceoff (Note: Not 50/50, there’s skill involved).
Over the past few years, the debate has grown more intense about the validity and reliance on #fancystats. The concept of quantifying the game has been a theme we have run with around here, albeit with the conceit that there is no perfect, all-knowing stat that can be universally relied upon to demonstrate a player’s ability level.
Statistics trying to quantify human athletic performance are inherently limited. There are very human characteristics in play; such as intelligence, judgment, emotion, situational awareness, etc. It makes it difficult to measure performance as if they were vital signs. I think that to fully expect that level of quantification or to vilify the statistic for being unable to is missing the point.
Much like politics, I think the emergence of these statistics and the resistance to adoption has pushed the two positions out to the extremes. The old school hockey community has written them off or marginalized their effectiveness, citing “games are played on the ice, not on a spreadsheet”, or taking pot shots at the Maple Leafs for hiring Kyle Dubas for their Assistant GM position, and various stats writers to make up a new analytics department. Read More→
If there was one thing proven about Alain Vigneault’s coaching style over the years, it’s that he tactically deploys his lines (after a whistle) depending on the zone start and the opposing line sent out. Last year we saw the Brian Boyle-Dominic Moore-Derek Dorsett line get the majority of their starts in the defensive zone, while the Mats Zuccarello-Derick Brassard-Benoit Pouliot line received primarily offensive zone starts. It is a common practice among coaches.
This year the lines have been shaken up with the turnover, and the matchups will likely change, both in zone starts and in quality of competition faced. Using the lines from practice the past few days, we can make some guesses.