HFBoards

Go Back   HFBoards > NHL Eastern Conference > Atlantic Division > Montreal Canadiens
Notices

Sabremetrics/Microstats in Hockey

Reply
 
Thread Tools
Old
04-03-2012, 08:13 PM
  #126
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 17,746
vCash: 500
Quote:
Originally Posted by MathMan View Post
Yah, it did. Gomez ended up being on the ice for about as many goals for as Plekanec, but couldn't get points for them. He didn't manage to collect second assists, notably, which was almost as freakish as his inability to get a goal but didn't get nearly as much attention. His first-assist total is perfectly in line with what little icetime he's had.

More importantly, he also had four different injuries, which do affect performance.
Wow, I guess he's just the unluckiest player in the history of the game. Well, excluding his linemates. Anytime anyone was paired with him they were screwed but uh, I guess that's just luck.


Quote:
Originally Posted by MathMan View Post
And yet again: points total is a very broad-brushed metric that hides any number of important factors and completely ignores half the game. Point totals tell us that Kaberle is a much better player than Josh Gorges and that Mark Streit is way more valuable than Drew Doughty.
No doubt. You can't just look at points to measure a player, never said you should.

I said that point totals can't be ignored when you're looking at a players' offensive ability and that it certainly was a better predictor of future offensive performance than CORSI is. I don't see how this is really disputable in light of the leaders of that particular stat.
Quote:
Originally Posted by MathMan View Post
Sigh. I've had this argument with you before. Oh, the particulars change, but it's basically always the same structure. You have an assumption that you believe to be impeachably correct. The assumption is challenged. You defend the assumption by asserting the assumption without support and dismissing all opposite facts because they contradict the assumption. In reality, you've not actually done anything but put window-dressing on the fact that you believe you're right simply because you're right. And yet another promising thread has devolved in this futility.
I've laid out objections that you have ignored man. Ignoring them doesn't mean they aren't valid. Again, when I look at those CORSI leaders, I'm not sure what I'm supposed to do with them. If Grabvowski is number one... what am I supposed to do with this? It might tell me that he's a little better than I thought he was but not much else.

Quote:
Originally Posted by Roke View Post
Exclusively? Not even close. The A's still employed scouts, they still used scouting and video work. Beane may not have watched his team play games but they had staff doing that, watching other teams play, and writing scouting reports.

I also don't recall him firing half the team sometime between 2001 and 2002 (2002 was the year where they had the 20 game win streak).

The 2001 A's lost a lot of talent to free agency (Jason Giambi, Miguel Tejada, Johnny Damon, and if you overvalue closers like most people still do Jason Isringhausen) but the 2001 team won 102 games and made the playoffs. The 2000 A's won 91 games and made the playoffs.

The teams before 2002 were hardly disasters.
Never said they were. Dude, I'm actually a baseball fan. I REMEMBER the A's losing Jason Giambi.
I was talking about JEREMY Giambi. The guy they brought in the next season.

And actually YES, he did tell his scouting group to go jump in the lake. Grady Huson (head scout) left the club out of frustration and was out there on the air ripping into Beane once he assembled that disastrouts start where they were starting to get blown out every night.

Yes, Paul DePodesta kept telling Beane to be patient and in the film specifically says there will be a regression to the mean and told Beane to stay patient. The team continued to suck and guys like Jeremy Giambi couldn't have cared less. THAT'S when Beane ripped apart the team. THAT club went on to have the 20 game win streak and make the playoffs - in the SAME season. That's what was so crazy about it.
Quote:
Originally Posted by SaskRinkRat View Post
This comment leads me to believe that maybe you didn't watch the movie closely enough (and certainly didn't read the book). The first half of the movie was essentially spent describing this whole concept of "regression to the mean" that MathMan has been talking about. The team was "built with a spreadsheet", as you put it, but the results predicted in the spreadsheet weren't manifesting on the field. It's not that the spreadsheet was wrong (i.e., it's not that the wrong stats were taken into account when evaluating the players), it's that those stats weren't being produced during the early part of the year due to small sample size.

Eventually, the team's performance regressed to the mean (i.e., reverted to what the spreadsheet predicted) and they won the division.
That's right... they did talk about regression to the mean. Only it never happened. It didn't happen until Beane ripped apart the club after a losing streak where they were getting blown out almost every night. He replaced a lot of cancerous players with other guys. THEN the team took off.

I'm not trying to say that Beane didn't use SABR... of course he did. I"m just saying that he learned that scouting does play more of a role than he thought it did.

Lafleurs Guy is offline   Reply With Quote
Old
04-03-2012, 09:17 PM
  #127
SaskRinkRat
Registered User
 
Join Date: Apr 2010
Posts: 340
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
That's right... they did talk about regression to the mean. Only it never happened. It didn't happen until Beane ripped apart the club after a losing streak where they were getting blown out almost every night. He replaced a lot of cancerous players with other guys. THEN the team took off.

I'm not trying to say that Beane didn't use SABR... of course he did. I"m just saying that he learned that scouting does play more of a role than he thought it did.
You're completely revising history. Dumping Giambi and Pena hardly amounted to "ripping apart the club". They specifically refer to the early part of the season amounting to a small sample size - i.e., regression to the mean had not yet occurred. There was nothing whatsoever in the film or the book that should lead you to conclude that the A's over-emphasized analytics in their assembly of the team. In fact, Beane traded Pena precisely because Art Howe had not bought into the analytics enough (playing Pena instead of Hatteberg). It's a logical fallacy to conclude the winning streak happened because those players were traded.

If you want to keep misrepresenting what was presented in either the movie or the book, and continue to revise history to fit your argument, I'd ask that you point to an excerpt from the book that confirms Beane made trades because he lost belief in the analytical approach.

SaskRinkRat is offline   Reply With Quote
Old
04-03-2012, 09:18 PM
  #128
Mathletic
Registered User
 
Mathletic's Avatar
 
Join Date: Feb 2002
Location: St-Augustin, Québec
Country: Canada
Posts: 11,336
vCash: 500
@Lafleurs Guy

You have to take into account that the Moneyball movie is different from the book to start with. There are plenty of segments in the movie that are not talked about at all in the book. Obviously, there was no way they could synthesize the whole book into a movie. The way they built the movie was still great because it got the point across but it's different from reality. That said, in the movie you see the A's moving from old-fashioned scouting to stats analysis overnight. In reality it wasn't the case. They talked about using scientific methods to evaluate players even before Billy Beane was hired as a GM. Even before 2003 they used analytics.

Also, Jeremy Giambi was already on the team. In the movie we see the A's hiring him in 2002 or 2003, can't remember, while he was on the team for a couple of years before that.

That said, it's quite obvious that as along as you have human beings involved there will always be a human element. Before dealing with numbers you deal with human beings. So that aspect of management will never go away.

Also, I strongly recommend reading the book. Honestly, it's a great read. Friends of mine who had no big interest in baseball ... before knowing there would even be a movie about Moneyball ... absolutely loved the book from beginning to end.

Mathletic is offline   Reply With Quote
Old
04-03-2012, 09:19 PM
  #129
EllertoKostitsynGoal
Registered User
 
EllertoKostitsynGoal's Avatar
 
Join Date: Nov 2011
Location: Mtl
Country: Canada
Posts: 1,055
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
No doubt. You can't just look at points to measure a player, never said you should.

I said that point totals can't be ignored when you're looking at a players' offensive ability and that it certainly was a better predictor of future offensive performance than CORSI is. I don't see how this is really disputable in light of the leaders of that particular stat.

I've laid out objections that you have ignored man. Ignoring them doesn't mean they aren't valid. Again, when I look at those CORSI leaders, I'm not sure what I'm supposed to do with them. If Grabvowski is number one... what am I supposed to do with this? It might tell me that he's a little better than I thought he was but not much else.
Corsi isn't really about offensive performance, it's not really about defensive performance either. It's about both. It weighs how much a player takes vs how much he gives up. Let's not forget it's only about 5on5 hockey too.

I would even say Corsi tends to make purely offensive player look better than purely defensive ones, mostly the purely defensive 3rd-4th liners who takes alot of tough minutes. Looking at the guys who have played signifiant toughs or second toughs on our team, the guy who was getting killed the most in Corsi isn't the soft Cammalleri, the lazy AK or the inexperienced Eller, it's actually Moen who looks like a terrible defensive player if you only look at his Corsi (and assume Corsi is only about defence). That's because Moen doesn't create offence so even if he doesn't give up alot (in the context of the minutes he has played), he takes even less.

And you can't use Corsi without context, you can't use any stats without context. Plekanec doesn't look that good looking at his Corsi but then you can look at his quality of competition, quality of teammates, zonestars etc... and he suddenly looks much much better.

And they are a lot more things to advanced stats than Corsi.

About Grabovski, if I'm not mistaken it was his RelCorsi that topped the league. RelCorsi calculates the Corsi of a player relative to his team (it's actually probably a better stat than Corsi (mostly for evaluating players on bad teams) but even that will be impacted by quality of competition etc.. since it's based on original Corsi).

So Grabovski, a really good 5on5 player (let's not forget that these fancy stats mostly concentrate on 5on5 play), a player who had 43 ES pts last year, had the best RelCorsi in the league last year. Looking at Toronto's roster last year, only 3 players had a positive Corsi, that team was absolutely dreadfull on possession. So it's pretty normal that Grabovski would have a great RelCorsi since that team significantly outshot the opposition with Grabovski on the ice but crashed in possession when he got off it (the other two players with a positive Corsi were his linemates).

And Grabovski is probably much much better than you thought he was. He's evolved into a Plekanec type, great two-way player, can carry the play at ES in tough minutes.


Last edited by EllertoKostitsynGoal: 04-03-2012 at 09:29 PM.
EllertoKostitsynGoal is offline   Reply With Quote
Old
04-03-2012, 09:53 PM
  #130
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 17,746
vCash: 500
Quote:
Originally Posted by SaskRinkRat View Post
You're completely revising history. Dumping Giambi and Pena hardly amounted to "ripping apart the club". They specifically refer to the early part of the season amounting to a small sample size - i.e., regression to the mean had not yet occurred. There was nothing whatsoever in the film or the book that should lead you to conclude that the A's over-emphasized analytics in their assembly of the team. In fact, Beane traded Pena precisely because Art Howe had not bought into the analytics enough (playing Pena instead of Hatteberg). It's a logical fallacy to conclude the winning streak happened because those players were traded.

If you want to keep misrepresenting what was presented in either the movie or the book, and continue to revise history to fit your argument, I'd ask that you point to an excerpt from the book that confirms Beane made trades because he lost belief in the analytical approach.
Changing two every day roster players is actually pretty significant man. It's a 25% switch in your roster. They were in last place the day they made that trade. Almost immediately after the moves the team began to win.

I never said he lost faith in the analytical approach at all and that's not what I argued. I said that picking players from a spreadsheet ALONE didn't work. SABR in baseball DOES largely work because baseball was made for it. Beane proved it. But even there you have to look at some intangibles when you build your roster.
Quote:
Originally Posted by Mathletic View Post
@Lafleurs Guy

You have to take into account that the Moneyball movie is different from the book to start with. There are plenty of segments in the movie that are not talked about at all in the book. Obviously, there was no way they could synthesize the whole book into a movie. The way they built the movie was still great because it got the point across but it's different from reality. That said, in the movie you see the A's moving from old-fashioned scouting to stats analysis overnight. In reality it wasn't the case. They talked about using scientific methods to evaluate players even before Billy Beane was hired as a GM. Even before 2003 they used analytics.

Also, Jeremy Giambi was already on the team. In the movie we see the A's hiring him in 2002 or 2003, can't remember, while he was on the team for a couple of years before that.
Of course it's exaggerated. No doubt about that. Point is though that there was a break from traditional scouting to analytics. You are right on Jeremy being there longer though. It's pretty far back and I just looked. You're right I forgot he was there with Jason for a season before his brother left for NY. So yes, I'm wrong on that detail, thanks for correcting.

Bottom line though is that 2002 really marked some big changes for him. He had lost key players in Jason G. and Damon as well as others.
Quote:
Originally Posted by Mathletic View Post
That said, it's quite obvious that as along as you have human beings involved there will always be a human element. Before dealing with numbers you deal with human beings. So that aspect of management will never go away.

Also, I strongly recommend reading the book. Honestly, it's a great read. Friends of mine who had no big interest in baseball ... before knowing there would even be a movie about Moneyball ... absolutely loved the book from beginning to end.
I own the book. BTW, if you want to read a cool book (different subject) go read Game of Shadows. THAT is a cool book.
Quote:
Originally Posted by EllertoKostitsynGoal View Post
Corsi isn't really about offensive performance, it's not really about defensive performance either. It's about both. It weighs how much a player takes vs how much he gives up. Let's not forget it's only about 5on5 hockey too.

I would even say Corsi tends to make purely offensive player look better than purely defensive ones, mostly the purely defensive 3rd-4th liners who takes alot of tough minutes. Looking at the guys who have played signifiant toughs or second toughs on our team, the guy who was getting killed the most in Corsi isn't the soft Cammalleri, the lazy AK or the inexperienced Eller, it's actually Moen who looks like a terrible defensive player if you only look at his Corsi (and assume Corsi is only about defence). That's because Moen doesn't create offence so even if he doesn't give up alot (in the context of the minutes he has played), he takes even less.

And you can't use Corsi without context, you can't use any stats without context. Plekanec doesn't look that good looking at his Corsi but then you can look at his quality of competition, quality of teammates, zonestars etc... and he suddenly looks much much better.

And they are a lot more things to advanced stats than Corsi.

About Grabovski, if I'm not mistaken it was his RelCorsi that topped the league. RelCorsi calculates the Corsi of a player relative to his team (it's actually probably a better stat than Corsi (mostly for evaluating players on bad teams) but even that will be impacted by quality of competition etc.. since it's based on original Corsi).

So Grabovski, a really good 5on5 player (let's not forget that these fancy stats mostly concentrate on 5on5 play), a player who had 43 ES pts last year, had the best RelCorsi in the league last year. Looking at Toronto's roster last year, only 3 players had a positive Corsi, that team was absolutely dreadfull on possession. So it's pretty normal that Grabovski would have a great RelCorsi since that team significantly outshot the opposition with Grabovski on the ice but crashed in possession when he got off it (the other two players with a positive Corsi were his linemates).

And Grabovski is probably much much better than you thought he was. He's evolved into a Plekanec type, great two-way player, can carry the play at ES in tough minutes.
Part of the reason why I posted this thread is that I'm trying to figure out why folks keep going back to this stat. Not sure if it was RelCORSI or not but that would explain a lot if it is. So far I haven't really seen a compelling reason for this stat to really be given the kind of importance I've seen it being given from some posters on here. All I hear is... go to this site or 'its' self explanatory'... I'm sorry but that doesn't cut it. When I see Gomez with a higher CORSI than Crosby I'm skeptical about what kind of value it brings.

Lafleurs Guy is offline   Reply With Quote
Old
04-03-2012, 10:12 PM
  #131
Roke
Registered User
 
Roke's Avatar
 
Join Date: Jul 2003
Location: Winnipeg
Country: Canada
Posts: 1,882
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post

Never said they were. Dude, I'm actually a baseball fan. I REMEMBER the A's losing Jason Giambi.
I was talking about JEREMY Giambi. The guy they brought in the next season.

And actually YES, he did tell his scouting group to go jump in the lake. Grady Huson (head scout) left the club out of frustration and was out there on the air ripping into Beane once he assembled that disastrouts start where they were starting to get blown out every night.

Yes, Paul DePodesta kept telling Beane to be patient and in the film specifically says there will be a regression to the mean and told Beane to stay patient. The team continued to suck and guys like Jeremy Giambi couldn't have cared less. THAT'S when Beane ripped apart the team. THAT club went on to have the 20 game win streak and make the playoffs - in the SAME season. That's what was so crazy about it.

Jeremy Giambi is not half the team. Jeremy Giambi wasn't even one of the ten most important players on the Oakland A's. Jeremy Giambi (who was an average outfielder for the A's) for John Mabry (who was an average player for the A's).

Jeremy Giambi's cancerness also didn't cause the A's to underperform in 2000 or 2001.

The only other transactions the A's made around the time Beane "ripped the team apart" by trading Giambi was releasing a utility player who wasn't even in the majors and releasing reliever Mike Holz who had appeared in 16 games (and been horrible, mind you) a couple of weeks later.

Quote:
That's right... they did talk about regression to the mean. Only it never happened. It didn't happen until Beane ripped apart the club after a losing streak where they were getting blown out almost every night. He replaced a lot of cancerous players with other guys. THEN the team took off.

I'm not trying to say that Beane didn't use SABR... of course he did. I"m just saying that he learned that scouting does play more of a role than he thought it did.
Regression to the mean did happen though, to oversimplify things the regression went "Too far" and the A's even over-performed their talent level for a short period by going 20 games without a loss. Trading an average outfielder for an average outfielder and releasing a bad back-of-bullpen reliever didn't make the A's a much better baseball team. Random variance has more to do with any long streak like that (in modern sports) than a single transaction

If going on a massive winning streak was as simple as dealing one player labelled a cancer the Cubs would have gotten a lot better after getting ird of Milton Bradley, or Carlos Zambrano. Teams letting John Rocker go didn't suddenly get better, nor Shea Hillenbrand, nor Carlos Silva.

Roke is offline   Reply With Quote
Old
04-03-2012, 10:20 PM
  #132
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,855
vCash: 500
Coincidence is the mother of narrative.

MathMan is offline   Reply With Quote
Old
04-03-2012, 10:27 PM
  #133
SouthernHab
Registered User
 
SouthernHab's Avatar
 
Join Date: Mar 2011
Location: USA
Country: United States
Posts: 8,304
vCash: 500
Quote:
Originally Posted by Mathletic View Post
That's funny because I actually had the Cards as favorite in the NL to play in the World Series to play Texas simply based on stats. There's more to statistical analysis than just picking the best record from each league and say, well if this team doesn't win, it means analytics don't work.

Also, the Cards rely on analytics. The new Astros GM used to work for the Cards and was part of their analytics department. Their analytics told them to draft players like Allen Craig, Jon Jay, Lynn and others. It also told them to make the Rasmus trade and so on.

As for the Canucks, I hope you understand that even though the Canucks had the better regular season record, the Bruins weren't that far behind. A team with the Canucks record would beat the Bruins about 55% of the time in a 7-game series. That's basically a coin flip. Not sure how that proves that anlytics can't help you in hockey.

As for chemistry. Like I said to somone earlier. Stop for a second and explain to me what chemistry is. Simply saying chemistry is so complex that it can't be measured won't lead you far ... so ironic btw, since actual chemistry is a science in itself but anyways ...

So what could account for chemistry. Maybe the n' of games 2 or more players have played together. The number of minutes they have played together. Number of teams they have been on together. Maybe it has to do with complementary skills. For instance, does a passer help a scorer score more. Does a player who creates turnovers helps another player to create even more turnovers. Do 2 playmakers offset one another. Do puck hogs help or stop one another from scoring. All this analysis can take place.

Now, maybe chemistry for you is simply well they like going to the restaurant together, so chemistry is just that. Then yeah maybe numbers won't help you. But then again, there's a science that studies group behaviour, social networks and so on.

Again, what does coming together mean? Does it mean that a team filled with similar players will start complementing one another. Start defining what your intangible is and I'm sure we'll find a way to find that in the data.
I will try to make it more simple.

Micro-stats or any other stat is an AWESOME tool when you are playing NHL 12 on the computer. There is no emotion. No brains.

If you cannot understand what an intangible is regarding sports and the performance of a human athlete with a brain and the capacity for emotion, well, enjoy your stats in the search for the meaning of perfect reliability in an imperfect world.

SouthernHab is offline   Reply With Quote
Old
04-03-2012, 10:28 PM
  #134
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 17,746
vCash: 500
Quote:
Originally Posted by MathMan View Post
How often does the best team in the league win the cup? This is a good analysis: http://objectivenhl.blogspot.ca/2011...-team-win.html

Given the structure and the samples sizes involved, the best team in the league is much less likely to win the Cup than it is not to.
A couple of things here...

1. Over the course of a 7 game series anyone can win. No doubt about it. It's not surprising at all that the 'best' team doesn't usually win over the course of the playoffs.

2. For some reason, the league seems to allow rougher play in the playoffs than in the reg. season. Don't know why that is but it sure seems to happen a lot. Witness the Canucks vs. Bruins. If the parameters of what's permitted suddenly changes, then the 'best' team in the reg season may not be the 'best' team suited to win in the playoffs.
Quote:
Originally Posted by MathMan View Post
I asked you if the the force of hits matter outside of their immediate impact on the game. You're telling me "I don't know and nobody can find out, but I say it does and you must accept it on faith", which is essentially what you're saying when you say something is 'impossible to quantify'.

You bet that's a cop out.
?

I don't duck questions and you should know that by now. And if I'm wrong on something, I have no problem admitting it.

The whole point I was making was that a hit from Chara isn't distinguished from one from Deharnais... As we all know though, there IS a difference. My question was HOW do you quantify it? Where does it show up? I've said that it's very difficult if not impossible to quantify. However, those who've actually played the game know that when you get rocked by a player you're less likely to go to those danger zones than you otherwise would. What I'm saying is though... it's difficult to quantify this. As I said, the player may still go to the net but might not go to the net as hard. But it's difficult if not impossible to capture this.

Not sure why you are now saying that this is somehow a cop-out when this is the point I was starting from to begin with. Maybe I'm misunderstanding something here or maybe you are...
Quote:
Originally Posted by MathMan View Post
Because he's a super-special in a way that even Lidstrom isn't, or because he's a very good defenseman and very good defensemen help teams win hockey games?

I can buy the latter. The former is an article of faith. You're gonna have to support that claim if you're gonna make it.
He's a completely different player than Lidstrom. Pronger sucks on international ice surface and under tight Olympic rules but in the NHL playoffs he's pretty awesome. Both players can do things the other can't and both are HOFers. If you asked me who I'd want on my Olympic team it would be Lidstrom regardless of what team I had. If you asked me who I'd rather have for the NHL playoffs though? I'd look at my roster first and then make the decision. Tough to choose anyone over Lidstrom as he's arguably a top 5 blueliner of all time but I'd look at my team and then decide.

As another example, if I'm starting from nothing as an expansion team and had to choose between Yzerman vs. Messier, I'd personally take Yzerman. I think he's more skilled and I feel like he'd help me more than Mess would. (Many people would disagree with me on this point but it's not important here...)

However, Messier brings a different skill set than Yzerman does. And if I'm already packed with skilled scorers I may want a different dimension on the team. As an example, if it's 1987 and I know that I've got Gretzky, Lemieux and tons of other scorers in my lineup already, I might forgoe another scorer in favour of a guy like Messier who might not be as skilled but is skilled enough and brings a different dimension to my team.

Make sense?
Quote:
Originally Posted by MathMan View Post
So was Marian Hossa. So was Ty Conklin. I'm sure Conklin made three different finals because he was so intimidating from the bench or maybe he has magical powers. Or maybe it's just coincidence.

Ty Conklin, man... three Finals on three different teams. No way that's a fluke.
The difference is that Pronger was the best player on his team in most of those cases (arguable in Anaheim where there was Niedermayer and Selanne) those other guys weren't. Moreover, Pronger took a team that had no business being in the playoffs to begin with and led them to the finals. Pronger's points per game actually goes UP in the postseason man. He's a monster in the playoffs.
Quote:
Originally Posted by MathMan View Post
How come Chara didn't make them awesome the way he apparently did Tim Thomas?
Because he didn't have a magic wand.

I don't think Chara made Thomas great. But I do think he made his job easier. How much easier... well that's the point I was trying to make - it's tough to actually know. You can try the equivalent to WARP in hockey (whatever that stat is) to try to figure it out and maybe it will give you some idea but I think it's tough to quantify this as you're trying to figure out if the chicken or the egg came first. It's a symbiotic relationship and it works.

I think Carey Price wishes he had the kind of protection in front of him and if he did I think it would make a difference in his numbers. You're able to make a lot more saves when you can actually see the puck. And you can see the puck a lot better when your giant blueliner is clotheslining people in front of your net for half the game.


Last edited by Lafleurs Guy: 04-03-2012 at 10:35 PM.
Lafleurs Guy is offline   Reply With Quote
Old
04-03-2012, 10:38 PM
  #135
Roke
Registered User
 
Roke's Avatar
 
Join Date: Jul 2003
Location: Winnipeg
Country: Canada
Posts: 1,882
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
Part of the reason why I posted this thread is that I'm trying to figure out why folks keep going back to this stat. Not sure if it was RelCORSI or not but that would explain a lot if it is. So far I haven't really seen a compelling reason for this stat to really be given the kind of importance I've seen it being given from some posters on here. All I hear is... go to this site or 'its' self explanatory'... I'm sorry but that doesn't cut it. When I see Gomez with a higher CORSI than Crosby I'm skeptical about what kind of value it brings.
Like EllertoKostitsyn said, you have to consider the context. Crosby faces the best opposition and (generally... this season is a bit wonky because of the injuries and Pittsburgh is ridiculously good) starts in the defensive zone more than the offensive zone.

Talking about Gomez last year (this year injuries made things wonky and the coaching change has him used in weird fashions) - he generally faced the 2nd-best opposition and started in the offensive zone more than the defensive zone (because the Habs were pretty good at pushing the puck toward the opposition goal and Halpern's lines were able to take a lot of defensive zone draws).

By looking at the context of usage and using intuition (your scepticism works well here) you can conclude that while Gomez' "Corsi Rel" was better than Crosby's last year and their "Corsi On" was close Crosby is still a better player than Gomez is and there's no debate.

Corsi (or it's brother Fenwick which just excludes blocked shots) is an excellent tool at the team level. When looking at individuals it can still be useful, it just means you have to look at what kind of matchups the individual gets as well as which zone the player starts in (or zone-starts...if you're familiar with the behindthenet stuff look at Vancouver's Zone Starts for their centres... ingenious coaching by Vigneault)

Looking for context in the microstats is not all that different than in goals/points (though comparing the shooting % to the player's career is another thing to keep in mind).

An easy example, last year Ville Leino put up 19G and 53Pts which looks like a decent 2nd-line player. Looking at how he was used though, Leino started a lot in the offensive zone at even-strength (significantly more than Gomez last year) and faced 3rd-line opposition. Looking at the percentages the Flyers' shooting percentage when he was on the ice (on-ice sh%)was high by about 2% (10% where ~8% is normal), and his personal shooting percentage was 16.2% when in his career he's about a 13% shooters.

Add in the fact Leino was easily the 3rd-best player on his forward line (IIRC he played 80%+ of the time with Briere and Hartnell) and it's not surprising that the signing blew up in Buffalo's face.

Context is everything and at the individual player level I find the Quality of Competition and Zonestart reports at the behindthenet website as valuable as the individual's Corsi. Even if you ignore Corsi and just look at the traditional goals/assists/points, looking into the percentages, the Zonestarts, and quality of an opposition give me a lot better view of what a player produces.

Roke is offline   Reply With Quote
Old
04-03-2012, 10:44 PM
  #136
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 17,746
vCash: 500
Roke, Jeremy Giambi didn't have as big a role in the previous years. There was his brother and Damon there. When they left, he assumed a bigger role. Of course he was never a star, but replacing him and Pena was still big. Changing two everyday roster players is a big move regardless of the players involved.
Quote:
Originally Posted by MathMan View Post
Coincidence is the mother of narrative.
In the case of Beane's A's, it's possible. But those wins started the day after they made the moves. Coincidence? Maybe... Sure seems weird how those wins piled up starting on May 23rd after they made the moves on the 21st.

Man... I'm going to have to take a break from this thread. I'm getting tired.

Lafleurs Guy is offline   Reply With Quote
Old
04-03-2012, 11:09 PM
  #137
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,855
vCash: 500
It's called post hoc ergo propter hoc, and it really is the source of many cool sports stories.

MathMan is offline   Reply With Quote
Old
04-04-2012, 09:33 AM
  #138
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 17,746
vCash: 500
Quote:
Originally Posted by MathMan View Post
We're derailing the thread and I should know better than to get into this argument with you again, but I think it bears pointing out that your recollection of the events may be a bit... off. And you may be misrepresenting what I say as a result, so for the record:

Montreal's microstats were horrible in 09-10. Ghastly, godawful, bottom-5-in-NHL awful. Maybe you weren't paying attention, maybe you just forgot, but I spent much of that year whining about it. They ended up riding absurd goaltending and a strong PP to the playoffs in a weak conference, despite a 5-on-5 game best described as abysmal. We had a lot of fun in the playoffs, but the run was no different than the regular season: constantly outplayed and surviving on goaltending.

They were terrible. I personally blamed Martin which, in hindsight, was either wrong or premature, because he completely turned it around the very next year (he had much better personnel to work with).

Montreal's microstats in 10-11 were very good. Unfortunately, they couldn't buy a goal, which is a "luck" thing, but there's no denying the club was far superior 5-on-5 despite this and it was, again, buoyed by strong special teams. The Habs were five wins better than the previous season (a significant amount). They slightly outplayed the Bruins the playoffs, but Tim Thomas Halaked them.

Montreal's microstats in 11-12 started out excellent, despite a punishing injury situation. And the 5-on-5 goals for/against numbers matched that, showing another leap of improvement. Despite all the injuries, Montreal had excellent 5-on-5 and excellent PK. The PP was not generating goals, but it was generating shots, so this was bound to turn around. The Habs had trouble with narrowly losing games, often after dominating play. This happens and it's generally not a bad sign because teams that lose this way tend to win in the long run. It was an extremely promising season despite the record and the injuries. Then Gauthier panicked, Cunneyworth happened, and Montreal's microstats, goals, and record went entirely to pot.

So yeah, lumping the three years together will paint a picture and that picture will be muddied and inaccurate. We're not talking hypothetical intangibles that may or may not affect the game. These were significant differences between the three seasons in the way the team played and the results they got. Analytics help us see and understand these changes but really, they ought to be obvious if one simply paid attention to more than the boxscores.
It doesn't change my original point. You consistently have overrated our team.

In '10 I told you Gomez was a 70 point guy MAX. You pointed to the microstats and said we were going to have a great year. As you said, it was horrible. Yes, the microstats were horrible but going into the season, you said the microstats would be great because players with good microstats tend to have good stats. I asked you about the declining point totals of Gomez, you said don't worry about it... he'll be awesome.

In '11 Gomez continues to decline despite having good CORSI. The team as a whole has good CORSI. We still finish bottom 3rd in goals and Price has a spectacular year and we wind up with 96 points.

In '12 We actually score 5 on 5 but the PP tanks. We're outside the playoffs when Martin leaves and we were bottom 3rd in scoring... again. Now we stand at the bottom of the league and are 22nd in goals.

If the results were different I'd be less skeptical. But the bottom line is that CORSI hasn't really been a great predictor of success for us.
Quote:
Originally Posted by MathMan View Post
It's called post hoc ergo propter hoc, and it really is the source of many cool sports stories.
And maybe that was the case. Maybe the win streak beginning two days later was coincidence or maybe it was a kick in the butt that the club needed.

My point wasn't to try to tear down SABR in baseball or say that Beane learned that it sucked... not the case. My point is that I think he learned that you have to look beyond the spreadsheet. You can still get guys who conform to your model... just get guys who aren't at the strip joint every night.

Lafleurs Guy is offline   Reply With Quote
Old
04-04-2012, 10:49 AM
  #139
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,855
vCash: 500
Lafleurs' Guy, we've been over all that before. Personally, I'm not interested in spending more hours putting together the same answers just so you'll reject them out of hand again.

I've said this before, but if you have interest in serious hockey analysis, you're going to need to start looking at things below the surface level. Analytics are about finding out the right process. It is a basic prerequisite when discussing them to understand that the quality of the results do not always (and in a game as variable as hockey, often do not) reflect on the quality of the process; the right process gives you the best odds of the right results, but nothing is guaranteed. In other words, the team that wins may not have been playing well and the team that loses may not have been playing poorly.

This is a difficult lesson for sports fans, because it forces one to give up some much-cherished notions, most notably that the team that wins always "deserves" to win. It took me a while myself. It only really fully resonated with me after looking back at the way the Habs fluked their way to the ECF.

This, I suspect, is why you don't really understand analytics; you only view the game at the results level, and then only the shallowest, least granular level. The kind of stuff you find in the average sports section in the newspaper, basically, and you aren't motivated in actually putting those assumptions in question.

Now. Montreal, it is true, has gone through a number of upheavals and odd occurences lately which makes it a difficult example for analytics, unlike most other teams; it hasn't landed in the middle of the bell curve very much, it's been either feast or famine from 2007-2008 on out. It happens, but it means the team we follow the most is not the best case study for analytics compared to most others. But personally, I'm interested in finding out why the Habs end up where they are, on a level of detail you don't seem to be interested in. That's fine, but maybe it means that analytics are not for you.

ADDENDUM: Also, if you want to cogently discuss analytics in any depth, it's important to acquire at least a basic understanding of probability science. Not hockey analytics, just basic statistics-and-probabilities stuff. If you don't want to do that, that's fine, but again: analytics may not be for you.


Last edited by MathMan: 04-04-2012 at 11:05 AM.
MathMan is offline   Reply With Quote
Old
04-04-2012, 11:19 AM
  #140
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 17,746
vCash: 500
Quote:
Originally Posted by MathMan View Post
Lafleurs' Guy, we've been over all that before. Personally, I'm not interested in spending more hours putting together the same answers just so you'll reject them out of hand again.

I've said this before, but if you have interest in serious hockey analysis, you're going to need to start looking at things below the surface level. Analytics are about finding out the right process. It is a basic prerequisite when discussing them to understand that the quality of the results do not always (and in a game as variable as hockey, often do not) reflect on the quality of the process; the right process gives you the best odds of the right results, but nothing is guaranteed. In other words, the team that wins may not have been playing well and the team that loses may not have been playing poorly.

This is a difficult lesson for sports fans, because it forces one to give up some much-cherished notions, most notably that the team that wins always "deserves" to win. It took me a while myself. It only really fully resonated with me after looking back at the way the Habs fluked their way to the ECF.

This, I suspect, is why you don't really understand analytics; you only view the game at the results level, and then only the shallowest, least granular level. The kind of stuff you find in the average sports section in the newspaper, basically, and you aren't motivated in actually putting those assumptions in question.

Now. Montreal, it is true, has gone through a number of upheavals and odd occurences lately which makes it a difficult example for analytics, unlike most other teams; it hasn't landed in the middle of the bell curve very much, it's been either feast or famine from 2007-2008 on out. It happens, but it means the team we follow the most is not the best case study for analytics compared to most others. But personally, I'm interested in finding out why the Habs end up where they are, on a level of detail you don't seem to be interested in. That's fine, but maybe it means that analytics are not for you.

ADDENDUM: Also, if you want to discuss analytics cogently, it's important to acquire at least a basic understanding of probability science. If you don't want to do that, that's fine, but again: analytics may not be for you.
I can see how it works extremely well for baseball. I haven't seen the evidence of it working well for hockey. Maybe it does... I just haven't seen the evidence. I'm not saying it can't work for hockey or that it won't in the future. I do think it will be harder to do and I doubt that it will be able to capture all the subleties of the game.

Moreover, I don't think anything should be taken at face value. Too often in your explanations on things that don't support your arguments you resort to terms like 'luck' or dismiss objections with 'analysts have thought of that' but not following it up with any substantive argument. I think formulas have to be challenged. It's the challenges that force these analysts to re-think their formulas. And if they can withstand challenges then you know that the formulas have validity. But if they can't withstand the challenges, then we're all the better off.

Baseball SABR is far more advanced than hockey is. And yet people are STILL working on new formulas to capture subtelties of the game. Like I said, some formulas are revised and others discarded in favour of new ones. We're just at the start of these kinds of analytics now in hockey. Fact is that 'counting things' is a lot harder in hockey than in some other sports.

Saying 'maybe this isn't for you' doesn't really do anything but make you look weak. Dismissing questions or objections doesn't do anything but raise more skepticism. I'm not satisfied with your arguments. You've been wrong most of the time on the results and have continuously referred to microstats as being the basis of your beliefs. Gomez was supposed to be putting up 70+ points a season according to you. The Habs were supposed to putting up 100 point seasons according to you. It hasn't worked out. And it's gone more or less the way I thought it would.

I don't think that it's wrong to challenge you on this. I think that challenging you on this is pretty important if we're going to try to test the value of this kind of analysis. I understand 'regressing to the mean' unfortunately, all I've seen from Gomez over the past three years is regression. At what point does the guy stop being mean to us and start actually producing?

Lafleurs Guy is offline   Reply With Quote
Old
04-04-2012, 11:33 AM
  #141
Ohashi_Jouzu
Registered User
 
Ohashi_Jouzu's Avatar
 
Join Date: Apr 2007
Location: Halifax
Country: Japan
Posts: 20,715
vCash: 500
Problem is, Mathman, whatever numbers your system comes up with will only report what a player did. It won't tell you how or why. And since this is a (non-"turn-based") team game, where coaches and GMs constantly tweak rosters attempting to find all the "right fits", whatever model you come up with to show you the "what", you'll never get the "how" or "why". IMO, "understanding hockey" is all about the "how and why", and the "what" is simply what played out.

Any statistical breakdown/analysis will show you what the individuals were able to "produce" under a certain "set" of conditions. What Lafleur's Guy are saying, I think, is that the "how" and "why" are likely more important when it comes to understanding what makes players able to contribute what they do to the fortunes of a team. If a significant enough variable is changed (say, a Chara or Pronger introduced to the equation), all of a sudden the model of production you came up with is rendered useless. Someone is going to get bumped down in depth, changing the situations/opportunities they get in a game, other players are going to play better because of the confidence knowing someone like them is behind you, etc, etc. Another extreme would be to imagine the impact of changing a variable like the coach, or fewer/more team injuries to work around, and figuring out how that statistically impacted each individual, but I don't think further elaboration is necessary. It obviously becomes meaninglessly cumbersome, at the very least.

At a certain level, the depth of Sabrmetrics you can get to is simple cerebral masturbation for those who spent years counting things and/or learning advanced statistical methods that they were disappointed to find out no one really cares about (no matter how "interesting"), and in fact, often do a better job without through observant experience. Watching the game, studying the human side of players, and recognizing patterns to find "the right fit" remains the best way for GMs to evaluate an asset that they are looking to trade/acquire/draft, but all these advanced stats are definitely useful for fairly extensively describing how productive a player was (in terms of countable events, at least).

Ohashi_Jouzu is offline   Reply With Quote
Old
04-04-2012, 11:41 AM
  #142
EllertoKostitsynGoal
Registered User
 
EllertoKostitsynGoal's Avatar
 
Join Date: Nov 2011
Location: Mtl
Country: Canada
Posts: 1,055
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
.

Part of the reason why I posted this thread is that I'm trying to figure out why folks keep going back to this stat. Not sure if it was RelCORSI or not but that would explain a lot if it is. So far I haven't really seen a compelling reason for this stat to really be given the kind of importance I've seen it being given from some posters on here. All I hear is... go to this site or 'its' self explanatory'... I'm sorry but that doesn't cut it. When I see Gomez with a higher CORSI than Crosby I'm skeptical about what kind of value it brings.
Like I said, there are a lot more things to advanced stats than Corsi and Corsi is about how much a player gives up relative to how much he takes. Now, the results (goals created and the points that comes with them) will depends on a couple of things. For example, a player outshooting 325-300 will have the same Corsi as a player outshooting 650-600 but the latter will, all thing equal, provide double the offence.

Another thing, on-ice shooting % (wich is the % of the shots taken while X player is on the ice gets converted into goals, to put in simply, if X player as a 10% on-ice shooting then opposing goalies have a. 900 SV% at while he's on the ice at ES) will make the results varies, especially over a relatively small sample.

A 10% on ice shooting is considered pretty high and not really substainable for most players but some players can maintain that over a long period (their offensive prime). Crosby is one who can, wich is why even if he ends up with the same Corsi as player Y even if the shots taken numbers are the same he'll create more goals while giving up the same (on ice save being equal) and thus be more valuable. Crosby has never had a on ice shooting lower than 11% (not since those things have being getting counted anyway) , that is in itself something pretty valuable considering that it's proven that he can that up. that's part of what makes him an elite player.

EllertoKostitsynGoal is offline   Reply With Quote
Old
04-04-2012, 12:01 PM
  #143
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,855
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
Moreover, I don't think anything should be taken at face value. Too often in your explanations on things that don't support your arguments you resort to terms like 'luck' or dismiss objections with 'analysts have thought of that' but not following it up with any substantive argument. I think formulas have to be challenged. It's the challenges that force these analysts to re-think their formulas. And if they can withstand challenges then you know that the formulas have validity. But if they can't withstand the challenges, then we're all the better off.
Fair, but the problem is that you don't seem to be interested in going into the research about it (such as the list of links I posted above). If you're going to challenge everything, maybe you need to understand analytics at least at the FAQ level, if not deeper.

The evidence you want is also there. I'll admit that it's not necessarily well-organized at all and there isn't really such a thing as a FAQ (though arcticicehockey.com has a primer that comes close), but everything is there.

If you want to learn about it at the level you seem to demand, it's going to take effort because there's a lot of material and it can't be easily spoon-fed. And you need to look at it with an understanding of general probability as well.

Quote:
Originally Posted by Lafleurs Guy View Post
Baseball SABR is far more advanced than hockey is. And yet people are STILL working on new formulas to capture subtelties of the game.
Here's the thing about those subtleties: they're subtle. They do make a difference, but it's a small difference. The Big Things That Matter have been understood for a while, and now things are being refined, but as more and more refinements happen, they don't challenge the basic stuff, they just make it more accurate. Formulas aren't being "discarded". They are being "refined". This is true both in hockey and in baseball.

A subtlety that explains maybe half a percent of winning is valuable to research, but that doesn't mean you can point to that subtlety not being completely analyzed as invalidating the whole model.

Quote:
Originally Posted by Lafleurs Guy View Post
Saying 'maybe this isn't for you' doesn't really do anything but make you look weak.
It might, but here's the thing: there's some stuff you need to understand that you don't, among other things a basic understanding of general probability, which underpins all sports analytics. And if you're not interested in learning that, then there's not much point in discussing the analytics. And you're certainly giving me the impression you're not really interested in learning.

Until you understand why the notion of one player being the primary element in three different teams to the Cup Finals or the notion that a baseball team making a few trades can directly lead to a 20-game win streak as the main factor is silly (not questionable, silly), this discussion is not going to be very fruitful. And if you hold to those notions too closely for them to be challenged, then no, analytics and probabilities are not for you.

Quote:
Originally Posted by Lafleurs Guy View Post
Dismissing questions or objections doesn't do anything but raise more skepticism.
And now you know how I feel whenever I debate with you.

Quote:
Originally Posted by Lafleurs Guy View Post
You've been wrong most of the time on the results and have continuously referred to microstats as being the basis of your beliefs. Gomez was supposed to be putting up 70+ points a season according to you. The Habs were supposed to putting up 100 point seasons according to you. It hasn't worked out. And it's gone more or less the way I thought it would.
I'm pretty sure when I talked about Gomez getting 70 points I called it a best-case scenario, which was based on the notion that he'd get 40ish points and benefit from the Habs' continually strong powerplay. Only half of that equation happened in 2009-2010. Setting up straw men by misrepresenting what I say is another favorite tactic of yours, I've noticed.

By the way, you didn't exactly predict a lottery finish for the Habs this year, did you? How come?

Quote:
Originally Posted by Lafleurs Guy View Post
I understand 'regressing to the mean' unfortunately, all I've seen from Gomez over the past three years is regression. At what point does the guy stop being mean to us and start actually producing?
Hilarious. And while you were trying to make a joke, it also tends to show that you don't really understand what an important concept like "regression to the mean" means...

MathMan is offline   Reply With Quote
Old
04-04-2012, 12:11 PM
  #144
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,855
vCash: 500
Quote:
Originally Posted by Ohashi_Jouzu View Post
At a certain level, the depth of Sabrmetrics you can get to is simple cerebral masturbation for those who spent years counting things and/or learning advanced statistical methods that they were disappointed to find out no one really cares about (no matter how "interesting"), and in fact, often do a better job without through observant experience. Watching the game, studying the human side of players, and recognizing patterns to find "the right fit" remains the best way for GMs to evaluate an asset that they are looking to trade/acquire/draft, but all these advanced stats are definitely useful for fairly extensively describing how productive a player was (in terms of countable events, at least).
There's an interesting, hidden contradiction implied here: you don't believe studying the "what has gone before" will lead to useful prediction of the future, but you value observant experience and "observant experience" is nothing else than knowledge of "what has gone before".

One way to look at analytics is precisely as the wide-spread collection, quantification, communication, and normalization of observant experience, with the goal of sharing this experience, recognizing patterns, and understanding what factors matter and to what degree. It does exactly what you describe as valuable, except with more data and less subjectivity.

MathMan is offline   Reply With Quote
Old
04-04-2012, 12:44 PM
  #145
Miller Time
Registered User
 
Miller Time's Avatar
 
Join Date: Sep 2004
Posts: 7,736
vCash: 500
Quote:
Originally Posted by MathMan View Post
There's an interesting, hidden contradiction implied here: you don't believe studying the "what has gone before" will lead to useful prediction of the future, but you value observant experience and "observant experience" is nothing else than knowledge of "what has gone before".

One way to look at analytics is precisely as the wide-spread collection, quantification, communication, and normalization of observant experience, with the goal of sharing this experience, recognizing patterns, and understanding what factors matter and to what degree. It does exactly what you describe as valuable, except with more data and less subjectivity.
i think your misconstruing his point...

"observant experience" isn't simply about a knowledge of "what has gone before", it's the expertise (from hours of dedicated practice) that allows for one to appreciate/recognize the subtle differences in performance that explain their effectiveness.

some examples:

a novice wine taster can't properly describe/analyze "why" they like a certain wine better than another, whereas an experienced taster will be able to breakdown exactly what he is "tasting" & describe what about the wine is causing the experience he's observing...

someone with no experience/training with classical music may very well appreciate the performance of a great work by mozart (or whomever), but can't explain/describe why or what they are hearing, or distinguish it much-if at all, from a far "lesser" performance of the same work, whereas an experienced observer will be able to break down the experience and accurately (or to a strong degree of accuracy) describe what makes the given performance so strong/weak.


could go on and on... point is that as expertise grows, its not so much the ability to predict the future that is the result (although that is arguably one of them, more on that below), its the ability to accurately "know" what it is you are seeing... to be able to break down the shots/passes/hits/takeaways/possession time/skating et.et. and understand not just "what" happened (which is all that the stats can tell you, at least until neurimaging makes its way into helmets) but "why" it happened... and that's the most important piece in figuring out if a certain player is likely to "do it again".

going further, there is growing evidence that the "intuitive" responses of expert observers (which is actually very akin to the amazing performances of expert performers when they are "in the zone") reveals the ability of human brain to problem solve (for lack of a better word) at a predictive level.


there are tons of interesting books/research on the matter, pm me if you'd like a list of some.

bottom line is that there definitely is something to the quality of an expert observer that existing statistical modelling simply can't touch... at least not yet.

that said, "expert observers" are not all of the same quality, and aside from track record (which itself is subjective... how do you comparatively measure a scouts impact/ability vs another?) there isn't really a concrete measuring stick to discern which scouts you should be hiring/listening to...

and that certainly opens the door to relying more and more on the developing statistical models, though I'd argue they are best used as a complement to a competent scouting department... you still want to have the best "human observer" you can find running the show and making the final decisions.

Miller Time is offline   Reply With Quote
Old
04-04-2012, 01:06 PM
  #146
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 17,746
vCash: 500
Quote:
Originally Posted by MathMan View Post
Fair, but the problem is that you don't seem to be interested in going into the research about it (such as the list of links I posted above). If you're going to challenge everything, maybe you need to understand analytics at least at the FAQ level, if not deeper.
I've looked at the links. I've seen the math. I haven't seen anything though that suggests that these formulas are worth anything. THAT is what I'm trying to get at. Apparently there are some clubs who've begun using this kind of analysis. I'm just trying to figure out how useful these kinds of analytics actually are.
Quote:
Originally Posted by MathMan View Post
The evidence you want is also there. I'll admit that it's not necessarily well-organized at all and there isn't really such a thing as a FAQ (though arcticicehockey.com has a primer that comes close), but everything is there.

If you want to learn about it at the level you seem to demand, it's going to take effort because there's a lot of material and it can't be easily spoon-fed. And you need to look at it with an understanding of general probability as well.
The fact that it's not well organized is part of the problem. Not only from an information gathering standpoint but also from the standpoint that it just doesn't seem to be there yet. It's in its infancy right now...
Quote:
Originally Posted by MathMan View Post
Here's the thing about those subtleties: they're subtle. They do make a difference, but it's a small difference. The Big Things That Matter have been understood for a while, and now things are being refined, but as more and more refinements happen, they don't challenge the basic stuff, they just make it more accurate. Formulas aren't being "discarded". They are being "refined". This is true both in hockey and in baseball.

A subtlety that explains maybe half a percent of winning is valuable to research, but that doesn't mean you can point to that subtlety not being completely analyzed as invalidating the whole model.
I don't think it invalidates the model and I don't think I said this. If I did then I mispoke. I do think that those 'subtle' differences are 'subtle' because they're hard to capture if not impossible on a spreadsheet. The force of a Chara hit does more to win than I think you believe. And I think it's going to be really hard to capture this. You say that it's subtle because it doesn't really have an effect on the game... maybe it does. I say it's subtle because it can't be measured... doesn't mean the effect isn't there. Stats will try to capture WAR in baseball and I suppose you can try to do this in hockey. There are problems with WAR to begin with but in hockey I'd think it would be even harder. I'm not sure how else you really try to capture the force of those hits though.
Quote:
Originally Posted by MathMan View Post
It might, but here's the thing: there's some stuff you need to understand that you don't, among other things a basic understanding of general probability, which underpins all sports analytics. And if you're not interested in learning that, then there's not much point in discussing the analytics. And you're certainly giving me the impression you're not really interested in learning.

Until you understand why the notion of one player being the primary element in three different teams to the Cup Finals or the notion that a baseball team making a few trades can directly lead to a 20-game win streak as the main factor is silly (not questionable, silly), this discussion is not going to be very fruitful. And if you hold to those notions too closely for them to be challenged, then no, analytics and probabilities are not for you.
See... this kind of thing is where you really lose me. I'll use a more extreme example here. Look at Gretzky. The guy goes to LA and leads them to a cup final. You think it's sillly to cite him as the principle reason they got to a cup after trading for him? You think it's coincidence that he was on that club? You don't think that one player can make a huge difference to a clubs' fortunes? Seriously... you believe this?

If thats' what your analytics are telling you then one of the following is going on here.

1. Analytics offers little value in the way of predicting how good a team is going to be and is not a useful tool in building your team.
2. Analytics is not developed enough to provide meaningful analysis.
3. You don't understand the analytics yourself and are misinterpreting the data or drawing the wrong conclusions.
4. I've misunderstood what you're saying here. Because what I think you're trying to say doesn't make any sense.

Right now I've got to go with number 3. There's no way that any meaningful analysis would suggest that a player like Gretzky wouldn't be a primary reason for LA getting to the finals. You've got to be missing or misinterpreting something there. Ditto with Pronger.
Quote:
Originally Posted by MathMan View Post
And now you know how I feel whenever I debate with you.
The difference is that I have the results on my side.
Quote:
Originally Posted by MathMan View Post
I'm pretty sure when I talked about Gomez getting 70 points I called it a best-case scenario, which was based on the notion that he'd get 40ish points and benefit from the Habs' continually strong powerplay. Only half of that equation happened in 2009-2010. Setting up straw men by misrepresenting what I say is another favorite tactic of yours, I've noticed.
It's not a straw man argument. When we dealt for him you told me he was a lock for 70+. I asked you about declining point totals, I told you that it was very possible that he was on the decline and you dismissed it. I'm not trying to 'rub it in your face' man... I'm trying to force you to confront what you've argued on here for a long time now. These stats have not produced the results that you've continuously told us it would. I'm giving you the reasoning for my skepticism.
Quote:
Originally Posted by MathMan View Post
By the way, you didn't exactly predict a lottery finish for the Habs this year, did you? How come?
Because I thought we'd have Markov, he'd go down and we'd get a decent replacement. Didn't happen. And we were far worse than I thought we'd ever be. Think about it though man... you continuously say that I"m too hard on the club. If anything, I've overrated them. And if that's the case, doesn't it just make your opinion on the club that much more absurd? If I'm overestimating them and you feel like I've been 'doom and gloom' about them as you've argued many times, then you're really way, way off here.

Quote:
Originally Posted by MathMan View Post
Hilarious. And while you were trying to make a joke, it also tends to show that you don't really understand what an important concept like "regression to the mean" means...
I"m doing the best I can. I thought it was kind of witty but uh... enjoy the veal.

Lafleurs Guy is offline   Reply With Quote
Old
04-04-2012, 01:15 PM
  #147
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,855
vCash: 500
Quote:
Originally Posted by Miller Time View Post
"observant experience" isn't simply about a knowledge of "what has gone before", it's the expertise (from hours of dedicated practice) that allows for one to appreciate/recognize the subtle differences in performance that explain their effectiveness.
And exactly how would this "expertise" be obtained? Invariably, it will be either via direct experience or via teaching, which is the transferring of experience.

There's no way to make salient predictions about the future without matching it to something that has happened before and recognizing the same pattern. It's an amazing and complex ability, but it is something that humans do all the time, so naturally that they don't ever think about it, but it's the basis of how we learn things. In fact part of why analytics are useful is because humans are so good at seeing patterns they commonly see them where there really aren't any.

(Unless we are talking about aesthetics, which are wholly the matter of opinion and preference, and thus outside the purview of analytics -- I only mention this because you bring up examples about wine tasting and listening to music. And even then, the experienced wine taster and music expert will be able to make a more granular opinion because of what they've encountered in the past. Knowledge simply does not spring from nothing.)

MathMan is offline   Reply With Quote
Old
04-04-2012, 01:53 PM
  #148
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,855
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
I've looked at the links. I've seen the math. I haven't seen anything though that suggests that these formulas are worth anything.
Then you've either not looked enough, or else your expectations for being "worth anything" are unreasonable.

Quote:
Originally Posted by Lafleurs Guy View Post
I don't think it invalidates the model and I don't think I said this. If I did then I mispoke. I do think that those 'subtle' differences are 'subtle' because they're hard to capture if not impossible on a spreadsheet.
They are subtle because the effect is either small, or rolled up into other things. If they had a big impact, then they would be immediately noticeable and it would not be hard to model them.

Quote:
Originally Posted by Lafleurs Guy View Post
Stats will try to capture WAR in baseball and I suppose you can try to do this in hockey. There are problems with WAR to begin with but in hockey I'd think it would be even harder. I'm not sure how else you really try to capture the force of those hits though.
It is hard and yes, it is harder in hockey because context matters. But a "win share" stat is the Holy Grail, really: a full, nicely-collected, all rolled up into one estimation of player value. There are lots of stuff out there that can give us very valuable information despite not giving us one unique, all-encompassing number we can refer to.

Yes, it's complicated. This stuff isn't simple. If it were simple, we would not be having this discussion.

Quote:
Originally Posted by Lafleurs Guy View Post
See... this kind of thing is where you really lose me. I'll use a more extreme example here. Look at Gretzky. The guy goes to LA and leads them to a cup final. You think it's sillly to cite him as the principle reason they got to a cup after trading for him? You think it's coincidence that he was on that club? You don't think that one player can make a huge difference to a clubs' fortunes? Seriously... you believe this?
It's a matter of degrees. Gretzky was the best player in the history of the game and played in an era where there was less parity. Even then, by himself he was not the only factor in LA going to the finals. The world, including hockey, isn't a nice place where you can point to one thing and say "X happened because of A". It'll be 12% A, 8% B, 9% C, 4% D, and so on. And a lot of those letters will be outside the control of the team in question.

Gretzky was a factor. One of many. Pronger was a factor. One of many. Important, no doubt. Possibly each of them was even the biggest factor... but in the way that A is the biggest factor in the list above at 12%. A is the biggest thing, but there's still 88% of other stuff.

It is not because of a special quality these guys have that will make teams win that other players don't. One guy by himself will not account for 40, 50% of the ability of a team.

Superstars don't have some magical quality that make them different from other players that don't have it. They are very good players that contribute more than others.

Quote:
Originally Posted by Lafleurs Guy View Post
It's not a straw man argument. When we dealt for him you told me he was a lock for 70+.
I did no such thing. I presented that as a possible scenario, which presupposed that he'd get around 40 ESP just like previous years (which he, in fact, did) and that he'd increase his PP production from the Habs' strong PP (he didn't).


Last edited by MathMan: 04-04-2012 at 02:08 PM.
MathMan is offline   Reply With Quote
Old
04-04-2012, 03:01 PM
  #149
Cyclones Rock
Registered User
 
Join Date: Jun 2008
Posts: 1,240
vCash: 500
I think some basic and clearly stated definitions would help many of us start to try and get a handle on some of these mico stats.

For starters, a relatively precise definition of "puck possession", I would think, would be essential as this concept seems to be a cornerstone of micro stats.

A little primer on Corsi wouldn't hurt either. This seems to be the most favored analytical model of micro stats enthusiasts. I've been able to find a few articles on Corsi. But, it would be useful to get a basic definition of it and the reasons it is believed so powerful from those who find Corsi to be superior to traditional stats.

The reality is-as with any new product which seeks to replace a long used traditional product-that the new product is not going to be accepted until it is properly sold. Endorsement by a few statisticians and analysts isn't going to get the new product accepted. It's understood (by me, anyway) that a basic understanding of statistics (means, regression, standard deviations, etc.) is required to understand the metrics.

Corsi et al. shouldn't be too hard to sell if it is a much better tool than traditional stats. Not too many are going to hold on to their 8 tracks when they can use CDs.

Cyclones Rock is offline   Reply With Quote
Old
04-04-2012, 03:28 PM
  #150
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 17,746
vCash: 500
Quote:
Originally Posted by MathMan View Post
Then you've either not looked enough, or else your expectations for being "worth anything" are unreasonable.
I'm guessing that I didn't look hard enough. I think some of us were hoping that some of its supporters could show us some of this stuff. As you said, it's not really available in any kind of clean manner. I haven't seen where it's been used to predict anything either. I'm assuming that it has but I don't see where.

Quote:
Originally Posted by MathMan View Post
They are subtle because the effect is either small, or rolled up into other things. If they had a big impact, then they would be immediately noticeable and it would not be hard to model them.
I'd say it is noticeable. I'd also say it's hard to quantify. If a guy gets rocked and it changes the tempo and momentum of the game. We've seen that before. I know you'll say it's luck or coincidence but I don't agree. So I disagree with you on that.

Millertime has given you an objection that's similar to the one I gave you earlier. You seem to be dismissing and avoiding these things with the same answer... 'I can't measure it so it's not important' or 'If it hasn't been measured, it's simply not important enough to measure because analysts would've done this' and I don't think anyone buys this as satisfactory.
Quote:
Originally Posted by MathMan View Post
It is hard and yes, it is harder in hockey because context matters. But a "win share" stat is the Holy Grail, really: a full, nicely-collected, all rolled up into one estimation of player value. There are lots of stuff out there that can give us very valuable information despite not giving us one unique, all-encompassing number we can refer to.

Yes, it's complicated. This stuff isn't simple. If it were simple, we would not be having this discussion.
I think we get that. I think we get that the stuff is useful. Useful enough to solely build a team on though and forego things like grit and size as you've suggested? No. I think those attributes should still be looked at by our scouts.
Quote:
Originally Posted by MathMan View Post
It's a matter of degrees. Gretzky was the best player in the history of the game and played in an era where there was less parity. Even then, by himself he was not the only factor in LA going to the finals. The world, including hockey, isn't a nice place where you can point to one thing and say "X happened because of A". It'll be 12% A, 8% B, 9% C, 4% D, and so on. And a lot of those letters will be outside the control of the team in question.

Gretzky was a factor. One of many. Pronger was a factor. One of many. Important, no doubt. Possibly each of them was even the biggest factor... but in the way that A is the biggest factor in the list above at 12%. A is the biggest thing, but there's still 88% of other stuff.
Dude, you said it was silly to view a player as a main reason for getting to a cup. You can try to put percentages on it all you want... 12%? Okay, good luck to LA getting to the finals without Gretzky. Ditto in Edmonton without Pronger. Nobody is trying to suggest that those players could go on the ice by themselves and beat 5 other guys + a goalie by themselves. And I'd say that when Gretzky joined the Kings he represented a much bigger change than you're suggesting he did. Look at that club before he got there. Look at them afterward. Those kinds of players make a huge difference.

I have no idea why you've decided to challenge me on this point. Superstars are essential ingredients to winning cups. History has shown this over and over again. The more you have the better your chances. If you dont have any... good luck.
Quote:
Originally Posted by MathMan View Post
It is not because of a special quality these guys have that will make teams win that other players don't. One guy by himself will not account for 40, 50% of the ability of a team.

Superstars don't have some magical quality that make them different from other players that don't have it. They are very good players that contribute more than others.
Like I said, when you come up with calling it 'silly' to credit guys like Gretz or Pronger for being main reasons why their clubs make it to the finals, you lose your audience. Superstars like them very clearly do make a huge difference. That's why teams without those kinds of players rarely if ever win.

Mario Lemieux had a season where he contributed about 55-60 percent of his teams' offense. One player can make a huge difference in a teams' fortunes. Look at Roy's cups... we have NO chance of winning without him. Anyone who saw the Hartford series knows this. Ditto with those crazy overtime wins in '93.

You're trying to carve it into percentages and I guess that's a cool project to undertake. But anyone who suggests that we could've won the cup in '86 without Roy doesn't know what they're talking about. And to suggest that it's silly to assert that he directly led us to a cup is well, crazy. He was a huge reason for us winning and without him we flat out don't win.
Quote:
Originally Posted by MathMan View Post
I did no such thing. I presented that as a possible scenario, which presupposed that he'd get around 40 ESP just like previous years (which he, in fact, did) and that he'd increase his PP production from the Habs' strong PP (he didn't).
You talked about his splits and said he was a virtual lock to improve here. I repeatedly challenged you on his declinging totals and you talked about 5 on 5 etc... and said the guy was going to be great for us. I told you he's a 70 point guy at best but could easily be worse and pointed to the totals. You said that I didn't know what I was talking about, the splits would get him there. Whatever, I'm not going to argue this... it's getting away from the point I was making which was that your arguments (based on microstats) haven't really materialized for us.

Is it possible that you're allowing bias to enter your analysis of our team? Do you eliminate the stats that suggest we won't be good and only rely on the ones that suggest we'll succeed? I have to ask because every year you say we'll be better than we actually are.

Lafleurs Guy is offline   Reply With Quote
Reply

Forum Jump


Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT -5. The time now is 06:13 PM.

monitoring_string = "e4251c93e2ba248d29da988d93bf5144"
Contact Us - HFBoards - Archive - Privacy Statement - Terms of Use - Advertise - Top - AdChoices

vBulletin Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
HFBoards.com is a property of CraveOnline Media, LLC, an Evolve Media, LLC company. ©2014 All Rights Reserved.