214 replies to this topic

[squishyx]

Actually if a person knew how to interpret microstats (and indeed had the ability to think stochastically), they wouldn't have any idea who won based on Corsi.  But were they to make that judgment on Corsi alone, which is beyond moronic since who does that, they'd probably think that the Leafs won, but got incredible goaltending to do so, because it's rather rare that a team gets outattempted by that much while losing the game.  Usually teams get outattempted like that while leading in the 3rd period, but that's not really how this game went.

YOU. you did that. This very thread was you creating a topic to judge a player and all you opened with was his terrible corsi. Maybe you went on later to justify your argument but the opening post was essentially "Stephen Gionta is killing this team, and the reason is his corsi".
 
yikes.

It's worth nothing, just like plus/minus is worth over anything less than an enormous sample (and even then requires giant caveats to interpret properly)

I agree for the most part. But I see tons of people making arguments based on corsi, zone starts or fenwick when they show what the poster is trying to prove. It's worth highlighting when those numbers are the complete opposite of reality as well. Curious though, do you really think 15 games qualifies as an "enormous sample"?

[squishyx]

I don't think Corsi numbers are that valuable unless they are used in a decent sample size. Over a period of a time a team with a high Corsi% should win more often then lose but there are going to be games a team drops despite being better in possession. There are going to be times when players get beat in possession in a single game but are fortunate enough to convert their few offensive chances. Even though tonight Zidlicky +2 and Carter -1 were on the low end of Corsi for Devils players, they still had nearly even or plus numbers.
 
Corsi is valuable as a tool for long term analysis/trends. I'm not saying its the end all be all but I think it has its place in hockey. Despite losing tonight, if the Devils play a ton of games like this in terms of possession, then they should win the majority of them. Hitting 2 posts and Elias being robbed by Reimer completely changed the game along with Moose's poor goaltending.
 
Gionta was on for both Devils goals but did he really do that much to contribute to them? Coming into tonight Gionta's PDO was tied for 14th highest in the league and he benefited from the players around him again tonight. At some point it's due to regress as history shows.
 
I hope Josefson regains his form in Albany. I think you can make the arguement that Carter and Bernier would be even more effective playing on a line together away from Gionta. Both of them have a higher Corsi For% without Gionta, Carter (55.9) and Bernier (55.6) then with, Carter (41.4) and Bernier (45.2).

We're all probably debating this issue too much but I find it fun nonetheless.

I think it's fair to say you need a large sample size for any stats to mean anything. I don't think we are there yet, and we probably won't be for a 48 game season. My argument all along has never been that Gio is some diamond in the rough, but rather an average 4th liner who was unfairly targeted as a scapegoat when the team was doing pretty well.

It's a lot different now, I'd be fine with anything to shake up the team. I still don't think Gio is the problem but if it meant getting some fresh faces up here for whatever reason I wouldn't be all that bothered. That's the bad with being a 4th liner, I don't have high expectations for them, but they are also (rightfully so) the first to get shifted down.

[Pepperkorn]

Cool.
 

 

Actually if a person knew how to interpret microstats (and indeed had the ability to think stochastically), they wouldn't have any idea who won based on Corsi.  But were they to make that judgment on Corsi alone, which is beyond moronic since who does that, they'd probably think that the Leafs won, but got incredible goaltending to do so, because it's rather rare that a team gets outattempted by that much while losing the game.  Usually teams get outattempted like that while leading in the 3rd period, but that's not really how this game went.
 

 

It's worth nothing, just like plus/minus is worth over anything less than an enormous sample (and even then requires giant caveats to interpret properly)

Tri - can you be a little more of a teacher here rather than a bully?  Sure people dont want to do the legwork to communicate stochastically (holy f**k can you BE anymore supercilious and alienating?)

 

But people do want to know what you're getting at.  If you don't want them to, then save it for your blog, right? But if you DO - then practice putting things in laymans terms.  Doesn't that kind of seem less stressful for everyone?  If squishy is just here to fight he'll make himself look stupid all on his own.  If you try to rush the process you look equally as inept.  Just sayin'  (if you get your pedantry i get my colloquialisms)

Edited by Pepperkorn, 05 March 2013 - 09:59 AM.

[Pepperkorn]

I think it's fair to say you need a large sample size for any stats to mean anything. I don't think we are there yet, and we probably won't be for a 48 game season. My argument all along has never been that Gio is some diamond in the rough, but rather an average 4th liner who was unfairly targeted as a scapegoat when the team was doing pretty well.

It's a lot different now, I'd be fine with anything to shake up the team. I still don't think Gio is the problem but if it meant getting some fresh faces up here for whatever reason I wouldn't be all that bothered. That's the bad with being a 4th liner, I don't have high expectations for them, but they are also (rightfully so) the first to get shifted down.

Squish -- I know, I know.... I said the exact same thing...   but actually you don't.  I knwo there is something wrong with my logic -- but science depends on getting reliable results with some pretty minute data.  There are proven calculations that do make predictions on small amounts of data accurate.  That's why I'm thinking maybe it's the many variables in hockey that make it difficult to make lineup choices based on microstats alone.  OH -- I JUMPED THE GUN and didnt' read your whole post.    alas I refuse to edit out my stupidity.

 

Exactly how much weight do we want to give microstats?  Feisty?  I think that might be a question I need to start off asking.

Edited by Pepperkorn, 05 March 2013 - 10:09 AM.

[squishyx]

Squish -- I know, I know.... I said the exact same thing...   but actually you don't.  I knwo there is something wrong with my logic -- but science depends on getting reliable results with some pretty minute data.  There are proven calculations that do make predictions on small amounts of data accurate.  That's why I'm thinking maybe it's the many variables in hockey that make it difficult to make lineup choices based on microstats alone.  OH -- I JUMPED THE GUN and didnt' read your whole post.    alas I refuse to edit out my stupidity.
 
Exactly how much weight do we want to give microstats?  Feisty?  I think that might be a question I need to start off asking.

I've always felt that stats are just indicators for how a player or team has been playing, not on how they will (obvious statement much). I'm not sure there is value in putting weights to microstats, my gut tells me they are somewhere on the order of 25-50% better at "predicting" future results then regular stats, but that still isn't enough for me to ever look at 'any' given stat after 15 games and make a decision, especially about a 4th line player.

My hope is that any team has a small army of nerds that actually break down film after a game to analyze every play and take them all into context before making recommendations to the coaching staff.

[Pepperkorn]

I know -- my guess: first stat Lou calculates =  RoI on a stats guy's salary.  I admit I don't feel comfortable saying that's short-sighted of the organization.  I'm a homer about it.  That's why I think I need to make myself part of the discussion. I don't want to be close-minded.

 

is there a stat for backchecking?  what we want to do is find a stat for foresight.  Is there one?  Everyone hates +/- on the board - I've always kind of liked it, probably more as a motivator for players though.  It's me making a hard stat touchy feely for a player  - not providing real data upon which to base organizational choices. I'm tryign to get out of that mind-set to broaden my sight.

 

I honestly havent considered hockey microstats enough -- or... well at ALL really.  So i ought to give it a crack - the worst is I can better articulate why it is I'm reticent about their application at this point. 

Edited by Pepperkorn, 05 March 2013 - 11:52 AM.

[ATLL765]

I don't think Corsi numbers are that valuable unless they are used in a decent sample size. Over a period of a time a team with a high Corsi% should win more often then lose but there are going to be games a team drops despite being better in possession. There are going to be times when players get beat in possession in a single game but are fortunate enough to convert their few offensive chances. Even though tonight Zidlicky +2 and Carter -1 were on the low end of Corsi for Devils players, they still had nearly even or plus numbers.

 

Corsi is valuable as a tool for long term analysis/trends. I'm not saying its the end all be all but I think it has its place in hockey. Despite losing tonight, if the Devils play a ton of games like this in terms of possession, then they should win the majority of them. Hitting 2 posts and Elias being robbed by Reimer completely changed the game along with Moose's poor goaltending.

 

Gionta was on for both Devils goals but did he really do that much to contribute to them? Coming into tonight Gionta's PDO was tied for 14th highest in the league and he benefited from the players around him again tonight. At some point it's due to regress as history shows.

 

I hope Josefson regains his form in Albany. I think you can make the arguement that Carter and Bernier would be even more effective playing on a line together away from Gionta. Both of them have a higher Corsi For% without Gionta, Carter (55.9) and Bernier (55.6) then with, Carter (41.4) and Bernier (45.2).

We're all probably debating this issue too much but I find it fun nonetheless.

 

I don't think swapping Gio for JJ would be a bad idea. I know we don't like giving a young guy like JJ crap minutes on the 4th line, but we can squeeze him in on the PK and PP to get him minutes.

[Triumph]

Tri - can you be a little more of a teacher here rather than a bully?  Sure people dont want to do the legwork to communicate stochastically (holy f**k can you BE anymore supercilious and alienating?)

 

But people do want to know what you're getting at.  If you don't want them to, then save it for your blog, right? But if you DO - then practice putting things in laymans terms.  Doesn't that kind of seem less stressful for everyone?  If squishy is just here to fight he'll make himself look stupid all on his own.  If you try to rush the process you look equally as inept.  Just sayin'  (if you get your pedantry i get my colloquialisms)

 

If I thought Squishy hadn't seen these arguments before, I'd make them without being a bully (see also how I responded to that guy who coaches youth hockey).  But he has 7000 posts and has been here for a long time.   And again, if I thought Squishy was doing anything besides creating silly strawman arguments, I'd also respond thoughtfully, as I have done to many others who are much more skeptical than I of these statistics.  

 

As for the word stochastically, I don't like it either, but it's hard to write out something like 'is aware of the extreme randomness observed in small samples and the large role randomness plays in the game of hockey even in larger samples'.

 

YOU. you did that. This very thread was you creating a topic to judge a player and all you opened with was his terrible corsi. Maybe you went on later to justify your argument but the opening post was essentially "Stephen Gionta is killing this team, and the reason is his corsi".

 

 

Did I say that Gionta had a bad game Corsi wise and therefore he is bad?   Absolutely not.  Over at In Lou We Trust, the guy who writes the game recaps will often say things like Player X had a good Corsi thus he had a good game and I've often admonished him not to do that.  Single game Corsi doesn't tell much of a story.  It tells you where the puck was 5 on 5, but in a very limited way - it doesn't tell you WHY it was where it was.

 

So, no, I didn't lay out the case perfectly in my opening post.  We can't change the past.

 

I look at 4 main stats, which I already explained to NLInfante when I look at a player who I don't know that well:

 

Fenwick or Corsi - Where's the puck, generally, when this player is on the ice?  

Zone Start - How often is his coach starting him in the offensive or defensive zone relative to one another?

Corsi Rel Qualcomp or Corsi Qualcomp - Who is he playing against?  Is he playing against the other team's top territorial players generally?

Corsi Rel Qualteam or Corsi Qualteam - Who is he playing with?  Do his teammates suck?

 

I will also look at ice time (since this is a good indicator of how a coach feels about a player), shots on goal, goals, and assists, but it's rare that some guy will pop up out of nowhere with good offensive numbers who I haven't heard of.

 

Now there's issues with all of these which we can get into.  But right now, Gionta ranks middle of the pack in team in terms of Corsi Rel Qualcomp, which sounds about right - he faces top lines sometimes, fourth lines sometimes - the CBGB line is tops among NJ forwards in Corsi QoC which suggests they face the toughest competition, although it wouldn't surprise me if to a degree they've brought that on themselves.  They're doing better in terms of Corsi and Corsi Rel but they are still at the bottom among forwards with 10+ games.  So it is trending upwards, though I suspect that is in part because they are less frequently playing against top lines since Carter got hurt and Gionta got moved down for a couple games.  Gionta himself starts in the offensive zone more frequently than most of the Devils' forwards.

 

Right now they're being pushed back a tolerable amount, although I think this has more to do with the fact that the Devils are frequently now trailing in games instead of leading as they were at the beginning of the season.  They're also playing Sidney Crosby a lot less.

[Pepperkorn]

Thanks Tri - this is really helpful.  I want to go back and read more.  The NLinfant ref is ... a good way for me to crash the server using the search feature 

 

and I do think I make your posts more supercilious than they are even.  I have to practice reading words without putting my own attitudes on top of them.

 

(I dont' sound like Chico now do I?  Well ya gotta give those Leafs credit - they played a heck of a game and maybe the Devils outplayed them but the scoreboard tells a diff... SHUT UP CHICO! )

[Devilsrock01]

enough with this benching bullsh!t..

[squishyx]

If I thought Squishy hadn't seen these arguments before, I'd make them without being a bully (see also how I responded to that guy who coaches youth hockey).  But he has 7000 posts and has been here for a long time.   And again, if I thought Squishy was doing anything besides creating silly strawman arguments, I'd also respond thoughtfully, as I have done to many others who are much more skeptical than I of these statistics.

Fwiw I am not skeptical about microstats and I don't really care how you respond, you've always been robotic and I never take it personally.

I was not making strawman arguments. If I were I would have harped on more about my Richards examples (and even there I hedged for you), or pulled more more data from previous years finding the few examples that went against the flow of your argument. There will always be statistical anomalies which is why I didn't bothered to respond with a list of dozens of players who every playoff year completely reverse course from regular or micro stats they had from the regular season.

Instead I tried to focus on something more comparable to what you were suggesting that we do right now with Gio, using these microstat's to determine who we should play or not. And I think it's perfectly legitimate to point out that microstat's would have suggested benching 2/3rds of our 4th line last year before the start of the playoffs. What's so strawman about that?

[DH26]

I think Tri's better at explaining it than a lot of other people who are behind the corsi and fenwick and other advanced stats. A lot of them who harp them aren't as in tune with them. The problem, as has been said, is getting past the robotic kind of tone but the substance for them w/ him saying it is a lot better than most other people who tend to bully much more than him regarding these stats. He knows what he's talking about, a lot of others just parrot the stats or don't fully understand them. 

 

I dunno that much myself and I saw Cordell put on Twitter that Larsson had 37% offensive zone starts but 47% of his shifts ended in the offensive zone. Is that a good number or just a function of who he's playing with/against or some kind of combination? 

[ATLL765]

I think Tri's better at explaining it than a lot of other people who are behind the corsi and fenwick and other advanced stats. A lot of them who harp them aren't as in tune with them. The problem, as has been said, is getting past the robotic kind of tone but the substance for them w/ him saying it is a lot better than most other people who tend to bully much more than him regarding these stats. He knows what he's talking about, a lot of others just parrot the stats or don't fully understand them. 

 

I dunno that much myself and I saw Cordell put on Twitter that Larsson had 37% offensive zone starts but 47% of his shifts ended in the offensive zone. Is that a good number or just a function of who he's playing with/against or some kind of combination? 

It means he's advancing the puck, or at least those on the ice with him are. The way Larsson passes though, this is where he's going need to excel, right? I'm no statistician, but if his being elite comes from his passing ability(which is what everyone said was his skill, the breakout pass) this should be where he be doing better than others. Larsson, to the eye test, made strides from last year to game 1 this season for him to now and if I got the gist of this thread at all, it's that when the microstats back up your eye, it's likely that's a good indicator of a particular skater's play, if they don't seem to indicate the same things, well, then you might just have one that's hard to figure.

Edited by ATLL765, 05 March 2013 - 05:03 PM.

[thefiestygoat]

I dunno that much myself and I saw Cordell put on Twitter that Larsson had 37% offensive zone starts but 47% of his shifts ended in the offensive zone. Is that a good number or just a function of who he's playing with/against or some kind of combination?

I swear I tweeted something like that a few days ago. I'd say that's really impressive, especially given his age. He has the lowest Offensive Zone Start% of all d-men in the NHL with at least 15+ GP. I think it shows the PDB really trust him in tough situations. Having Greene by his side has helped but I think Larsson has pulled his fair share.

[SterioDesign]

I swear I tweeted something like that a few days ago. I'd say that's really impressive, especially given his age. He has the lowest Offensive Zone Start% of all d-men in the NHL with at least 15+ GP. I think it shows the PDB really trust him in tough situations. Having Greene by his side has helped but I think Larsson has pulled his fair share.

 

at this point i think it's not that Deboer trust him because he's so great, i think its more that he doesnt trust the other dmen that we have... with good reasons.

[Triumph]

I think Tri's better at explaining it than a lot of other people who are behind the corsi and fenwick and other advanced stats. A lot of them who harp them aren't as in tune with them. The problem, as has been said, is getting past the robotic kind of tone but the substance for them w/ him saying it is a lot better than most other people who tend to bully much more than him regarding these stats. He knows what he's talking about, a lot of others just parrot the stats or don't fully understand them. 

 

I dunno that much myself and I saw Cordell put on Twitter that Larsson had 37% offensive zone starts but 47% of his shifts ended in the offensive zone. Is that a good number or just a function of who he's playing with/against or some kind of combination? 

 

That is a good number, but it needs a lot of caveats - I wish I had bookmarked all the articles on this - but basically it's been found that players who start a lot in the offensive zone finish a lot in the defensive zone and vice versa.  For instance, the Sedins started nearly 80% of their shifts that began in either the O-Zone or D-zone in the offensive zone, but only finished 58% of their shifts in the offensive zone.  That might seem like they went backwards, but on the contrary, they were a huge + in terms of shots and goals.

 

My problem with Zone Finish is that it relies too much on the coach's discretion and how he deploys his lines against the other coach.  Also, it doesn't give credit for scoring a goal - since the ensuing faceoff is in the neutral zone, that doesn't count as an offensive zone finish.  The real number to take away from Larsson is that while he's starting 5 out of 8 faceoffs that take place in either the O-zone or D-zone in the D zone, he's break-even in Corsi.  That's a really good rate.

fiesty:  Wow, when I saw that number I said 'boy, fiesty's out to lunch, no way that's the lowest'.  But it is.  Weird.  I guess it's hard to get a D man's number lower than that because there's only 3 D pairings.

 

at this point i think it's not that Deboer trust him because he's so great, i think its more that he doesnt trust the other dmen that we have... with good reasons.

 

The Devils' defense is giving up fewer shots than all but 2 teams 5 on 5.  They've had 3 straight road games where they've held the opponent to 25 or fewer shots overall.

Edited by Triumph, 05 March 2013 - 05:20 PM.

[Devils Dose]

I think you can make the arguement that Carter and Bernier would be even more effective playing on a line together away from Gionta. Both of them have a higher Corsi For% without Gionta, Carter (55.9) and Bernier (55.6) then with, Carter (41.4) and Bernier (45.2).

hmmmm... Well we got to see this tonight. It seemed kind of true.

[maxpower]

Bumping this back from the dead.

 

Anyone taking a look at his faceoff stats for this season?   This has to be historically bad for the number of them he's taking.   I don't know if he should be benched, but he sure as hell shouldn't be taking any faceoffs.

[SterioDesign]

Bumping this back from the dead.
 
Anyone taking a look at his faceoff stats for this season?   This has to be historically bad for the number of them he's taking.   I don't know if he should be benched, but he sure as hell shouldn't be taking any faceoffs.

Well whos fault is that? Really? Gionta is certainly busting his ass out there every shift, hes just not that good. Is it his fault? No. Lou didnt get anyone better than him and deboer is playing him, Gio certainly never had a gun to their heads to put him out there.

If hes out there in those situations its because we dont have anyone better. Were also sitting on like 10+million. Lou should try to get us a hockey player who can be useful in a hockey play. And stop signing and picking up goons who cant bring anything to the table.

[maxpower]

Well whos fault is that? Really? Gionta is certainly busting his ass out there every shift, hes just not that good. Is it his fault? No. Lou didnt get anyone better than him and deboer is playing him, Gio certainly never had a gun to their heads to put him out there.

If hes out there in those situations its because we dont have anyone better. Were also sitting on like 10+million. Lou should try to get us a hockey player who can be useful in a hockey play. And stop signing and picking up goons who cant bring anything to the table.

 

Huh?  I'm just saying he's 31% this year on faceoffs, which is hilariously awful.  One of his regular linemates is much better on faceoffs than that.   It doesn't factor into whether he should be playing or not... just why on god's green earth is he taking any faceoffs?    He's 83 for 263!