Monday, December 03, 2007

Things That Surprise Me

Over on Lowetide's San Fernando game day thread I asked why so many people are so ready to see Geoff Sanderson sit as soon as Edmonton can ice a healthy line up. Thankfully, I didn't word my question very well and the rush of responses mostly dealt with explanations surrounding his current injury (which sounds pretty painful btw) and I've had the opportunity to do my own digging.












Posted here are numbers taken straight off of behindthenet.ca and they include Oilers who have played at least 20 games this season. The only stat that I have added is the final column, "net effect", which shows you the difference between a player's on ice and off ice +/- per 60 ES minutes. Immediately this shows that Sanderson is 3rd worst among Oiler regulars in this regard but after scanning up and down, #8 has become the least of my surprises.

Jarret Stoll has looked to be a shell of a man at times so far this year. He's gotten some puck luck of late and a few points have followed and in addition, it's always tough to fault him when it comes to effort regardless of how well he is actually playing. All this said, I hope that he either a) recovers to the half decent ES powerplay specialist he used to be or b) there is a GM out there somewhere who hasn't noticed #16 is barely better than our 18 year old junior kid in terms of results this year.

In the WTF category on the positive side of things, or HTF as in how the fuck are Dustin Penner and Raffi Torres near the top of this list? Seriously, I am a bigger fan of DP than I suspect most on the blogosphere would admit but how is he #2 to only Shawn Horcoff in this regard? Maybe a quick peek at Vic's location of faceoff thread might help give context to this but it's just odd to see so many players above Ales Hemsky.

Props are to Tommy G who is a surprise in how much he is kicking Mr. Steve Staios' ass so far this year... but it's time for the single biggest surprise of all on this list: How is the difference in quality of competition between Horc/Hemmer vs., say, Sam Gagner, so small? Is it possible that Edmonton is so bad this year that opposing coaches (almost) don't give a shit who plays who against us? Or scarier yet, is it possible that the Oilers are so bad that MacT is the one not giving a shit?

26 Comments:

Blogger rananda said...

im not sure why you dont recognize that the first order qual comp limitation is significantly skewing that statistic. do you really think that alexeev, fedotenko, perrin, tarnasky all played higher competition than lecavalier last year? do you think that callahan, hollweg, dubinsky, and orr played better competition than jagr. havent examined the 07/08 numbers (dont seem to be as easily sortable as last year's), but im not sure that (outside of those players that only go up against top scoring lines, the maddens, pahlssons) those numbers are helpful at all. petr prucha, as much as i love him, does not play against better players than jaromir jagr. i suspect this is what is making it look like hemsky and and gagner are playing against the same people.

12/03/2007 11:44 am  
Blogger Hawerchuk said...

I think the 07-08 numbers are pretty easily sortable - just click on the column heading.

Jagr has a higher Qual Comp than Prucha this year, btw. But if you look at the game-by-game logs, you can see that Prucha has had tougher opponents in several games - games where he lined up against Ovechkin and Crosby all the time, and Jagr didn't.

12/03/2007 2:10 pm  
Blogger Mr DeBakey said...

Let’s see

Fedotenko Played The Most ES Against,
Forwards – Ovechkin, Kovalchuk, Olesz, Jokinen, Laich
Defence - De Vries, Salei, Van Ryn

Alexeev Played The Most ES Against,
Forwards – Walker, Staal, Ovalchuk, S Kozlov, Ovechkin
Defence – Eminger, Commodore, Vishnevski

Lecavilier Played The Most ES Against,
Forwards – Pandolfo, Horton, Gordon, Brylin, Whitney
Defence – Morrisonn, Bouwmeester, Havelid

That the top 5 forwards in order, and
the top 3 dmen in order

Fedotenko played pretty tough minutes

12/03/2007 3:10 pm  
Blogger Lord Bob said...

The numbers guys tell us that we should ignore the deceiving evidence of our eyes and concentrate on math. Therefore, Penner for MVP!

(I do think that he is, to no small extent, a victim of the "huge guys always look slow" syndrome we're familiar with, and because he's a battleship he stands out on the ice. I like Penner. But, still, WTF?)

12/03/2007 3:48 pm  
Blogger Hawerchuk said...

This comment has been removed by the author.

12/03/2007 11:27 pm  
Blogger Hawerchuk said...

I think the 'numbers guys' say "Penner looks like he sucks. Why don't his stats suck too?"

But look at Game 3 at Detroit, where Penner was even in 13:30 against these guys:

NICKLASLIDSTROM D 89.8 %
PAVELDATSYUK C 79.5 %
HENRIKZETTERBERG L 78.1 %
BRIANRAFALSKI D 77.3 %
TOMASHOLMSTROM L 64.0 %

Or game 8, where he was +1 against these guys in 11:15 of ES time:

JAROMEIGINLA R 67.0
CRAIGCONROY C 64.9
ALEXTANGUAY L 61.5
DIONPHANEUF D 53.8
ADRIANAUCOIN D 52.6

Game 18, where he was even in 13:18:

MATTIASOHLUND D 62.2
HENRIKSEDIN C 56.6
MARKUSNASLUND L 55.9
DANIELSEDIN L 52.3
MIKEWEAVER D 48.5

And Game 20, also even in 13:57:

ALEXTANGUAY L 56.3
JAROMEIGINLA R 54.1
CRAIGCONROY C 53.6
ROBYNREGEHR D 47.2
CORYSARICH D 46.7

The first of these three were on the road, so opposing coaches got what they wanted - Penner - and it didn't work out for them. Penner was out there with Horcoff most of the time and either Hemsky, Nilsson or Gagner. Maybe it's all Horcoff?!

12/03/2007 11:28 pm  
Blogger rananda said...

hawerchuck, you can view each team's players on a single page for 06/07, and then click on a diff team to take you to their numbers. i didnt see a place to do that team by team for 07/08. sortable wasnt the right word. easier to view for me is more accurate. if there's a way to do this for 07/08 that i am not seeing, apologies.

My eyes tell me jagr plays against either the other teams top offensive players or the checking line and always against the top d pair. the stats show him playing against ovechkin's line against wash this season and morrison-jurcina (the shut down pair under hanlon) and against talbot (pitt's shutdown center) against pitt. the seems about right, and it's not really any different for any of the other big guys (the lecavaliers, crosbys, spezzas, ovechkins). im not sure it would be easier for jagr to play against talbot, whose only job that game is to stop him) than crosby. jagr, to my eyes, seems to better against the other big guys as opposed to the checkers. this is why the first order limitation is so significant, crosby looks like a much tougher opponent than talbot, but im not sure the difference is that big. this is why the 06/07 numbers make no sense to me for a lot of players. the rangers are the team i watched the most last year so it's easiest for me to think there's something wrong by looking at their numbers.

another problem beyond the first order limitation: some players play more or less depending on how the game is going. this touches on a point tom benjamin made in a post on these type of statistics a while back, but he was basically saying that ohlund would play a lot more when the canucks were ahead and jovo when they were behind. jovo had to take risks and consequently his numbers could suffer. similarly, i think most coaches get their big offensive guys out there when theyre down and are more likely to roll 4 when ahead and everyone's playing well. well, if you assume that it's easier to score/defend when youre ahead and harder to score/defend when youre behind, then i think youre going to get skewed numbers. the big guys on ice/off ice differential is going to look worse than it should. my guess is that this effect is more than insignificant. there's prolly other stuff out there as well, i dont think the qual comp numbers are airtight yet, or if they will ever be. i would take the eyes and brain of an intelligent and knowledgeable person who has watched all of a team's games over the numbers all the time right now. put it this way, dont you think mac t has a better idea of hemsky's qual comp compared to gagner's. if his assessment matches the numbers, fine, but if they disagree, who would you believe?

12/04/2007 2:42 pm  
Blogger Hawerchuk said...

First, "quality of competition" is defined solely in terms of a player's +/- relative to his team. So the checking line will never look like tough competition using this definition. Clearly, you could compute it another way if you wanted.

I think you're expecting a lot out of a simple stat, though. Look at another simple stat: Even-Strength PPG per 60 minutes TOI. It is a valuable thing to look at, perhaps the most valuable simple stat we have, and yet it does not capture any of what you mentioned.

I guess what I'm saying is that I understand that this is an imperfect stat, and there's a whole lot of error built into it, but so what? If it's total garbage (like most of the ideas I came up with before) then I'll toss it...So far it has been valuable despite its limitations.

12/04/2007 3:19 pm  
Blogger Hawerchuk said...

anyways, I think this was what you wanted - sortable stats by team. Here's Edmonton:

http://www.behindthenet.ca/2007/5_on_5.php?sort=8&mingp=&mintoi=&team=EDM

12/04/2007 4:26 pm  
Blogger rananda said...

right, i think my problem is not with the stat but with my perception of how it's being used here and in other places, which is basically, this stat tells us how good the people a player plays against are. look at how it's being used by showerhead in the original post, which is basically asking, are coaches really playing the same quality guys against horcoff as gagner? the answer should obviously be "hell no, that's not what the stat is telling us." part of this misperception comes from the name of the stat, "quality of competition." as you point out, it's really only giving you the relative +/- of the opponent, which certainly means something, but cannot be confused with pure quality. i think it's an important distinction and one that needs to be made as i suspect more and more people are going to hop on this train and go to your site and start using this speak.

i think the way in which the stat is most helpful, is it tell us who the pahllsons and maddens are, who plays against the big boys every night. that's nice. but im not sure that really is the same as quality of competition. obv playing sammi and john are no slouches themselves (tho with a diff skill set) so playing against them has to count for something. and w the qual com stat, playing against them doesnt. that's the rub.

i thought you implied over on mudcrutch that more iterations and using the elo chess model or whatever could solve this problem. your statement here and me thinking about this more makes me think that it wont. which is fine, it's a simple stat, a good stat, but let's not use it as a metric to identify who plays against the good players. if that's not how everyone is using it, apologies, but that's how ive read it in some places and i could see it getting worse.

12/04/2007 4:54 pm  
Blogger voxel said...

When Penner is on ice he's like Torres + Pisani last year... the opposition ceases to score at the normal rates (using the ESGA metric)

It's not really surprising to me, because this has happened since the beginning of the season. But the reason why... I can't say because Nilsson is also a low GA bleeder currently (may change after he plays more games) whereas Pouliot is just a heart-attack (awful).

Is Penner being played mainly in O-zone draws and soft ES situations? Horcoff we know - isn't. Constantly taking draws in your own zone is a death sentence to your +/- in the long run.

12/04/2007 5:13 pm  
Blogger Dennis said...

Well, SH, anytime 89 plays with 10 at home, or at least before the 14-16-46 checking line was erected, that meant he was playing tough min.

Plus, there was also a stretch, the home loss to Cgy was one of the games, where MacT had 13-89 together and Keenan and other guys were going nuts with their line rotations just to get their big guys out against our kids. And, yes, sometimes MacT's had a chance to get our kids away from the top lines and he hasn't done it. I can't remember the opponent but there was one game where the Oilers had 89-13 taking an own zone draw after a bloody TV timeout, for fucksakes. So I think there's a little bit of seeing what he's got in the youngsters.

How 27 is racking up these numbers, I don't know. I don't know how slow he really is but he isn't fast either. I'll just guess and say he's grabbing those numbers though because he's a bear on the forecheck and he wears out time in the O zone.

12/04/2007 5:58 pm  
Blogger Showerhead said...

If you make comparisons between the highest and lowest quality of competition rankings on each team, you'll find that the overwhelming majority of them have greater differences between their top and bottom player. I haven't made any assertions that the stat is flawless but have pointed out an observation, however black and white I make a grey situation read. My question is absolutely not "are coaches seriously playing the same players against Horcoff as Gagner" (though, as Dennis points out, those two particular players have been linemates from time to time). My question is "my eye and these statistics suggest that there is a chance other coaches aren't maximizing their ability to match lines against Edmonton - why is this?". I definitely prefer to post in a black and white manner if it provokes responses but let's not stray from the true intentions of my post.

My understanding is that the quality of competition stat does a good job of pointing out who plays against opposition posting high ES +/- relative to their team. These types of players are going to be your Crosbys, your Lecavaliers, etc.. but they are also going to be your Getzlafs, Cammalleris, etc (the examples I'm aiming for are kids who are successfully making hay against lesserlites) moreso than Pahlssons or Madden. I am cool with this.

12/04/2007 6:57 pm  
Blogger Jeff J said...

rananda: "it's a simple stat, a good stat, but let's not use it as a metric to identify who plays against the good players."

I pore over every shift chart for the Canadiens, and to my eye the qual comp stat is exactly right for them. This is two years running. Because of the +/- thing, you can't trust it on a game-to-game basis, but long term it seems to sort itself out within a team. I wouldn't use it to compare the level of competition for players on different teams.

12/04/2007 7:04 pm  
Blogger RiversQ said...

But the reason why... I can't say because Nilsson is also a low GA bleeder currently (may change after he plays more games) whereas Pouliot is just a heart-attack (awful).

Sure Pouliot was bad in 9 games but he had the lowest GA/hr rate on the team last year. (And by last year, I mean before Smyth was traded)

If I had the choice between Pouliot and Nilsson, I take Pouliot every time. Nine bad games and a handful of GA doesn't outweigh 46 games of solid hockey last year for a young player.

12/04/2007 11:27 pm  
Blogger RiversQ said...

Jeff J said...

I pore over every shift chart for the Canadiens, and to my eye the qual comp stat is exactly right for them. This is two years running. Because of the +/- thing, you can't trust it on a game-to-game basis, but long term it seems to sort itself out within a team. I wouldn't use it to compare the level of competition for players on different teams.


Sure it does. That's also a team that has gone strictly towards the checking line strategy which is part of the basic assumption made by Desjardins.

Now the Oiler numbers for this year?

Not enough GP yet to worry about it and what Dennis said.

Showerhead: I agree that this gives a ballpark measure, but I have found that under close scrutiny it often doesn't hold up, especially with this sample size. (Although I do agree with jeffj's assertion that it's generally better within a team than between different teams.)

12/04/2007 11:30 pm  
Blogger Vic Ferrari said...

rananda said:

another problem beyond the first order limitation: some players play more or less depending on how the game is going. this touches on a point tom benjamin made in a post on these type of statistics a while back, but he was basically saying that ohlund would play a lot more when the canucks were ahead and jovo when they were behind. jovo had to take risks and consequently his numbers could suffer.


Do you have a link to this rananda? The reasoning seems upside down to me. If there was evidence to support this claim I'd like to see it, surely it would just be a one-off for a player in one year, no? Run the same script for different years and you'd see the opposite effect on Jovo's numbers I'm sure.

First off, I suspect that the increase in icetime of Jovo over Ohlund, relative to score, was marginal, it would be the 3rd D pair that would see more pine.

Mostly though, this is exactly the sort of icetime that improves a player's numbers, so it is a poor rationale.

Most teams, if not all, play to the score (as Moreau would say, a soccer term really), meaning that if they have the lead in the third the play more conservatively, try to avoid taking penalties, are happy to keep the play to the outside at both ends of the rink, never shoot far post from outside the dot, etc.

Now of course playing uber-conservatively slows the game down in terms of shots and chances of course, that's the point. There is less likely to be a goal in the next ten minutes, so to speak. But it also means that the trailing team is a shade more likely to get the goal if there is one. Because the leading team is willing to sacrifice some out-chancing in order to cut down the overall level of scoring chances. We see that in almost every game we watch.

So if we pick on Ken "The Thrill" Hitchcock ... I think it's a safe bet to say that his teams underperform expectations at outscoring when they are leading by two goals in the last half of the game. But that they overperform expectations in terms of shot and goal prevention in that situation.

It's really obvious in soccer, if two teams are evenly matched, then inevitably the trailing team will have the better of the possession, and usually by a wide margin.

12/05/2007 12:31 am  
Blogger Vic Ferrari said...

SH:

Back on topic, the guy who has surprised me is Nilson. He has had good linemates, and is the first choice to be subbed off for a centre for an own zone draw. Still, results be results, and Nilson has good underlying even strength numbers as well (Shots +/-, Fenwick +/- and Corsi +/-).

I mean I didn't like the Penner deal, and the first rounder given up, but he looks like a guy who is driving possession and chances. Nilson, to my eye, just doesn't. The play dies with him too much in the offensive end imo. But he keeps getting it done, with only the occasional glaring error. I don't think he's ever going to be a difference maker or anything, just doesn't bring enough offense, but he may very well be a useful NHLer.

12/05/2007 12:32 pm  
Blogger rananda said...

here's the link to tom's old post re jovo and ohlund. there's no evidence to support the assertion re the two but it makes sense to me.

http://www.canuckscorner.com/weblog/nhllog/archives/2005/09/on_the_stats_ag.html

my point is slightly different. it's that on those nights when the team is going (maybe theyve been sitting at home for 4 days and the opponent flew in on a red-eye from dallas), the team is likely to jump to a quik lead and then roll 4 lines the rest of the way. the team beats up on a tired opponent and the 3rd and 4th liners pot goals, and their relative +/- takes a jump. conversely, in the opposite scenario, the coach has to ride the big boys to get even, but they give up two late ones when pressing (like jagr's v car two nights ago). consequently, their relative +/- takes a hit because of the situation guys like that have to play in (when trailing and when playing a superior opponent that night). this is a subtle effect, but that one that skews relative +/- and qual com. that said, i think the first order limitation probably has a greater effect than this situational (who do you play when) concern.

12/05/2007 2:01 pm  
Blogger Hawerchuk said...

I think that assertion needs to be quantified. You could do it by comparing actual +/- results from the game database to what you get by simulating a game with a constant probability of goal-scoring. I don't imagine it's a huge effect.

12/05/2007 2:24 pm  
Blogger MikeP said...

Penner doesn't surprise me at all. I looked at his numbers a while ago, but only at evens, and he was flattered by them then. He was on the PK against LA, and judging by his evens play and his PK against LA, I don't see any reason MacT shouldn't give him an extended audition on the second unit. I think he's a much better player than he's gotten credit for.

12/05/2007 6:56 pm  
Blogger Vic Ferrari said...

rananda:

Thanks for the link. The reasoning from Tom Benjamin there is terrible. Firstly, as Hawerchuk guesses, the effect on overall results is minimal. Secondly, the effect, though minor in terms of individual players, skews things a smidge in the opposite direction of what his eyes are telling him. For the simple reason that I've stated above. Cosh's comment smacks of simple reason, Tyler's probably would have as well, but it was too long to read.

Overall, I think the problem you are having with the behindthenet.ca stats is that you are looking at the wrong things. I think people are gravitating to the player ratings and expecting them to be a magic bullet. Am I wrong? I don't think that Gabe is advertising it that way, otherwise he wouldn't go to the trouble of providing us with the underlying data.

If you look at Jagr's QUALCOMP rating it's actually one of the highest on the team, so that supports your assertion. And if you scroll across you see that his Corsi number is terrific (that's actually a Fenwick number at behindthenet, but it's near enough the same thing. You see that the thing that's killing Jagr's EV+ rate this year is the Rangers' abysmal shooting percentage while he is on the ice. At a quick glance, it looks like you'd expect him to have nearly double the + rate. That's just shit happening, all the smart money says that Jagr leads the Rangers at evens in terms of outshooting the opposition, and in terms of EV shooting percentage while he is on the ice (I'd say 9%ish from here on out). Don't sweat the bounces, it's still early enough that quite a few guys have numbers on the sheet that don't match their play on the ice.

Now obviously everyone on the Rangers benefits in terms of EV- rate because they have Lundqvist behind them. And Jagr gets sent out there for a crazy number of offensive zone draws and relatively few own zone ones. So he's eating up a lot of the opportunity that would otherwise fall to other wingers. Drury is at the opposite end of that spectrum, just edging out Malik. Presumably that pattern holds true fro changes on the fly as well, that Renney is trying to get Jagr over the boards on the fly when the puck is headed north as well. Therrien did the same with Crosby in the third last night(Odd that McGuire never pointed that out). In any case that's simply not the kind of usage that hurts a player's numbers, in the long run the opposite in fact.

12/06/2007 1:23 pm  
Blogger Tyler said...

Therrien did the same with Crosby in the third last night(Odd that McGuire never pointed that out).

McGuire doesn't like to go on about Crosby too much. He prefers the understated approach.

12/06/2007 4:55 pm  
Blogger voxel said...

Sure Pouliot was bad in 9 games but he had the lowest GA/hr rate on the team last year. (And by last year, I mean before Smyth was traded)

If I had the choice between Pouliot and Nilsson, I take Pouliot every time. Nine bad games and a handful of GA doesn't outweigh 46 games of solid hockey last year for a young player.


Until Nilsson has his bad 5-9 games (the probability is very high), I'm going with the player who currently produces and not the one with "unreached potential."

Pouliot's play has regressed and that is a big fat warning sign: Bust. It wasn't even 9 games... all through TC also. So far Oiler pre-season scoring has correctly hinted who would be contributors in the regular season.

12/07/2007 3:37 am  
Blogger PunjabiOil said...

While Pouliot could end up as a useful 3rd line center, it's becoming quite clear he doesn't have the 2nd line upside. Nilsson is more dynamic, and I believe will end up as a better pro hockey player.

By now though, we can rest the Parise vs Pouliot debate. It's clear who the better hockey player is by now.

12/09/2007 1:17 pm  
Blogger rananda said...

vic:

the problem i am having is that people are looking at the qual com stat using it to identify who plays "the toughest opposition" (quoting your post today on mudcrutch), when really, that stat tells us who plays against those players with the best relative +/- on their teams. these are not the same thing, as evidenced by jordan staal having a higher qual com than crosby.

12/10/2007 9:11 pm  

Post a Comment

<< Home