• Mike Green, Karl Alzner/John Carlson: Who Stops The Opposition’s Best?

    by  • June 7, 2014 • Hockey • 25 Comments

    I did some posts a few days ago looking at famous defencemen and CorsiRel. I touched on three Washington Capitals defencemen: Karl Alzner, John Carlson and Mike Green. I’m going to reproduce the graphs here and my comments about them.

    Screen Shot 2014-05-31 at 2.31.36 AM

    I know he was at Team Canada’s camp in Calgary last August; I don’t really see it.

    Screen Shot 2014-05-31 at 2.35.48 AM

    Yeah, I don’t really get the Carlson popularity either. His Corsi falls in lockstep with Washington’s.

    Screen Shot 2014-05-31 at 2.41.40 AM

    I’m going to have to look at some open play numbers for Green, Carlson and Alzner. Washington’s fall back to earth over the past few years has affected all of them, but Green still stands well above the Caps when he’s not on the ice, which isn’t true of the other two.

    DCsportsfan85 had this to say:

    Green is definitely the best puck moving defenseman for the caps and drives the play. He also is given more offensive zone starts and plays against significantly weaker competition than Alzner/Carlson.

    Zone adj. possession metrics would paint a better picture, but that still doesn’t account for the quality of competition.

    OK. I figured that I’d take a look at this. I identified 23 star forwards that I figured coaches would want to match their best defencemen against: E. Staal, Malkin, Bergeron, Kessel, Crosby, Nash, D. Sedin, Hall, Giroux, Pacioretty, Toews, J. Thornton, Kopitar, Toews, St. Louis, Zetterberg, Datsyuk, Tavares, Parise, Backes, Seguin, Ladd, Getzlaf and Doan. Murderer’s row.

    For those of you who don’t follow the Capitals closely, Alzner/Carlson generally play as a pairing. Green then plays with someone else, most commonly Dmitry Orlov last year. So the Capitals played 824.2 5v5 minutes against those forwards last year. The defensive matchups break down like so.

    Screen Shot 2014-06-07 at 12.50.55 AM

    So DCsportsfan85 has a point. The Caps definitely shelter Mike Green from the corrosive effects of the other team’s star forwards. They can’t do it completely but Carlson/Alzner are taking on a big chunk of the burden. So what would happen if we looked at the results put up by these seven pairings when there’s a star forward on the ice? Surely Carlson/Alzner will be better, right?

    Screen Shot 2014-06-07 at 1.35.39 AM

    Well, it’s close enough that the difference is basically nothing but it turns out that the answer is “Not really” – it doesn’t seem to make much of a difference whether or not Green is out or Alzner/Carlson – the Capitals do about the same. I mentioned above that I wanted to look at some open play Corsi% stuff for these guys, so I decided to do that for star forwards.

    Screen Shot 2014-06-07 at 1.35.51 AM

    Now that’s a little more interesting. All of a sudden, a large gap opens up between Green and Alzner/Carlson, with the Capitals doing much better when Green is on the ice than they do with the Alzner/Carlson pairing. It looks to me like the overall Corsi% looks much closer because of the Caps habit of putting Alzner/Carlson on the ice for offensive zone faceoffs against star forwards. Those are basically structural Corsi fests, so they look better than they actually are.

    I’ve put together a breakdown of how the Caps used these guys for faceoffs against star forwards along with the faceoff percentages when they were on the ice against star forwards. It is, I think, telling:

    Screen Shot 2014-06-07 at 2.07.17 AM

    The gist of it is this: Carlson/Alzner were on the ice for many, many more won OZ draws than Green and more OZ draws relative to their other draws. The Caps also had a horrible faceoff percentage with Green on the ice against a star forward for some reason. This explains why the overall Corsi% is closer than the open play Corsi%, I think – the numbers are misleading but to Green’s detriment.

    This is a bit of interesting trend that’s emerged in two cases where I’ve taken a closer look at a defenceman who was being sheltered but put up better numbers than the people providing him shelter. In both cases, the guy being sheltered did better even if you just looked at star forwards and open play (Anton Stralman v. Ryan McDonagh was the other guy I looked at, although I didn’t write this part up).

    Six points of Corsi, which is the edge that Green has over the Carlson/Alzner pairing in open play against stars (it’s also very similar to the overall edge, for what it’s worth) is a pretty significant margin. It’s difficult for me to imagine circumstances in which a guy could be bad enough when he is defending to overcome that difference.

    Think of it in terms of 100 shots and assume that none of these guys affect on-ice S% (a fairly safe assumption). We’ll call the on-ice S% 8%. The Caps get outshot v. star forwards 52.3-47.7 with Green on the ice and 58.1 to 41.9 with Carlson/Alzner on the ice. At 8% shooting, that’s 3.816 goals on 47.7 shots when Green’s on the ice and 3.352 on 41.9 shots when Carlson/Alzner are out. Assume a league average save percentage behind Carlson/Alzner: .920. On 58.1 shots allowed, that’s 4.648 goals against, a difference -1.296 per 100 shots with star forwards on the ice.

    Even if you think that defencemen have an impact on opposition shooting percentage, in order to post a worse goal difference against star forwards, the save percentage with Green on the ice would need to be .902 or worse. There simply aren’t defencemen who do that over any extended period of time. I am exceedingly skeptical that, whatever deficiencies he might have, Green’s true talent on-ice save percentage is anywhere near that low.

    If you add it up, Green looks like he had the better season against the opposition’s best, although I haven’t taken into account linemates. It doesn’t really seem like it’s even that close though – I doubt the forwards that they’re paired with would make a significant difference. For what it’s worth, the three year on-ice save percentages for Mike Green (.920), Karl Alzner (.925) and John Carlson (.919) don’t really suggest a gap anywhere near big enough to negate Green’s big possession edge.

    I have to think that we’re heading towards a bit of a re-thinking of how we understand defencemen. This isn’t really stats guys v. traditionalists – guys within the game are saying the same thing. Here’s Darryl Sutter:

    “The game’s changed. They think there’s defending in today’s game. Nah, it’s how much you have the puck. Teams that play around in their own zone (say) they’re defending but they’re generally getting scored on or taking face-offs and they need a goalie to stand on his head if that’s the way they play.”

    I think he’s right. But if he is right, one of the logical implications is that we need to reconsider the entire idea of a defensive defenceman at 5v5. If a defensive liability like Mike Green outperforms Washington’s shutdown pair against star forwards, doesn’t that kind of suggest that maybe we’re wrong when we think that you need stay-at-home defencemen, guys with grit and defensive skill? If what guys like Green and Stralman do works, shouldn’t teams want to get guys like that?

    Email Tyler Dellow at tyler@mc79hockey.com


    25 Responses to Mike Green, Karl Alzner/John Carlson: Who Stops The Opposition’s Best?

    1. woodguy
      June 7, 2014 at

      Awesome, awesome stuff.

      I think the worm is turning on rugged, defensive Dmen.

      It’s just starting so the market inefficiency should last a while.

      Good puck moving D like Stralman who can move the puck the right way against the best but not put up points are probably the best $/result possible.

      Therefore MacT will sign Orpik and Matt Greene.

      Thanks for work Tyler.

    2. D'ohboy
      June 7, 2014 at

      This is great work, and a nice follow-up on your stats dump from earlier. I’ve felt that Green has been underutilized since Boudreau left, and that his underutilization under Hunter and Oates has negatively impacted the Caps’ possession stats (albeit at the margins, and not as significantly as the atrocious systems played under those two coaches).

      One caveat is that Green tended to play with good puck-movers when he was apart from Alzner. His most frequent partners were Dmitri Orlov and Nate Schmidt, both of whom had good CorsiRel ratings, even accounting for their time with Green.

      My initial guess as to the “why” behind the Corsi differential comes down to zone entries and exits. Green is a proverbial “one-man breakout.” Carlson is a decent puck-handler in space, but when pressured in the D-zone he’s more willing to chip it off the glass than Green. Alzner has regressed badly with the puck on his stick. He’s become far too willing to chip it off the glass or eat the puck along the boards. Carlson is generally willing to carry the puck into the offensive zone, but he’s less aggressive (and arguably less successful at it) compared to Green. Alzner looks to dump the puck as soon as he can.

    3. Hawerchuk
      June 7, 2014 at

      Remember when I had someone track passing for a DET-CHI game years ago? One of the funny things was that Brian Campbell’s possession stats were much better than Keith and Seabrook. He was clearly a better puck-mover, he broke out better, etc…One game does not a dataset make, but maybe you’re running into the same things…

    4. jonrmcleod
      June 7, 2014 at

      Interesting stuff.

      Have you revealed your math for your Open Play Corsi calculations?

    5. Cam Charron
      June 7, 2014 at

      Well, MY takeaway is that John Carlson is just damn good at making his centreman feel comfortable enough to win offensive zone faceoffs.

    6. Skraft
      June 7, 2014 at

      It’s always good when stats and eyes-on evidence mesh. Watching two really good possession teams work their way into and now play in the finals puts the ! on the stat. Question: when did the game shift to the importance of possession. My theory is that the blueline move was the point. Increasing the size of the O zone encouraged possession rather than a more open game. It took sometime for play to adjust as teams better at possession emerge as winners. Any data to support the theory?

    7. woodguy
      June 8, 2014 at

      If the straight corsi was close (45.3 vs 46.2), but the open play was not (41.9 vs 47.7), then should we even use corsi to compare players without game state info?

      Up until now when we have talked about context of CF% its been ozone starts, most common line mates, quality of competition, with and without you analysis, and even then its easy to make mistakes.

      Most common mistake I see in people using WOWY is failure to take into account a different assignment when “away” from a player. i.e. Dougie Hamilton having a better CF% away from Chara than with him doesn’t make Chara a drag on DH’s corsi as when he’s not with Chara his assignments are probably much easier.

      Now with what you have shown us here, even thoughtful and reasoned analysis using the available info still doesn’t show us enough to actually come to anything near a conclusion.

      We would have come up with the 45.3 vs 46.2, but only if someone took the time to pull out that CF% against the top end players. So maybe .01% of people who use this stuff to look at players, and still been off the mark.

      So what’s the next step for us arm chair hobbyists?

      Hope that someone reproduces your game state info and gives access to their data base?

      Hope that you give access to your database?

      I still think we can glean a pile of information and come to reasonably sound conclusions, but missing this by such a huge amount (0.9 vs 5.8 difference in CF%) is a bit disheartening.

      • Tyler Dellow
        June 9, 2014 at

        If the straight corsi was close (45.3 vs 46.2), but the open play was not (41.9 vs 47.7), then should we even use corsi to compare players without game state info?

        This is probably a bit of an unusual situation, with the combination of zone starts and offensive zone faceoff wins. I’m ok with using Corsi because these circumstances won’t usually arise. That said, if I was spending money on players, I’d be getting something a lot more detailed that includes all this stuff.

        Hope that someone reproduces your game state info and gives access to their data base?

        Hope that you give access to your database?

        I still think we can glean a pile of information and come to reasonably sound conclusions, but missing this by such a huge amount (0.9 vs 5.8 difference in CF%) is a bit disheartening.

        I’m not really set up for that sort of thing. Maybe something Extra Skater can get into for next year. It’s difficult to do it without it just turning into a blizzard of numbers though.

    8. Bruins fan
      June 8, 2014 at

      Tyler, really great work, I’ve learned a ton reading your stuff. Had a question about this work on defenseman. Shouldn’t defensive blocked shots be taking into account in some way in this study? I get why looking at corsi for matters, and why you wouldn’t want to take into account the opposition blocking shots, but is there a way to adjust corsi to take this into account? Not familiar enough to know if fenwick would be more appropriate? I’m probably just still recovering from how effective the Habs were at clogging up the shooting lanes. Thoughts?

    9. Mick Warshaw
      June 8, 2014 at

      “League average” sv% is .920? Check your links again, league average js more like .915 the last few years, .920 is GOAT territory. Hasek is the all time save % leader at .9202, a few ten thousandths of a point better than Lundqvist.

    10. Pingback: Spectors Hockey | NHL Blog Beat – June 9, 2014

    11. Tom Benjamin
      June 9, 2014 at

      Two points:

      1) These look to be ridiculously tiny sample sizes.

      2) Even if I ignore that objection and conclude that Green is better than Alzner against top opponents, it does not necessarily mean the Caps should use Green this way. When the Canucks had both Ohlund and Jovanovski, they used Ohlund as the shutdown guy and “sheltered” Jovo by letting him freelance with Naslund and Bertuzzi. That was the right call even if Jovo was better than Ohlund at stopping opposition stars.

      If Alzner is nearly as good as Green at the shutdown role, why not use Green in a role that he handles much better than Alzner? I probably want Green with Ovechkin and I probably don’t want Ovechkin out against top scorers…

      • Simon Lamarche
        June 9, 2014 at

        This is a great point I think. I was looking at the sample sizes and thinking this might prove to be a problem.

        However, smallish (I wouldn’t say ridiculously tiny) samples pointing in the same direction as the biggest ones (CorsiRel) are not the same as if those small samples were disproving what bigger ones tell us.

        As for player usage, I would agree with you and I would add that I wouldn’t expect a big swing in win expectancy were roles reversed either way. Until it is proven that one of them is better against stars or scrubs, they’re still top-4 defensemen who play approximately the same amount of 5 on 5 minutes every game. Giving more to the first line to beat on the second or playing better against the first to give more to the second line is kind of splitting hairs in terms of strategy.

        That’s incredibly useful info to a GM who might want a second Mike Green (Brian Campbell, Dustin Byfuglien, etc.?) and could trade either Alzner or Carlson to get one.

      • Tyler Dellow
        June 9, 2014 at

        1) Yeah, I agree. At the same time, I’m responding to a specific argument here: “Player X, a known defensively soft player who moves the puck plays easier minutes than Player Y, which is why his Corsi% is better.” We can argue about whether Green would continue to out-Corsi Alzner/Carlson; we can’t argue that, this year at least, the difference lies in facing the best players. As mentioned above, I’ve found this with Stralman/McDonagh too – McD plays more v. toughs, Stralman puts up better numbers.

        2) I’m not sure I agree that I don’t want Ovechkin out against top scorers. For all the ridiculous video and fun jokes, the advantage of playing guys who push possession against top scorers is that they get fewer chances because the puck’s in your end less. Caps did better against stars with Ovechkin out there than when he wasn’t.

        I agree with you that this isn’t an end point – we’re still left with a decision in terms of matchups for which we require more information. It may be (and I suspect it is) that it’s a game state specific thing. If I’m losing, I want to trade chances. Maybe then I don’t match my best players against their best – I want to up the chaos in the machine and hope my number comes up. If I’m winning the opposite is true. Easier to do at home than on the road.

    12. dcsportsfan85
      June 9, 2014 at

      Great article. I think Green is extremely under appreciated by Capitals fans because of some of his deficiencies in his defensive game. He definitely makes some terrible turnovers and mistakes in the defensive zone, but Adam Oates’ system on breakouts probably contributed to these issues to an extent by taking away Green’s best attribute in being able to carry the puck up ice to lead the rush.

      He was still given easier minutes and more favorable zone starts in total this past season, but he always has been given those minutes as are most elite offensive defensemen for example the revered Duncan Keith. John Carlson and Karl Alzner have both regressed in terms of their puck possession in their “tougher minutes” under Adam Oates. If the Caps can provide Green with a shutdown defenseman at a partner and unleash him from Oates’ system I think Green still has the talent to post great offensive numbers. He still lacks defensively, but if he can keep the puck away from the opponents and get a shutdown partner that sounds like a great top 4 pairing.

      I still believe Green is the Capitals best PP quarterback option on the point, even though Carlson showed great strides in handling those duties this past year.

      Completely agree with someone who said the today’s NHL emphasizes mobile puck-moving defensemen, but you do need some grit and physical presence on your blueline. Preferably from a 2 way defensemen, but those can be hard to find.

    13. Jay
      June 9, 2014 at

      Fantastic article, gotta love debunking commonly held beliefs.

      What made you include Doan in the list of 23 star forwards? Thought that was a bit odd.

    14. chelch
      June 12, 2014 at

      There’s a place for big, sturdy defensive dmen who chew up minutes as long as there’s big, sturdy forwards who like to pound dmen. You have to have the Matt Greene’s and the Willie Mitchell’s to protect your Drew Doughty’s and Mike Green’s from the Milan Lucic’s and the …well there’s only one Milan Lucic… but there are a lot more hitters than scorers in the league, so you need guys that can take a hit and get the puck up the boards. There are also way fewer quality puck moving dmen so it’s a big risk to put 4 in your lineup and lose a couple to injury. 2 or 3 and a shutdown pair has always been the right mix.

      You’re right in pointing out that the Caps have misplaced Carlson as their number 1. Same situation as Phaneuf. But if Green is consistently getting injured, and their lack of depth on defense is epic (with only 3 dmen playing more than 54 games and a total of FOURTEEN different dmen playing at least one game), it’s hard to blame them for sheltering their star.

    15. Pingback: WHERE ARE THE TRADES: Official Off Season Trade Thread :D - SportsHoopla.com Sports Forums

    16. Anun
      June 27, 2014 at

      Why are the graphs so skewed? You used different year values for the Green corsi, which makes it seem like he’s been a consistently good possession player, it’s completely misleading. Comparing the years where they have played together on the same team you can see that zone starts are largely correlated with corsi (for Carlson and Green at least, I didn’t check Alzner).

    17. What?
      July 10, 2014 at

      How can you say that their line mates don’t matter. 3/5 of the players on the ice at any given are forwards. Maybe green played more with good Corsi forwards like Grabovski against star players in this small sample size and Carlson and Alzner played with poor Corsi forwards like Laich. Oh and Green’s on ice save percentage last season against all competition was .907, second worst on the team, so it’s definitely plausible his on ice save% against stars was lower.

    18. Pingback: Should the Sharks Trade for Mike Green

    Leave a Reply

    Your email address will not be published. Required fields are marked *