• Wish I believed like they do, Pt. I

    by  • October 30, 2008 • Uncategorized • 13 Comments

    I was trying for a while before the season to put together a post explaining why it is that I’m so pessimistic about the Oilers’ chances this year and struggling to do so. It just kept getting too big and unwieldy and then the baseball playoffs and then next thing I know, we’re eight games into the season. I kind of figured that it might be better to break it down into bite sized pieces.

    One of the high points of the Oilers run towards the end of the year was the PP. Souray came back from injury, Gagner got some games under his belt…things started rolling. I’ve seen detailed arguments about how much better the PP was and that it can be directly tied to the development of the youngsters and Souray’s return. Guys like Lowetide, who is a very smart observer of the game, are suggesting that “…this team…should punish opponents with a splendid powerplay.” In fairness to him, he isn’t explicitly basing this on the results achieved at the end of last year, but, and I’m sure that he’ll correct me if I’m wrong, I’m guessing that it factors into things a little bit.

    The Oilers really were a mixed bag on the PP last year. Through their first 42 games, they were scoring 4.5 PPG/60. In their final 40 games, they scored 7.9 PPG/60. 4.5 PPG/60 would leave you in last over the course of an entire season in terms of PP efficiency, without even really being within hailing distance of the team in 29th, St. Louis, which finished with 5.1 PPG/60. 7.9 PPG/60 would mean that you’ve got one of the best PP in the NHL, behind only the utterly ridiculous Montreal Canadiens at 9.0 PPG/60, the surprising Philadelphia Flyers at 8.4 PPG/60 and the Stanley Cup Champion Detroit Red Wings, at 8.0 PPG/60. It really was quite a ridiculous swing, going from an horrific PP to one of the best in the NHL.

    What troubles me, in terms of deciding whether or not I think that the gains were real, is the manner in which the change came about. Through 42 games, the Oilers were taking 37.7 PPS/60, shooting 11.9%. In their final games, they took 41.3 PPS/60 and shot 19.2%. It was completely a shooting percentage driven jump. It will surprise nobody, I’m sure, to learn that the Oilers PPS/60 through the first 42 games was good for last place, as was their PPS/60 in their final 40 games. Their shooting percentage on the PP through their first 42 games would have put them 28th over the course of an entire season; their shooting percentage over their final 40 games would have led the league, handily beating out Philadelphia’s 17.1%.

    In light of the above, I think it’s clear that in order for one to reasonably expect the Oilers PP to be a splendid one that lights up opponents’ penalty kills, one has to believe either of three things to be true: a) a low shot volume/high shooting percentage PP can be sustained over the long haul, b) if not, that it’s realistic to expect the Oilers to get luckier than Barry Fraser between 1979-1983 again or c) the combination of the players that they’ve brought in and the development/meshing of the players who were already in Edmonton will be such as to remedy the underlying problems that afflict the PP.

    I’ve assembled four years of data to poke around – 2003-04, 2005-06, 2006-07, 2007-08. What I’ve done, for the sake of poking around a little bit, is split the data into halves. So, in addition to the totals for each season, I’ve also got the numbers for each half. I’m trying to just come up with some general principles here that emerge from the years over which we have this data. This is a subject that I think a lot can be done with, something that I might come back to a few times over the course of the season, because there’s a lot of stuff that you can do as far as digging into this sort of stuff. If anyone wants the data, let me know and I’ll figure out a way to put the spreadsheet up.


    Anyway, for this post, some general stuff. I’m tossing up a graph of how the thirty power plays have done over the past four seasons just for kicks and so we can all revel in how terrible the Oilers PP has been. I’m a MacT supporter, as I’ve said time and time again, but if he’s going to be fired at some point over the next season, the stink emanating from the PP ought to be the reason, not a failure to build on last year’s 88 points. As an aside, it’s kind of interesting how the PP shooting percentages drift into a pretty tight band – they’re all basically within two percntage points of the mean over that period.

    The Oilers inability to generate shots really is a bad joke. MacT is a smart guy – I’d like to hear him explain what the hell he perceives the problem to be here. I’ve probably watched 70% of their games over the past few years and I’d lean towards an inability to get possession in the offensive zone. It’s appalling. Anyway, on to my points.

    1. A team’s PPG/60 is pretty tightly tied to its shooting percentage. It doesn’t really matter how you break this down, whether it’s by looking at the 240 team halves in the chunk of data that I have or the 120 team seasons. In the former case, the correlation is .828; in the latter it’s .784. Shot rate, by comparison, isn’t so tightly correlated with the goal rate. For the 240 half seasons I have, the correlation between the shot rate and the goal rate is 0.542; for the 120 full seasons, it’s .641. My take on this is that there are so relatively few shots taken on the PP over the course of a season (and particularly in a half season) and a wide enough spread in the shooting percentages, that a good run of shooting percentage, whether it’s luck or not, can plaster over a team’s failure to generate a lot of shots on the PP. This finds a little support, I think, if you look at the data over a four year period – the correlation then between PPS/60 and PPG/60 is .694, compared to a correlation between shooting percentage and PPG/60 of .755.

    2. Past shooting percentage doesn’t predict future shooting percentage on the team level particularly well. This is fascinating to me. I’m not sure how valuable doing this on a full season level will be here, given the lockout. It doesn’t really seem to me to make much sense to compare 2003-04 and 2005-06. If you look at team shooting percentages in 2005-06 versus 2006-07 and in 2006-07 versus 2007-08, the correlation is negligible at .294. If you look at the correlations in shooting percentage for the various halves between the first and second halves for 2003-04, 2005-06, 2006-07 and 2007-08, it’s virtually non-existent, at 0.08. Crazy shit happens in the course of a single half. Oddly, the best shooting percentage in a single half on the PP since 2003-04 belongs to the 2003-04 Penguins, the horrific pre-Crosby/Malkin edition of the team. That probably made Dick Tarnstrom a few million bucks though. Edmonton’s second half in 2007-08 finishes second on that list.

    3. Shooting rate predicts future scoring rates about as well as actual scoring rates do. The correlations are pretty weak but still – for S/60 to G/60 from the first half to the second, it’s .238; for G/60 to G/60, the correlation isn’t much better, at .274. If you compare 2005-06 to 2006-07 and 2006-07 to 2007-08. the correlation between S/60 to G/60 is .283; for G/60 season over season, it’s .308. Weak predictive values for each.

    At the end, what I take from this early sort of a look and what drives my skepticism is that the stuff that the Oilers were good at – basically G/60 and S% – isn’t the sort of stuff that tends to repeat. I have a theory that there’s a limit to what skill on the PP can get you in terms of shooting percentage and above that, you’re into luck. Does anyone really think that Edmonton is currently icing better offensive talent on the PP than teams like Montreal, Ottawa, Detroit and Pittsburgh have over the past four years? Those teams have shot 14.8%, 13.6%, 15.0% and 15.5% respectively over four years. If the Oilers shot 15.5% on the PP in the last half of the 2007-08 season, their PP wouldn’t have been nearly as impressive, scoring just 6.4 PPG/60. Not the horror show of the first half but nothing respectable either. With their inability to generate shots on the PP, they simply have to have a shooting percentage in the range that I think of as lucky, in the sense that they can’t expect it to continue in the long run, in order to put up good results on the PP. We can all hope that that happens but, as I’ve said here before and no matter what we saw last year, hope isn’t much of a replacement for a plan.

    I really think that the PP shooting rates are a systemic problem for the Oilers, something that goes beyond the players. Even when Pronger was in town, they were up to nothing more than average. Whatever they’re doing, it hasn’t worked for four years and the guys who’ve been responsible for gaining the zone, guys like Hemsky and Horcoff, they’re going to be big players on the PP again this year. I think that MacT is a hell of a coach, a guy capable of getting a lot out of his players but this just seems like an incredible weakness of his.

    In order for me to believe in this team the way that guys like Allen Mitchell and Jonathan Willis do, I have to believe that they’re able to start generating power play shots at a realistic pace. I haven’t seen anything to suggest that they’re at that point yet. We’re eight games into the season right now and the PP is scoring a respectable 7.0 PPG/60. The same old problems are in place though – they’re getting just 33.9 PPS/60. They’re currently shooting 20.6%. As long as the shooting percentage is through the roof, the good times will roll. When the shooting percentage dries up – and it will – they’re going to be in a world of hurt if they haven’t figured out a way to start getting the shots into at least the high forties. Given that the best non-Pronger half they’ve put up in the past four years is 44.3 PPS/60 in the second half of 2006-07, I’m not going to hold my breath in this regard.

    Unrelated: If you’re looking for something funny out of the whole Dave Berry thing, check out this post at Pensblog. They’re unduly harsh on Brownlee (even though I wasn’t completely in line with his take on the Oilers thing) but otherwise fair.


    13 Responses to Wish I believed like they do, Pt. I

    1. October 30, 2008 at

      Inability to gain possession in the offensive zone sounds about right to me too. Or at least, inability to keep it when it matters. They have been quite good at passing around the perimeter over the last several years, with Hemsky’s seeing-eye passes leading the way. Where they’ve been losing it this year for sure is when they try cross-ice passes. Those keep getting deflected. Then next time they try passing to Souray, who immediately has a defending forward on him, so he dishes to Visnovsky. Vis gets either a shot that the goalie sees all the way, or tries to pass it to somebody else, and they play keepaway for a while longer. They don’t seem to attack much from down low. When they do have that option, they seem to do better. It’s no surprise that that seemed to happen most when Penner was on the ice.

      Want to fix the power play? Drop Gagner to the second unit, put Flying Fridge in his spot with Horcoff and Hemsky. Break up Vis and Souray. First unit’s job is to cycle low and draw the D down, exposing Souray’s shot. If they don’t bite, well, you’ve got Penner’s butt in the goalie’s face. Second unit, tell Visnovsky to cheat down to the faceoff dot on the goalie’s right, where Stoll and Carter used to get those terrific one-timers off. Get Gagner to feed him. If Vis is blocked off, play the opposite side of the ice.

      I dunno, maybe I’m no better a PP coach than Simpson and MacTavish, but that’s what I’ve been seeing this year. Single option makes defending easier, and low percentage cross-ice passes are killing them.

      • Jerry Seinfeld
        November 21, 2013 at

        If shot percentages vary wildly, then analyzing shots is irrelevant. By definition. You cant pretend you have an actual science if what you say is only based upon “after the fact”. For example, yes the power play had an unreal shooting percentage. Thats why the power play was unreal. Is this a revelation?

    2. October 30, 2008 at

      I like to say that if I were a coach, the two tendencies I would beat out of my players with a broken composite would be diving and stopping up five feet into the zone and trying to thread the fucking needle to the trailer across the ice. I don’t know what they call that shit in the Czech Republic, but here in Canada, it’s called a “turnover,” and it’s one of the biggest flaws in Hemsky’s game. At least get past the top of the circles if you’re going to try that, come on.

    3. HBomb
      October 30, 2008 at

      So you’re meaning to tell me that all those obnoxious idiots at Oiler games yelling “SHOOOOOOT” as soon as the Oilers hit the opposing blueline on the powerplay…..they’re actually right?

      Ain’t that a kick in the junk….

    4. October 30, 2008 at

      I don’t think it’s prudent to dismiss shot% because it jumps around a lot more than shot totals. Just because it varies a lot more than the shot rate numbers doesn’t mean it isn’t a factor – it just makes it harder to deal with.

      Now, before you-know-who jumps down my throat expounding on repeatability, I’m not suggesting that any PP is capable of sustaining a shot% of 20%. It’s not unreasonable to think that with a sufficiently huge sample to look at, the inherent PP shot% might *repeatably* range from 12% for the bad to 15% for the elite.

      “If you look at team shooting percentages in 2005-06 versus 2006-07 and in 2006-07 versus 2007-08, the correlation is negligible at .294.”

      I don’t think .294 is negligible. Does anyone really think that with a sufficiently huge sample all PPs would produce an identical S% of 13.5%?

      From the Wikipedia page on correlation:

      “A correlation of 0.9 may be very low if one is verifying a physical law using high-quality instruments, but may be regarded as very high in the social sciences where there may be a greater contribution from complicating factors.”

      Sample size has a lot to do with the greater year by year correlation of shot rate. The complicating factor(s) getting in the way of a greater year by year shot% correlation would be the fact that it depends on 1) producing high quality shots, and 2) having those shots turn into goals. It’s the “shit happens” aspect of #2 that causes the shot% to jump around so much year by year, or half by half. If only there was a way to separate shot quality from shot%…

      Consider this:
      It’s a bit of a leap, but let’s assume the four year totals in the chart above are teams’ inherent PP shot production (PPS/60) and shot quality (S%). As a sniff test of the S% column, it looks like several points jive with popular opinion of the clubs I’m most familiar with: Ott, Buf, Pit, Mtl have PP skill and are mostly good at producing quality shots over this period while Fla, NYI are bad at it.

      Now look at the distributions of the PPS/60 and S%.
      PPS/60 ranges from 43 to 56. The mean is 48, the SD is 3.3. The average distance from the mean is 0.8 standard deviations.
      S% ranges from 11.5 to 15.5. The mean is 13.5, the SD is 1.0. The average distance from the mean is 0.8 standard deviations.

      *If* you take the four year sample to be teams’ inherent abilities, they are separated from one another by S% just as much as by PPS/60.
      An average PP takes 48 PPS/60, shoots 13.5%, and scores 6.5 PPG/60.
      Improving the PPS/60 by one SD to 51.3 yields an additional 0.5 PPG/60.
      Improving their S% by one SD to 14.5% yields an additional 0.5 PPG/60.

      There is more than one way to skin a cat. If you take the S% distribution above to be an accurate reflection of the real PPs shot quality, then shot quality is just as much a contributor to PP success as shot rate.

    5. mc79hockey
      October 30, 2008 at

      It’s not unreasonable to think that with a sufficiently huge sample to look at, the inherent PP shot% might *repeatably* range from 12% for the bad to 15% for the elite.

      I basically agree with this.

    6. October 30, 2008 at

      Another complicating factor that has to be considered is personnel. In 2003-04, Montreal had:
      Ryder, Zednik, Ribiero, Dagenais, Hossa the Younger, Hainsey, Perrault, and Souray

      and didn’t have:
      Kovalev (until the deadline), Streit, Kostitsyns, Plekanec (20 GP, no PP time), Higgins, Hamrlik, Latendresse, Tanguay, Lang

      Turnover in the NHL is huge, and different personnel will naturally affect the entire functioning of the power-play, in terms of ability to gain the zone, ability to get shots on net, and ability to put pucks in net, though probably more so between seasons than between halves.

    7. October 30, 2008 at

      Am I You-Know-Who, Jeff?

      In any case, I agree with you I think, though I’d have to get back to it and make an effort to understand your mathy stuff, some time when the mood strikes I suppose. Though PP performance is just so hard to peg with any reliability at all.

      Just generally I think you need to paint with a really big brush with PPs. Riversq posted quite a few things on the subject on OilFans in the past, and it seemed to work the best.

      Just generally looking at the player’s long term history of being able to put up points on the PP. And knowing that you need two, if not three, guys with good track record at that to have a decent chance of having a good powerplay yourself.

      And some coaches just always seem to have crap powerplays. Neilson for the longest time tracked every little bit of minutae from PP video, adjusted tweaked and planned … and it didn’t go so well for him most years. Ron “plan, who needs a plan?” Low would probably done a hell of a lot better by doing nothing at all. Calling practices off early and sending everyone home happy.

      When he arrived in Ottawa Neilson was charged with reviving OTT’s long underachieving powerplay, obviously Keenan had rubbed off on him, because he picked six players as his PP go-to guys and insisted they not practice the PP. That must’ve pissed of Martin, a man who plans planning meetings to develop new plans, but that’s what he did. I mean it’s hard to practice your PK if your best PP guys aren’t allowed to go against them.

      Then, presumably, Roger went for brunch. That was it. And the OTT PP was very good during his tenure iirc.

      Go figure this stuff. With EV play, just the crudest commonsensical methods churn out repeatabilties of underlying numbers in the r=.80 to r=.90 range and predictive correlations to future results that brush damn close to what is ever going to be possible, or would be possible with trivia craps. And we haven’t really even given a collective ‘try’ at that yet, we can do better.

      But special teams stuff is off the hook.

      Maybe we should be looking at it more like baseball. Where nothing repeats like it should, and recent results seem to matter a bunch at every level.

      Recent results have almost no bearing 5v5 hockey results (practical reasons, like ‘Lidstrom broke his leg! aside), they add absolutely nothing extra in trivia craps. But EV hockey at this level just doesn’t leave players time to think, it’s mostly automatic reactions and skills, learned or instinctive.

      PPs though, there is time for guys to think. Different ball of wax.

    8. October 30, 2008 at

      Am I You-Know-Who, Jeff?

      Bingo! I just thought it would be best to try to preempt another shot quality/quantity war.

    9. October 30, 2008 at

      Bingo! I just thought it would be best to try to preempt another shot quality/quantity war.

      I didn’t know that there was a war. It’s a different set of hockey fans I think.

      We’ve got the chemistry guys … why did the Oilers goal diff drop by about 60 the year after Gretzky left? Why did the Kings goal diff go up about the same amount that year?

      Chemistry, of course. In combination with ‘gel factor’, ‘will to win’, scoring BIG goals and changing the momentum of the game. Plus whatever keLLY HRUDEY SAYS NEXT. YES! I KNOW I’M TALKING IN CAPS NOW!!! THE 80S OILERS HAD CHEMISTRY, I’M ABOUT TO IMAGINE A WHOLE NEW ANECDOTE!!!!!

      Gretzky was a really, really good hockey player.

      Some people find solace in shit that flat out didn’t happen in hockey games. Bless ‘im. We’ll call them the Staplites for now.

      The people who gravitate towards A2 just want to talk hockey with each other.

      The people who answer A1 seem to live for any slight validation from the crowd that picked A2. Why the hell they just can’t go back to the fan boards, or ring in to talk radio … well that’s beyond me. The heart wants what it wants, I suppose.

    10. October 31, 2008 at

      I really think that the PP shooting rates are a systemic problem for the Oilers, something that goes beyond the players. Even when Pronger was in town, they were up to nothing more than average. Whatever they’re doing, it hasn’t worked for four years and the guys who’ve been responsible for gaining the zone, guys like Hemsky and Horcoff, they’re going to be big players on the PP again this year. I think that MacT is a hell of a coach, a guy capable of getting a lot out of his players but this just seems like an incredible weakness of his.

      “Systemic problem” is the key phrase here. You probably know that I agree with you about the quality of MacTavish, but he can’t coach a powerplay. Some of it is personnel, I’m sure, but not all of it. Hemsky has consistently been among he best options in the West, but MacTavish refuses to use him nearly enough. Horcoff/Gagner is pretty much a trade off in terms of PP efficiency in my opinion, but using Cole over Penner is indefensible on any level. Using 44 and 71 on the same unit when their key strength is identical also boggles the mind.

      And none of this is even getting into tactics.

    11. David Staples
      August 24, 2010 at

      Staplites . . . I like that.

      Crazy bunch of kooks, I hear.

    12. David Staples
      August 24, 2010 at

      Just reading this again, after following a link from Gabe.

      Seems to me the bottom line here is that the teams that generate the best scoring chances scores the most goals, but that if you generated great scoring chances as a team in the past, it doesn’t mean you will do so in the future.

      Counting up powerplay shots isn’t so crucial, either way.

      At the risk of being a Staplite, I think I’ll keep trying to figure out which players actually help to create goals and the best scoring chances, and which do not. Dustin Penner’s high number of unofficial assists in 2008-09 for screening the goalie were real and significant.

    Leave a Reply

    Your email address will not be published. Required fields are marked *