I don't want to open this up to the main board, since discussions get out of whack there, but keeping this Islanders-focused, I simply don't get the fascination with advanced stats. Seems to be that people (the media, some posters) are simply throwing out numbers as science without enough scrutiny of whether they are even valid. It's pretending to be scientific without any scientific method or thought.
Lets hammer out a few predicate issues. No one is pretending to be scientific, although disregarding stats is significantly LESS scientific. It is an issue of credibility. Either you can support your assertions, or you cannot. If someone makes an assertion (Travis Hamonic sucks) and their only support is (Didn't you watch him? He sucks!) conclusory, then that person has no credibility. Supporting an assertion is what I'm interested in.
Next issue is that I don't think these stats are advanced in any way. They are nothing like WAR, or VORP, or any of the ACTUAL advanced stats that people use in baseball. None of them are any more advanced than +/-, however they are infinitely more useful.
I am by no means an expert on advanced stats. I read the reports, visited the websites, heard from the Mirtles of the world, understand the discussion from both sides.
What I haven't heard from ANYONE is an example of how you would use it to make a roster decision and let's use the
NYI as an example.
I think I've said QUITE a bit on this topic, even building a predictive model and agreeing to input any lineup anyone has asked for... However, putting that aside, I'll take the bait and continue on. I will re-answer the first inquiry in a bit here.0
My hypothesis is that these "advanced stats" are hardly advanced.
Wholly agreed. Nothing advanced at all. Really quite simple stuff.
The results, sorted top to bottom, are more random than insightful. I've sorted the NYI players (40games plus) top to bottom by Corsi Relative. At the top, deHaan (who was outstanding last year), then Brock Nelson, Matt Donovan and Josh Bailey. Hardly the "top four" of anything on this team, by EVERYONE's opinion. Our best players (by EVERYONE's opinion, Tavares, Okposo, Hamonic, Frans are RIGHT IN THE MIDDLE).
First off, you're starting from the assumption that you already know the answer. That is a problem but one we can ignore for now.
Second, Corsi relative merely tells us one simple thing: How does the TEAM perform, in terms of shots towards the other net compared to shots towards its own, when this guy is on the ice. There are countless factors that go into this inquiry. Most importantly are, whether the guy is playing against tougher competition, whether he was sheltered by playing with the best teammates, whether he started in the OZ or the DZ, how many minutes he played, and others.
If you rely on relative Corsi alone, you'll get a good picture of how a guy does relative to his similarly situated teammates. That picture always has caveats attached, but generally it will tell you something valuable, such as whether a guy was performing well at keeping the puck in the offensive zone.
And lets consider the guys you are discussing:
Josh Bailey - everyone's complaint is that he doesn't shoot enough. So what does that tell us, he's in the offensive zone PASSING too much. THIS is entirely in line with his good corsi numbers. The goals come because he doesn't shoot enough. This is a guy who lines up entirely with the anti-stats crowd, but they don't realize the implications.
Matt Donovan - everyone's complaint is that he gives up bad turnovers (on bad pinches) that turn into goals. This tells us that he is IN the offensive zone moving the puck around and he gets hung up and makes a mistake that turns into ONE corsi event but also turns into a goal. This is a complicated issue that I've discussed at length, and I won't belabor it yet again, but the above comports entirely with his corsi numbers. Corsi isn't telling us whether someone gets a "bellringer" chance. It tells whether more shots go to the opposing net while he's on the ice. That is true of Donovan. If those handful of bad turnovers are saves, no one is complaining about Matt Donovan.
Brock Nelson is an outstander player and so is de Haan. With regular minutes on the top line Brock Nelson could very well be a star on this team. We can discuss his play further at another time, if you like.
The Islanders have some strangely skewed numbers because one of the worst players in the NHL had a lot of minutes on our team: Andrew MacDonald. He dragged down some of our best players because they played with him so much. We can get into that deeper point later.
This seems to be the same random (if not random, then certainly showing no actionable pattern) list regardless of what metric you choose (Fenwick, PDO, etc.)
The problem is that you want ONE metric to tell you all the answers. It doesn't work that way. We use them all in conjunction to get a good picture of what has happened on the ice and how we know who is letting us down and who is bringing us up. Corsi is the most valuable of any of the individual metrics.
So, can someone please explain what insights Corsi (or advanced stats in general) we can gleam based on last year's NYI season?
Where do I begin:
1) Matt Donovan is a good possession player who made some bad defensive mistakes that were made glaring by bad goaltending.
2) Andrew MacDonald was a complete disaster who wasted more of JT's time than any other player to step out on the ice with him. Anytime MacDonald was on the ice with JT, shots generated goes down, and shots allowed goes up. The opposite is also true.
3) Travis Hamonic is playing some of the most difficult minutes in the NHL - and still excelling.
4) Brian Strait is a possession disaster.
5) Matt Carkner is not as bad as we all think (but still bad)
6) de Haan is an absolute stud (and Matt Donovan is potentially a COMPARABLE player if he can solve some issues)
7) Vis is STILL an excellent possession defenseman (when he can play)
8) Casey Cizikas and Matt Martin were horrendous together, coming out on the ice after every goal allowed and getting the team outshot almost every time.
9) Michael Grabner is one of the best PKers in the league, suppressing more corsi events during shorthanded time than anyone else in the NHL.
10) Jack Capuano is a moron because he did not use Michael Grabner in first PK situations.
I could go on but I think I've given you from food for thought. All of the above comes from the "fancy stats". All of it helps us understand the value of our assets. All of these points could trigger great discussions, but it is my view that it would be best be done in individual threads: Player Discussion: Josh Bailey, Fancy Stats Discussion.
It's quite easy to cherry pick any stat to make a point, which is easily done with any of the several stats out there. But upon further analysis, using that same criteria/logic, the same number shows results that simply defy logic (as in Brad Boyes, Mark Streit and John Tavares having the worst PDO on the Isles in 2012-13, Cizikas and Okposo the best) YET, in 2013-14 Zeeker is among the WORST and Tavares at the top.
PDO Is a stat that I don't put much stock in. It supposedly tell you whether someone is underperforming, or overperforming. Supposedly it is "puck luck". I, however, do not agree with that and I think it has little value and therefore never found a use for it. I can't defend PDO because I just don't see the merits of it.
I think you're missing the value because you're looking at it as "traditional stats" and "advanced stats" instead of reviewing individual statistics in conjunction with others which is where the value is. PDO isn't very useful, in my opinion. Corsi relative is very useful. Team Corsi is undeniably useful.
I don't want to dismiss the spirit of advanced stats out of hand, but when you apply some critical thinking, the stats simply don't stand up.
These kinds of assertions are what I take issue with. This simply isn't true. You're basically saying that I, BroadwayJay, do not apply critical thinking. This is insulting and untrue. I know you don't mean it that way, so I'm not taking it personally; but the way it is phrased is unfair. Really you mean that you couldn't draw out what you were expecting to see. Your expectations may be too high. Alternatively, you may not know what you're seeing. Nothing wrong with either of those.
Now, I also believe that maybe we don't (currently) have the right stats to measure effectiveness of a player in the context of a free-flowing game like hockey, with so many variables continually in play.
Probably true, but we're working on it and we have some really good stuff already.
Just because we have some numbers available and some, seemingly logical hypothesis can exist, that doesn't mean they are valid in any way.
Theories are theories. You can replace your argument here with climate change and it would be the same. We work with what we have to try and understand the game better. The numbers are real. They are recorded on-ice events. That's it. None of these stats are advanced more than +/- in any way. Ignoring what we have for no reason at all is not a useful exercise.
And for context, I've looked at the same sets of data on several teams, over several seasons, and the lack of real insights is fairly consistent.
You're indicating that you started by knowing what SHOULD be there, which is a problem of course. If what you see in the stats doesn't comport with what you think you know, you should rethink what you think you saw and not just say the stats are wrong.
In teams that I've seen the most hockey first hand (like the Leafs, Isles) and even teams like LA, CHI (Kane is 13th on the team in Corsi Relative for example) or DET where Datsyuk is 16th in PDO - I'm sorry, this just doesn't add up.
Patrick Kane is really NOT a superb possession player. He's a shooter, like Thomas Vanek. Really valuable player, but not necessarily for possession. Also, the Blackhawks use their players in a unique way so that, for example, some forwards play extremely sheltered minutes (Morin) and others do not. We can talk more about Coach Q's unique deployments in another post too.
Would love to hear from those who see this differently and can point out specific examples where a roster decision can be made based on some stats, that, by applying the same logic, would yield consistent results to give the metric real, actionable value.
I would LOVE to do that. And thank you for saying this.
The most obvious would be (1) pulling MacDonald from any contact with the Tavares line (or any line) [already done] and (2) moving a possession driver [someone without big points but great corsi, for now] who can create more shooting opportunities for Tavares and Okposo, someone like Brock Nelson or possibly Josh Bailey.
I made a predictive model based on Corsi which I then used to "predict" last season based on our starting roster. It returned awful numbers that were surprisingly consistent with our actual numbers. Obviously we had many lineup changes, etc... but it gave a pretty good picture of the predictive power of corsi and corsi relative. I've been planning to make it more accurate based on certain additional statistics (penalty +/-, zone starts, and others) and we can slot in players and see where we end up.
Right now, the model has us, assuming no injuries, as probably the fourth best team in the East (were it 13-14). That's not too shabby.
This has been irking me for some time, and it's the off-season, so maybe now is the best time to discuss this, if anyone's into it. If not, camps are only a few weeks away
I'm really glad you brought this topic up, but I think there is just SO SO SO much to talk about that one thread doesn't cover it.
I'm overjoyed to discuss this, but I would ask we confine it to one topic at a time.
I should probably prepare a "fancy stats" primer to help folks understand the usage.
Another poster recently told me that the way we statpeople talk about the stats makes it hard to follow:
Example:
"So I was looking at Nelson's CF%rel and I think when you look at the ZS% and his QoT you have to wonder where he'd be if...."
That's hard to understand.
I liken it to my own line of work:
JUDGE: "Alright I have a decision on the motions. I'm ordering Wade, Mapp, Dunaway/Huntley. We'll do Sandoval right before trial and anything in limine at that time."
Its that jargon that no one else can understand. Whereas, "we're going to have a hearing about whether the initial stop was lawful" is something you can understand.
I'm working on that and I would encourage anyone, at any time, to say "Hey Jay, what did you mean when you said 'Kulemin's WOWY numbers with Grabo are off the charts', speak English to me bud".
I really think we can have some great exchanges and get a good grasp on where our team has been and where it is going if I can learn to communicate a little better. So feel free to ask!
And thank you for asking this question.