Diary of a Madman

Someone recently told me that some of my posts assume an unrealistic degree of familiarity with some rather obscure stats. This post is intended to be a bit of a primer for other Habs fans, demonstrating how I interpret some of these numbers for the Habs. If you think this is all hogwash, you're not alone. If you see something that piques your interest, I recommend reading some archives of some of the more numbers-oriented blogs listed on your right. Especially the Oilogosphere ones.

These are 5-on-5 numbers from behindthenet.ca.

The columns are Games Played, Time on Ice per 60min (that's 5-on-5 time per 60 minutes), Quality of Competition (v. v. v. important), Goals For (by all players) while on the ice, Goals Against while on the ice, and the Corsi Index. The Corsi number is like a plus/minus for shots directed at the net by your team minus those from the opposition (shots for + missed shots for + opposition blocked shots - shots against - missed shots against - blocked shots). The number shown doesn't include the blocked shots, and it's a number per 60min.

If you want to get a handle on who the important players are and who is doing a good job given their role, these numbers are valuable. 5-on-5 hockey is at least two thirds of the game. In addition to the goals scored/allowed, events at 5-on-5 determine who gets the PPs and determines whether you make it to OT/SO. PP/PK performance is certainly valuable - probably moreso on a per-minute basis - but the bulk of the game is decided 5-on-5.

If you accept that, the next important piece is who the player in on the ice with/against. Most Habs fans are aware of the team's lines and defence pairings. Less well known is how those lines are used by Carbonneau and how they are addressed by opposing coaches. The most common misconception revolves around just who is on the checking line (hint: it's 'none of the above'). The second most common misconception is the even strength effectiveness of the Kovalev/Plekanec line. Sure, they score some at 5-on-5, just not as much as they surrender. And they do it against weaker opponents than the Koivu line. And their penalty differential is worse. Anywho, the QUAL COMP column is the important piece of the above table. If you Google 'Desjardins' and 'quality of competition,' you'll find lots of interesting discussion. It's not a perfect stat, to be sure. I find that when I compare QUALCOMP to the players each Hab regularly faces according to the shift charts and Hockeyanalysis stats, they mesh very well. YMMV. Now we have an idea of the difficulty of the players' icetime, and that provides context for the GF/GA.

Finally, the Corsi column. Obviously GF/GA are the currency of wins and losses. The problem with looking strictly at goals is sample size. A few extra saves in one end of the rink and not the other can make a good player look bad, or vice versa. Shots directed at the net are a good proxy for GF/GA, and they provide a much larger sample size. It tells you which end of the ice is seeing most of the action.

So.... what can we take from these numbers?

1. Markov and Komisarek are playing against the best players in the league. If you were to look at Hamrlik's +8, compare it to Markov's -9 and conclude "Hamrlik's been playing better defensively," you'd be wrong. It's easier to shut down Kyle Wellwood's line than it is Mats Sundin's line. There are some factors besides QUALCOMP that explain the huge +/- discrepancy between these two that I'll get to in another post.

2. Koivu, Higgins, and Ryder are (or were) also playing against the best players in the league. This is a major departure from last season, or any of the last 5+ seasons for that matter. Last year the Bonk/Johnson checking line faced the top opposition. Despite this shelter, the Higgins/Koivu/Ryder line were terrible at even strength. This season the trio has actually been in the black when playing together, and they've done it facing the top lines in the league way more often than last year. This has to be taken into consideration when looking critically at their performance.

3. Trevor Linden, Joe Juneau and Radek Bonk are all veteran players who were brought to Montreal to play a checking role. I assumed Bryan Smolinski would inherit that role this season, but it looks like Carbo was having a hard time trusting him. Judging by his +8/-16, it's understandable. Again, there are other factors at work that mitigate the +/- discrepancy. Smolinski's line has still been the usual 2nd option against tough opposition, after Koivu's. His loss due to injury is a big deal.

4. Dandenault has been awful. Just awful. It's shocking that he's been given so much responsibility, being a frequent winger for Smolinski (a big factor in Smolinski's +8/-16). Dandy's been shuffled from line to line and he's getting blown away everywhere.

5. Hamrlik and Brisebois have been the second pairing, and they've done well. Hamrlik is a vast upgrade over Souray, 5-on-5.

6. Kastsitsyn/Plekanec/Kovalev have performed no better than Koivu's line, 5-on-5. And they're facing mediocre opponents. There have been some recent games where the opposing coach focused on the Kovalev line rather than the Koivu line but that's only because of the bizarre decision to put guys like Streit or Kostopoulos or Dandenaullt on Koivu's wing. For the most part, the Kovy line has been getting plum icetime and not doing as much with it as we would hope. Same as last season.

7. Mark Streit looks bad in terms of GF/GA, especially considering his 3rd pairing status. However, his +8/-17 doesn't seem to jive with his Corsi number of -1.7. Shooting luck is a huge factor here (again, I'll discuss in a later post). He's been good.

8. Surprise, Chipchura is not this team's checking line centre.

9. Kostopoulos has been getting 4th line minutes and he's one of four regular forwards with Corsi numbers on the right side of zero. Maybe it's time to give him a little more responsibility.

10. Gorges has been getting a free ride from fans/media. He's been the worst defenseman while facing the softest opposition. With nine frigging defensemen on the roster, this guy should not be playing.

Conclusions and Contradictions.

With the hindsight of 29 games this season Brisebois has been OK, but the decision to bring him to town looks like a bad one.

Immediately after the offseason FA frenzy, The team had seven defensemen: Markov, Komi, Hamrlik, Streit, Dandenault, Bouillon and Gorges, plus the guys on the farm. Up front they had lost the entire checking line of Perezhogin/Bonk/Johnson, and replaced them with Smolinski and Kostopoulos. Now, Smolinski is a versatile player but had never in his lengthy career (to my knowledge) demonstrated that he could play a role like Bonk's. Kostopoulos was a 4th liner in LA. Even if you give Gainey the benefit of the doubt with Smolinski (maybe he saw something that indicated he could be a shutdown guy), it seems clear that there were holes in the bottom six forward spots.

As the summer wore on and Bonk and Johnson remained unsigned, Gainey picked up a retreaded Patrice Brisebois for $700k. So now we're up to eight defensemen and ten forwards. Interesting. That, in a nutshell, is why we are now seeing Streit and Dandenault playing on the wings.

I just don't get it. Why sign Brisebois when Dandenault is an adequate 3rd pairing RD, and then bump Dandenault up to the wing? Why not skip the middle man and just sign a competant RW?? I would have been happy to see Bonk back - he signed for a cap hit of $1.5MM over two years in Nashville. Gainey must not have wanted to commit to that 2nd year. Tangent time: Speaking of Bonk, it looks like he was the one in the driver's seat last season - not Johnson. Bonk's doing for the Preds what he did last year for the Habs. Johnson's in St. Louis and he's been used in a similar role too but with ugly results.

So. Here we are with the top two lines playing tougher roles but coping pretty well, and a weak bottom six. They were keeping things close to even at even strength, and doing great on the PP. So what does Carbo do? He loses patience with Ryder's stagnant counting numbers and screws up the top line. And he continues to give significant icetime to Dandenault, Begin, Latendresse, Chipchura - all bottom six forwards who are getting beat up regularly. There are deeper teams than Montreal who are limiting their 4th liners to significantly less icetime. It's puzzling.

Labels: , ,


Blogger Topham said...

Nice analysis. Very nice. I know there is more to say, but I would have to have a proper look at your stats, or risk looking like a fool.

I'll be linking you up on my blog (lionsinwinter.blogspot.com) if that's alright with you, now that you're back to posting on our team.

I'll have another look at the stats too and get back with a more useful comment...

12/12/2007 11:45 a.m.  
Blogger Kaz said...

Thanks Jeff. This definitely helps for those of us in the math-challenged category.

I think it was Mathman who was concerned about the difference between shots and shot quality. Of course "shot quality" is difficult to quantify. But what about looking at shots that actually make it on net? A decent substitute for shot quality?

Also, are there other reasons to look at shots instead of GF/GA besides sample size? If it's just sample size, then the end of the season ought to provide a definitive (if late) analysis.

Given all these stats, and from your previous posts, I think that we do have a checking line: Koivu's. And it looks like they were doing reasonably well too, at least with Ryder (even more surprising). So maybe Carbo was onto something, ie, none of the other "talent" in his lineup is good enough to take on the opposition's top line. Now that's depressing.

12/12/2007 2:57 p.m.  
Anonymous MathMan said...

I have a bit of trouble with this all, really. Not that there isn't a lot of interesting data to be found in it (I didn't realize that Dandy had such a shot differential, for example) but IMHO you push the interpretation much further than the value of the numbers allow.

I believe that statistics derived from shots on goal are by nature suspect, especially when they do not take shot quality into account. It's a larger sample size, but not all shots are created equal. There's too much context involved in shots on goal to make just the raw addition meaningful.

For example, there's the situational aspect. When the Habs (and I suspect other teams) blow someone out, typically they get badly outshot by the end of the game as they start to slow down, dump the puck in the opposing zone, hang on to it, and generally play defense, while the other team tries the patented "pray-and-shoot" method to try to score a few quick goals. Latest example of this phenomenon for Montreal was the Islanders game, where the Isles generated 40 shots but hardly any real scoring chances and richly deserved their 4-1 loss. And I've lost count of the times Montreal generated a respectable number of shots only to have it called, fairly or unfairly, something along the lines of "the easiest set of shots goalie X has ever faced"...

It just seems to me that the Corsi number's value appears dubious when considered in isolation. A guy who takes tons of low-percentage shots and/or misses the net regularly (eg. Ryder lately) will artificially increase his Corsi number while not generating a whole lot of offense or value, whereas a defenseman who concentrates on blocking shots (like, say, Komisarek) rather than deny them will appear to give up more offensive chances than he actually does. In this light, it seems like little more than a mildly interesting metric.

Sure, these stats make Ryder look good, but the game film shows that he's taking low-percentage shots when he should pass, and dekes or passes or loses the puck when he should simply take a high-percentage shot. It's not that he's turned into suck, but he's definitely in a slump and no, it is not only puck luck.

Likewise about Smolinski -- he's made a number of unseemly and costly defensive blunders that do not have anything to do with Dandenault. I'm not convinced losing him will hurt all that much defensively; he's just not played up to expectations.

12/12/2007 3:49 p.m.  
Blogger Jeff J said...

topham: Didn't know about your blog; I'll link it up ASAP.

kaz: "what about looking at shots that actually make it on net? A decent substitute for shot quality?"
mathman: "...not all shots are created equal.

Yes, you're absolutely right. Shot quality is a factor. Consensus is that it's just not a huge one. If a player gets outshot 2-1 for a game, it's generally a bad game for that player and his mates. There just aren't five man units in the league who can boast twice the shooting accuracy of other five man units. Of course a breakaway is a better shot than a dump in on goal from the red line. By and large, these things balance out over the long term. Those "40 easy shot" efforts happen on both sides. The situational differences happen to both sides.

For every criticism of shots as a metric, there are darn near as many problems regarding goals - and goals are ten times more rare. Not all goals are created equal. What about cheap goals late in a blowout against a backup goalie? What about empty netters? 100' floaters over Tommy Salo's head?

This post is a good read regarding the Corsi number. Vic's comment here addresses the issue of shots vs. Corsi vs. the 'Fenwick number' (Corsi minus the blocked shots). The correlations are high - very high. It's a good idea to take the bigger sample size.

kaz: Yes, I suppose you could call Koivu's line the checking line. Recent NHL history suggests most coaches go strength vs. strength. Alfredsson and Sundin go head-to-head rather than trading turns against the other team's checkers. I have a hunch that Carbonneau would prefer to use a dedicated checking line but doesn't feel he's been given the players to assemble one.

mathman: "Sure, these stats make Ryder look good, but the game film shows that he's taking low-percentage shots when he should pass..."

That's cool. You're not alone in your judgement of Ryder's game. You're in good company because an entire NHL coaching staff agrees with you.

12/12/2007 4:41 p.m.  
Blogger Kaz said...

I'd rather have Ryder shoot. He's not a good passer, and worse at carrying the puck. He needs to be with linemates that can do both. I'd like to see him and Kostitsyn swap places, and see how Ryder does on the left wing in 5-on-5 situations, just as he does on the PP. Playing on Plekanec's line would reduce his defensive responsibilities too.

I, too, noticed that trend of top lines playing against other team's top lines. Best-defense-is-a good-offense theory I guess. The problem is that the Habs' don't have a good enough offense. We don't have a line that can compare to Alfredsson-Spezza-Heatley. The more conventional hockey wisdom of matching lines is more applicable to team's like the Habs: match your top lines against weaker lines, and send out a shutdown unit against their top players. But I guess the Habs don't have such a unit, so even conventional wisdom doesn't apply.

Quite the conundrum. Carbo could do better (esp in dealing with Ryder), but I kind of feel sorry for him now.

12/12/2007 5:02 p.m.  
Anonymous MathMan said...

About Ryder: I do think the coaches need to shake him out of his funk somehow, because Ryder needs to do something different than "be luckier", IMO. One game in the stands, maybe a one-game demotion to try to shake him up.

However, I'm not saying I agree with swapping him long-term with Smolinski, Streit, then Dandenault on Koivu's wing. He wasn't hurting the team defensively (as you quite rightly point out) and he was getting enough icetime to possibly shake his slump. I think trying out Latendresse was worthwhile (because Lats was potting goals) but the other choices are very poor substitutes.

Basically I think Ryder's not so much playing badly as he's not playing the way that made him successful. His tentativeness is at least partly responsible for his drop in shooting percentage.

12/12/2007 6:46 p.m.  
Blogger Hawerchuk said...

This comment has been removed by the author.

12/13/2007 1:36 a.m.  
Blogger Hawerchuk said...

Complain and ye shall find:

Canadiens shooting detail

I've got a database of pretty much everything and as long as you don't ask me to do something that's just a simple database query, I can probably turn it around in a few days...

12/13/2007 1:37 a.m.  
Blogger Jeff J said...

Thanks Hawerchuk, that is fantastic.

Everything, huh? Sounds tempting, but there is already enough information available to keep a duffer like me busy for awhile.

12/13/2007 10:56 a.m.  
Blogger Jeff J said...

I took a quick & dirty look at forwards with between 45 and 50 shots to get an idea of how Ryder stacks up.

The group miss:shot ratio is 0.39. Ryder is 0.32. He misses less often than the others.

For the group, the shot type breakdown is roughly 20% slapshots, 20% snapshots, 40% wristshots, 20% other*. Ryder is pretty much the same.

Average shot distances are 44.7 for slapshots, 34.1 for snapshots and 31.8 for wristshots. For Ryder the numbers are 48.8, 33.3 and 28.7. The higher slapshot distance doesn't carry a lot of weight. For the group, 10% of goals are scored off slapshots, 20% off snapshots, 45% off wristshots, 25% off other. Ryder's shot distances are above average.

There is nothing in the table to suggest Ryder's shot quality is any worse than the group.

The group shot% (including misses) is 8.3%. Ryder is 3.1%.

These numbers include his recent service in the bottom six, where you can't expect his chances to be as good.

* - Not sure what the 'other' shots are. Deflections? Backhands? Dekes? Bounces off Bryan McCabe's ass? All of the above?

12/13/2007 11:04 a.m.  
Anonymous MathMan said...

Interesting, very very nice statistics. I find it particularly meaningful that Ryder doesn't miss the net as much as other shooters. Maybe I listen to Yvon too much.

Ryder may have just been unlucky after all, though I still think *some* part of his low shooting percentage is due to tentativeness and lack of confidence. Quite possibly induced by said lack of puck luck.

But I think that we can agree that he ain't gonna get scoring if he keeps getting scratched. Like, apparently, he is going to be tonight. Grumble.

Other shots: backhands and tip-ins, I think is what the NHL records, but Hawerchuk would know more than I.

12/13/2007 1:09 p.m.  
Blogger Hawerchuk said...

The other shots are backhands, tip-ins and wrap-arounds, and also, new for this year, deflections that miss the net. Not much to see there, and I imagine the results would be confusing because lots of scorers call a wrap-around a backhand, etc...

12/13/2007 3:16 p.m.  

Post a Comment

<< Home