Player Discussion Adam Larsson (Hall talk will result in a thread ban)

Discussion in 'Edmonton Oilers' started by DasDrai, Jan 29, 2018.

  1. Little Fury

    Little Fury Registered User

    Joined:
    Jun 21, 2006
    Messages:
    8,852
    Likes Received:
    1,082
    Trophy Points:
    139
    So let me get this straight: you hate it when people use stats like shot attempts for/against without context (despite using stats like +/-without context) but you like David Staples' utterly garbage and completely subjective "mistakes" stat? *EYEROLL*
     
    Aerrol likes this.
  2. MessierII

    MessierII Registered User

    Joined:
    Aug 10, 2011
    Messages:
    13,844
    Likes Received:
    770
    Trophy Points:
    139
    It corresponds with my eye test. Klefbom is and always has been bad in the D zone. Turns the puck over and blows coverage regularly.
     
  3. Little Fury

    Little Fury Registered User

    Joined:
    Jun 21, 2006
    Messages:
    8,852
    Likes Received:
    1,082
    Trophy Points:
    139
    "It confirms my preexisting biases" is a lousy criteria for evaluating a stat.

    Funny too how you don't account for quality of competition (Klef is seeing tougher opponents than the likes of Russell and Gryba who fare better by this bogus metric) or zone starts or any other contextual considerations. But the bottom line is "mistakes" is an entirely subjective and incredibly loosely defined, unlike neutral and actually quantifiable stats like shot attempts.
     
    Aerrol likes this.
  4. MessierII

    MessierII Registered User

    Joined:
    Aug 10, 2011
    Messages:
    13,844
    Likes Received:
    770
    Trophy Points:
    139
    Shot quality > shot quantity

    Shot attempts have literally 0 impact on the outcome of the game. I’ll take subjective stat quantifying scoring chances for and against over 100% accurate shot attempt stats 10 times out 10.
     
    guymez and CycloneSweep like this.
  5. Little Fury

    Little Fury Registered User

    Joined:
    Jun 21, 2006
    Messages:
    8,852
    Likes Received:
    1,082
    Trophy Points:
    139
    Then use a stat like XGF/XGA instead of this staples nonsense.

    No that's GF. But shot attempts are a useful proxy for puck possession.

    Yeah it's pretty clear you're less interested in using stats as a tool to evaluate players and more interested in cherry picking anything that confirms your pre-existing beliefs.
     
  6. MessierII

    MessierII Registered User

    Joined:
    Aug 10, 2011
    Messages:
    13,844
    Likes Received:
    770
    Trophy Points:
    139
    Staples got the idea from Dave Tippett. He used a similar stat to track and evaluate his own players.
    Shot attempts are not puck possession. If you want to measure puck possession measure puck possession. Time of possession is tracked in other sports. Pretty simple.
    No different than literally every analytics blogger in existence and 99.99% of aceboogies posts.

    Sorry for your hurting your feelings by the way.
     
    Last edited: Jun 1, 2018
  7. Drivesaitl

    Drivesaitl Civilized Worm

    Joined:
    Oct 8, 2017
    Messages:
    4,595
    Likes Received:
    1,841
    Trophy Points:
    99
    Gender:
    Male
    Location:
    Edmonton
    A key is theres a difference between the hockey analytical level of Staples vs Tippet. With Staples its garbage in/garbage out. His hockey observations are as fraught with bias as anybody.

    Its one thing for someone to borrow a concept or metric. But to think they have done that concept justice through their own level of analysis and observation is another thing.

    The judging of shot quality would be very subject to inter viewer reliability because unlike goals, or shots, the judging of quality shots is not an actual data amount. Its a subjective judged tally. Of course if a viewer likes a particular player, or think they are dangerous around the net the "shot quality" bell will be ringing like crazy. Doesn't mean the tabulation is quality..

    If you are familiar with statistics in order for observed data to be considered reliable inter observer reliability should be substantiated. Different viewers should be able to utilize the same metric to obtain consistently reliable data.

    What has often occurred instead in these metrics (Staples and that of analytics) is that interobserver reliability has been said to be satisfied through the use of several different viewers involved in tabulating. This does not establish reliability unless all those viewers tabulations demonstrate very similar scoring.

    This is one of the difficulties I have with a lot of hockey stats. Those that are producing them have limited familiarity with statistical methodology or such things as reliability. How many would even know how statistical reliability is obtained;

    Social Research Methods - Knowledge Base - Types of Reliability
     
    Last edited: Jun 2, 2018
  8. Little Fury

    Little Fury Registered User

    Joined:
    Jun 21, 2006
    Messages:
    8,852
    Likes Received:
    1,082
    Trophy Points:
    139
    So? It doesn't change the fact that it's a fundamentally subjective stat (also I expect that Tippet knows what to look for mor ethan staples does).

    Until such a time as the NHL starts tracking that and making it public or someone takes the time to do so by watching all the games, shot attempts are a useful proxy for possession. Shot quality would be even harder to measure as there are a million variables that influence it.q

    Whataboutism.

    Is this english?
     
  9. PerformanceMcOil

    PerformanceMcOil Registered User

    Joined:
    Aug 2, 2005
    Messages:
    360
    Likes Received:
    27
    Trophy Points:
    81
    Yeah, but it isn't fair to single out Staples in that regards. There are problems with how basically all the advanced stats in hockey are generated, and there are almost never measures of variability done for any of them. So while the numbers generated themselves are technically objective, the conclusions made on them are so open to observer bias, they might as well not be.

    The analyses done basically boil down to 'more is better', and at what level 'kicking ass' turns into 'getting ass kicked' tends to revolve around the narrative someone is trying to push on a given player.

    Staple's stats, and the scoring chance stats that others generate are very useful in principle, but yeah, unless they are done with rigor, they're just different chapters in the advanced stats fiction novel. An interesting read, but not scientific in the slightest.
     
    PGW likes this.
  10. PerformanceMcOil

    PerformanceMcOil Registered User

    Joined:
    Aug 2, 2005
    Messages:
    360
    Likes Received:
    27
    Trophy Points:
    81
    The problem is, you can't just say that you are unable to measure what you'd like to measure, so you'll just measure something else and use it as a 'proxy', and because it is a number, it is good. You have to show it is actually a useful proxy, and then in your analyses bake in how good it is.

    For instance, let's say at the level of the individual, Corsi shows a 60% correlation with possession, which would actually be incredibly high, and I very much doubt it is anywhere near that. Let's also assume that two player's have CF/CA percentages of 58% vs 48%, and that these players are machines and their CF/CA are exactly that, game after game and season after season, so we don't have to worry about their individual variance (which I hope you can understand is a ridiculous assumption). Then let's really go nuts, and assume they play on the same teams with and against the same players for the same amount of time (which is impossible). Even then, their true 'puck possession' values would range from about 35 to 81 for the first player, and 28 to 68 for the second player*. So, player one is more likely to be a better puck possession player for sure, but there is a substantial chance that player two is actually equal or superior. Yet, I am pretty sure no one who uses these metrics would have anything nice to say about player two. And this was after making a boat-load of assumptions that don't even come close to reflecting reality.

    So yeah, Corsi etc. may be proxies for possession, and they may be the best stats we have, and the numbers are numbers so they may be objective, but it doesn't mean the conclusions drawn have any statistical foundation to stand on.

    *NB: These numbers are generated purely from the R^2 of 60%, and aren't a true measure of what the ranges would be (you would need to have a significant number of data points measuring possession vs. CF/CA for the league). However, it is a reasonable estimate of magnitude, and there would certainly be significant overlap between what the true possession numbers of the two players could be.

    Edit: Also, a pretty strong argument could be made that if all the assumptions I made were true, the r^2 value between Corsi and possession would be much higher, and so the predicted values would be much tighter.
     
    Last edited: Jun 2, 2018
  11. McFuhryous

    McFuhryous Registered User

    Joined:
    Sep 24, 2007
    Messages:
    5,176
    Likes Received:
    1,060
    Trophy Points:
    139
    Location:
    Medicine Hat, AB
    I'm pretty excited to see what kind of changes Yawney brings to Larsson's game. I think we will see him turn Larsson into a complete d-man. It won't happen over night, but by 25 games in I bet we see the positive changes start to take affect.
     
    ManitobaKid and MessierII like this.
  12. ohheyhemsky

    ohheyhemsky Regehr DooDoo

    Joined:
    Nov 1, 2010
    Messages:
    19,903
    Likes Received:
    594
    Trophy Points:
    139
    Occupation:
    I attacc
    Location:
    DT Cowtown
    Legit, what if we see a 30 point season? Will everyone just shush?

    Probs not.
     
    McFuhryous likes this.
  13. Dazed and Confused

    Dazed and Confused Ludicrous speed, GO!

    Joined:
    Aug 10, 2007
    Messages:
    3,819
    Likes Received:
    291
    Trophy Points:
    124
    Location:
    Gothenburg. Sweden
    I'd be curious to see Larsson get some PP time. Let him grow his offensive game a bit, and for the most part he fits the needs of the top unit: RH shot, good passer, and not someone that telegraphs their moves. He's always looked good when given the opportunity with the national team, so try it here.

    But this would mean you'd need to free up Larsson from his heavy defensive usage.

    If you do that I'd like the idea of going after Pysyk or Petrovic from Florida. Bogosian Buffalo would be another name (if they'd be willing to retain salary), who would also work.


    Nurse-Larsson
    Sekera-Klefbom
    Russell-Pysyk/Bogosian/Petrovic
    Benning

    Maybe break up Nurse-Larsson if you want to spread the physicality around, but I actually like the look of that.
     

Share This Page

monitoring_string = "358c248ada348a047a4b9bb27a146148"