I get that certain moments in Oilers history are woven into the tapestry of the team and that they are true because they are true. Wayne Gretzky and Kevin Lowe won four Stanley Cups by peering into the Islanders dressing room after the 1983 Stanley Cup finals*, Grant Fuhr never let in the next goal, the Oilers were screwed over by the CBA because it forced them to trade all of their good players who were totally about to win a Stanley Cup, Tommy Salo was ruined by the goal against Belarus, the Oilers missed the 2004 playoffs because of a loss to Los Angeles on HNiC on January 31, 2004, Adam Oates’ contract was worth it because he taught Shawn Horcoff and Jarrett Stoll how to win faceoffs, the Ryan Smyth trade wasn’t about money…these are the stories that calcify over time and become part of the story of any sports franchises.
They drive me nuts.
The one that galls me at the moment is the discussion of the Oilers down the stretch last season. There seem to be two myths that have developed about that run. The first, which is relatively easy to dispel, is that they actually played pretty well. The second, and more difficult to disprove, is that their play was somehow related to them playing looser. I’m going to deal with the first, which is easier to disprove than the second.
Robin Brownlee jumped on the bandwagon today:
Funny thing how expectations work. When there are none, you get rolls like the Edmonton Oilers fashioned in their final 20 games last season as they went 14-5-1 and looked like the 1957 Montreal Canadiens doing it.**
Brownlee’s certainly not the only Oilers fan to have developed this point of view. It’s shared by virtually all of the media. Most of the fans were relying on this to explain why they figured that the Oilers were going to be in it this year. Presumably it was in the back of Craig MacTavish’s mind when he made his utterly insane comment at the start of the year about contending for the division title.
It is, I think, too easy to overlook that the Oilers record was in no way supported by their goal differential. I’m going to leave the SO goals in to help with the point that I’m making and give the Oilers every benefit (they were 3-1 on the SO in the final twenty). During their final 20 games last year, they scored 69 goals and allowed 63 goals, for a goal differential of +6. Teams do not commonly put up 29 points while going +6 over 20 games.
If you treat 1-20, 2-21 etc. of each team’s season as a 20 game segment, there had been 62,491 20 game segments entering the 2007-08 season. 1366 of those had seen a team post a +6 goal differential. 1494 of the 62,491 20 game segments saw a team post 29 points.
I’ve tabulated some information below to try and convey just how unlikely this sort of thing is. First, the points earned by the 1366 teams that had a +6 goal differential over a 20 game slice of the season. As you can see, the Oilers are at the right end of the curve; in fact, their 29 points in 20 games with a + 6 goal differential ties the record.
You can see that, post-lockout, you’ve got about a 1/40 shot at 29 points if you’re +6 over 20 games. Those are not particularly good odds. Historically, the odds are even worse, although the whole graph is going to be shifted to the right by the ease with which the NHL dispenses points these days. I think that you can see the beginnings of that in the information presented above.
What about teams that accumulate 29 points in 20 games. What kind of goal differential do they generally require to support such a thing?
Crazy things happen in those stints where teams put up 29 points in 20 games with an extremely low goal differential. The 1985-86 Caps lost only five games in their twenty, but they got absolutely crushed in two of them - a 7-0 loss to Detroit (17-57-6 on the season) and an 8-1 loss to Pittsburgh (34-38-8 on the season). The 2005-06 Predators went 8-3 in 1 goal games, which is just an astounding volume of 1 goal games along with a spectacular success rate in them.
While low goal differential/high point runs have become a lot more common since the lockout (2/3 of 29 point runs with +10 or lower goal differential happened in 2005-06 or 2006-07), they still aren’t particularly common. The Oilers 20 game run was still very unusual but because of the low goal differential that produced the record. Their goal differential looks nothing like the typical team that goes on a 29 point run and the typical team that accumulates 29 points over 20 games has a far superior goal differential. Even in the post-lockout era, where points are free and easy, a +6 goal differential is in the 3rd percentile of goal differential for 29 point teams.
Simply put, the 2007-08 Oilers were not full value for their results down the stretch. The argument that they somehow morphed into some sort of unstoppable hockey machine once the season was lost doesn’t fit the goals for and against. Any argument that their looseness was benefitting them in the standings has to explain not only how “looseness” produces results but how the Oilers managed to spread out their goals so perfectly. I don’t see a +6 as being particularly unusual for a team like the 07-08 Oilers. The unusual bit was the results. Strange that the looseness would affect that only.
*TSN tells that story like so:
As they made their way out, Gretzky and the Oilers noticed that while the family members and team personnel were popping champagne and sipping from the Cup, many of the Islander players sat in their stalls looking tired. They weren’t partying it up so much as they were icing down knees, tending to bruised ribs and rubbing ointment on their shoulders. As Gretzky later wrote in his autobiography, he realized that these players made the mental and physical sacrifices to reach the top.
Having witnessed this rather surprising sight, Oilers’ defenceman Kevin Lowe nudged Gretzky and whispered, “That is how you win championships.”
Maybe things were different then, maybe the cameras don’t get into the dressing room fast enough, maybe a lot of things…but when I watch teams win the Stanley Cup now and the dressing room interview proceeds, I’ve never noticed a lot of ointment rubbing. It seems to me that everyone pretty much seems to be running around and getting their drink on.
**I’m not really sure what he means by this to be perfectly honest, whether he means that they were playing firewagon hockey or whether he means that they were really good. Incidentally, the 1956-57 Montreal Canadiens had stretches in which they went +6 and stretches in which they put up 29 points. The 29 point stretches involved them going +24, +27 and +31. The +6 stretches saw them put up 19 points twice. I obviously didn’t see them play (no Centre Ice) but, whatever the Oilers were doing down the stretch, they certainly weren’t channelling them.
the Oilers missed the 2004 playoffs because of a loss to Los Angeles on HNiC on January 31, 2004
Wasn’t that Ryan Smyth Double Icing night?
Great post.
It’s funny — prior to reading this post, I had assumed that the Oilers GD during that stretch in question was at least somewhat commensurate with their record.
Without even knowing it, I guess I bought into all of the media hoopla.
Re: 1983 Gretzky/Lowe peering into the dressing room:
Why ruin a good story with the facts?
Great post Tyler. The truly scary thing for this season is that the Oilers’ current goal differential doesn’t support their record, especially if you isolate the road results. I think people have been tricked into thinking these guys are a bad home team when they’re actually just a lucky road team. I guess they must play looser on the road though, which helps with their clutch goal differential even if the regular goal differential isn’t helped much.
Great post.
Goal Differential has been the reason I’ve been waiting for the Oilers to quietly drop out of the playoff hunt — and now that the sample size is small enough, I’m hoping for a bit of luck!
I get that certain moments in Oilers history are woven into the tapestry of the team and that they are true because they are true.
Delete “Oilers” and replace “the team” with “the discussion” and the statement loses none of its veracity. It drives me absolutely up the wall. And causes me to drink.
The luck cake beneath the beautiful luck icing.
http://www.timeonice.com/playershots0708.php?team=EDM&first=20945&last=21230
Funny thing, I remember at the time that the Oilers were getting outplayed and just getting the breaks. I wasn’t reading much of the sphere at the time, but I would imagine that was the concensus of the sharp kids. And I remember MacTavish on the radio telling us the same, in coachspeak. But by early summer he seemed to have convinced himself that the aurora was rising behind them. Odd stuff.
I’m glad some tackled this 14-5-1 thing though, over the past week it’s become the local sports media’s favourite three numbers with which to start an Oilers discussion.
@Vic: I was sifting through some quotes before I wrote this and you’re right. If someone wants something to do, it’s interesting to go back and look through the game stories on slam.ca. Lots of quotes from MacT and guys like Brodziak (odd, given that he’s one of the guys now talking about how good they were last year) about how they were lucky to get the point, etc etc. This seems to be a case where the 14-5-1 looked a lot better once people got some distance from it and the details were lost in the mists of times.
you get rolls like the Edmonton Oilers fashioned in their final 20 games last season as they went 14-5-1 and looked like the 1957 Montreal Canadiens doing it.**
Gee, did the 1957 Montreal Canadiens typically get outshot by 9 shots per game, as the Oilers were during those 14 wins? (497-371) What was exceptional about that run was how many of the wins — and losses too — went against the flow of play.
The 1957 Canadiens are about the last team I’d want to be comparing that 2007-08 Oiler club to. Ridiculous.
As far as the luck cake goes, the PP was going berserk too. This year, same shitty shot rate but so-so shooting percentage.
By the time the Oilers would have finished changing and started leaving after the 83 cup loss it would have been long after the Media had left the Islanders dressing room. By that time I would assume that some of the adrenalen would have worn off for some of the Islanders players (especially the pretty banged up ones) and they would have started treating their injuries.
Wayne Gretzky and Kevin Lowe won four Stanley Cups by peering into the Islanders dressing room after the 1983 Stanley Cup finals*,
IIRC there were comments made to that effect in the immediate aftermath of the ‘83 Finals, it wasn’t something that cropped up out of whole cloth a year or a biography later. A quote from Kevin Lowe to the effect of “this is what we need to do” was in the next day’s paper, probably on the same page as that epic picture of an utterly dejected Paul Coffey sitting in the open gate while the attendants cleaned up the litter after the empty net goal that sealed it.
It was apparent to any neutral observer, and even some inherently biased ones like yours truly, that the Islanders were prepared to pay a heavier price, do whatever it took to win. Don Cherry was blah-blah-blahing about toughness and character and was, as is so often the case when he is actually right, belabouring the obvious.
The 1983 Oilers who reacted negatively to Billy Smith’s slashery and other insults and distractions learned a lesson. By 1984 they had matured into a much more disciplined, focussed unit that walked the talk of “this is what we need to do”. Key plays of the ‘84 series included Glenn Anderson skating through a Smith two-hander without a concern, or absorbing a vicious Bob Nystrom elbow/fist/butt-end to the head and not retaliating and instead earning the Oilers a powerplay (that’s in the first period of Game 5 for those who got the box set). Adding hardrock players like Kevin McClelland (who pummelled Islanders’ bully Duane Sutter in a memorable Game 4 scrap) and Jaroslav Pouzar (who crushed Billy Smith in an “accidental” collision in Game 3) to the line-up also made a significant difference, and for the first time in their three playoff series the Oilers were able to match the Islanders’ toughness and discipline, and press their advantage in the skill department to win going away.
This is
I hate it when Bruce leaves us hanging like that
Ty: it’s bad enough that you and I only visit Oilersnation in order to belittle Gregor and Brownlee but now you’re doing it at your own site as well?
I stopped reading this article when I realized it was how ridiculous it is. The bulk of what I read bathed itself it odds, as in how unlikely it is for a team to have a plus 6 goal differential and still get 29 points. The argument seems to be that this is proof that the Oilers weren’t actually playing well during the last 20 games. And it almost makes sense until you realize that it’s stupid.
Just because the odds aren’t in favour of 29 points in 20 games that does not mean that a team is not playing well should that occur. It certainly doesn’t mean they just got lucky, it’s possible, but the statistics shown are far from proof.
Further more, any one who watches sports with a critical eye can spot when a team is playing well and working as a unit, and I don’t think there are very many people who would suggest that the Oiler’s weren’t playing well at the end of last season.
This gentlemen refers to the media last year all reporting on how well the team was playing at season’s end and seems to imply that the majority of these reporters go it wrong.
Let’s play the odds on that, dingbat. What’s more likely, that a large team of hockey analysts made the correct assessment about a team, or a single dork with numbers got it right.
Do the math.
I think you’ve clearly demonstrated in the past that, over the long-term, goal differential is a pretty nice proxy for “team quality”.
But I think we need to be careful to not let that obscure the fact that it is still Points Earned that is ultimately the most important measure. In principle — and especially over the short-term — there are other routes to a good Record than high GD. “Luck” is the usual culprit fingered, but I wouldn’t be at all surprised to learn that other factors are also important.
In that light, the only thing I’d take issue with in this post is the statement that the “Oilers were not full value for their results down the stretch.”
I think that’s grabbing the wrong end of the stick. They were, in fact, getting fantastic value for their quality of play (if GD is, as we think, a reasonable proxy for “quality”).
It still doesn’t change the overall conclusion that they Oilers were — more than likely luckier than normal during that stretch; but it places the priority where it should be IMHO.
BTW — have you ever posted on GD in early regular season vs late regular season vs playoffs?
I’d be interested to see if GD is partially context dependent or if these weird runs of good record/low GD are randomly distributed throughout the season.
Beevbo: That’s no way to talk. There are iron-clad numbers to support all the arguments put forth here.
They prove without any doubt the Oilers in their final 20 games weren’t actually the 1957 Montreal Canadiens - forget that it was said tongue-in-cheek for effect to reflect the record rather than how they attained it. Thank goodness that’s been clarified and to hell with any narrative licence that falls outside the numbers.
I’d love to know if Robin and Beevbo posted from the same IP address…
Uh, no we didn’t.
It’s interesting, laughable actually, that anybody would think I was suggesting the Oilers of the last 20 games a year ago were literally the 1957 Canadiens, let alone take the time to disprove it.
With nothing to lose, the Oilers were loose — which played a part in things, despite Tyler’s doubts about that — got some bounces and got on a roll.
I wasn’t fooled by the 14-5-1 record. I didn’t forecast great things for this season based on the results of that stretch. I didn’t think the Oilers were the Canadiens. Suggesting as much makes for some fun and another poke at a gullible MSM guy, but it ain’t so.
Had I been all warm, fuzzy and fooled by the last 20 games, I probably would picked the Oilers — with the addition of a healthy Souray, Horcoff, Moreau etc and Visnovsky added to the fold — to be division contenders, like a lot of people did. I picked them ninth. That would rate as a disasterous off-season for the Habs, no?
Good question PDO. Surely there can’t be two people that dense.
Tyler, whip up some odds on that and check the IP addresses while you’re at it.
Nah, Brownlee’s honest. I’m as surprised as everyone else that (at least) two people on the internet think that I’m an idiot but it is apparently true.
It’s interesting, laughable actually, that anybody would think I was suggesting the Oilers of the last 20 games a year ago were literally the 1957 Canadiens, let alone take the time to disprove it.
If I thought that you were suggesting that they were literally the 1957 Habs, I wouldn’t have relegated my comment about the 1957 Habs to a footnote. It would have been the entire post. Your post served as a jumping off point for the discussion on how strange it was to see a team put up 29 points with just +6 goal differential and how the Oilers, even while “on a roll” still didn’t post particularly great numbers.
They prove without any doubt the Oilers in their final 20 games weren’t actually the 1957 Montreal Canadiens - forget that it was said tongue-in-cheek for effect to reflect the record rather than how they attained it.
Huh. I wasn’t really sure what in the hell you meant. Like I said: “I’m not really sure what he means by this to be perfectly honest, whether he means that they were playing firewagon hockey or whether he means that they were really good.” That’s why that was a footnote.
I still don’t understand how “they…looked like the 1957 Canadiens doing it” is a reference to their record rather than to how they got it. If you’d written “The Oilers roared down the stretch like the 1957 Canadiens, going 14-5-1″, I’d get your point but I just find this confusing. Again though, not the main point.
With nothing to lose, the Oilers were loose — which played a part in things, despite Tyler’s doubts about that — got some bounces and got on a roll.
Except that they didn’t really do exceptionally well in terms of scoring or preventing goals. We know, from Vic’s link above and from paying attention, that they were still getting hammered in the shots. It seems that the only thing that changed was that a few more went in, a few more stayed out and the wins went through the roof. People who read this site know that the percentages are notoriously fickle.
The trouble I have with the language of “went on a roll” (and this maybe should have been more explicit) is that it implies that they started playing markedly better hockey. They don’t seem to have. The goals for/against was somewhat better and the wins skyrocketed. My view is that they were getting a bit of puck luck and a lot of win luck. I don’t really think of that as being on a roll.
Tyler,
Yeah, the Oilers weren’t playing particularly well down the stretch. I got yelled at a lot on the intarweb every time I pointed it out, but this didn’t change the fact that this was what was happening on the ice.
Next time the Oilers come from behind with an obviously fluky goal in the 3rd period, like a bounce off of a stanchion or a shot that goes off of two legs and in. Check out the game day threads around the web … that bounce will have turned a bunch of Oiler players from assholes into great guys. Mad shit, this sports fandom.
Granted, something about the playoffs turns me into ‘that kind of fan’ every time the Oilers make the post season. Of course when I do that, it’s delightful.
In my opinion, Alan Reifman’s website is a good jumping on point for anyone interested in sports streakiness in particular, and randomness in sport in general. More as a source of links and authors than anything.
I’ve been reading quite a bit of baseball stuff lately, Jim Albert and Bill James are easily the most interesting for me. A cool quote from a 2003 Bill James interview on NPR out of Boston, I was listening to it a couple of hours ago:
“Almost all fans and 100% of people who run afternoon sports call-in shows believe that players are either clutch players or they are not clutch players. Nobody has ever been able to demonstrate that this is true, probably because it isn’t.”
Dude has a way with words. Granted, it’s probably fairer to say that ‘clutchness’ is nowhere near as important as talk radio show hosts and callers would make you believe.
To show that I’m fair, I think that if RB truly believed in the club’s late ‘08 success, he indeed would have predicted more than just a 9th place finish for them in ‘09.
But, he’s like most - if not all of the MSM - and there’s no way we’re at the point where someone’s gonna dig through the numbers and come out and proclaim a streak as being as a fluke.
It’s just not the way these guys are wired. Now, that’s not to say these guys can’t learn to look at things in this light but how many years did it take before OPS was listed on baseball telecasts.
BTW, I harp more on letting Hejda go then Glencross - Hejda stood out to me and I was worried that Glencross was pan-flashing - but I had a limit that I wouldn’t go past for GlenX and he didn’t ask for that so, yeah, they should have brought him back as well and it was within reason.
Anyway, I’m scanning through faceoff.com today and there was a piece on GlenX playing with Iginla and Jokinen.
And here’s the quote that really made me miss him.
– In 63 games, the former member of the Alberta Junior Hockey League’s Brooks Bandits has generated 38 points, including 12 goals.He’s gunning for “20 and 20,” but claims offensive totals don’t drive him.
“Not as much as things like plus-minus,” says Glencross, who rates a lofty plus-14. “Because when you’re a second-, third-line player, it’s all about outplaying their second or third line.” –
Complete link
http://www.faceoff.com/hockey/teams/calgary-flames/story.html?id=1401043&p=2
And it almost makes sense until you realize that it’s stupid.
This made me laugh out loud.
Yeah Dennis, I read a bit on Howson in the Colombus Dispatch the other day. He’s really turned over that roster, it’s incredible when you look at the list of players moved since he arrived.
Anyhow, he is quoted as saying that the only trade he regrets making is the Glencross deal. That team has a whack of good forwards already though, Glencross undoubtedly had a bigger impact on the Flames than he would have in CBJ.
Mr.dB: Sorry ’bout that, just an editing/posting error.
The truly odd thing about that 14-5-1 stretch was that not only did the Oielrs get collectively outshot by 648-553, but the distribution of those shots w.r.t. the results.
In the 14 wins:
371 shots for (26.5 S/G)
56 goals (4.00/G)
497 shots against (35.5 SA/G)
38 goals against (2.71/GA/G)
In the 5 regulation losses:
151 shots for (30.2 S/G)
6 goals (1.20 G/G)
115 shots against (23.0 S/G)
20 goals against (4.00 GA/G)
And from the 20-game segment collectively:
Winners: 522 shots, 80 goals
Losers: 679 shots, 48 goals
It seemed like every game’s outcome was running contrary to the shot clock. It was clear in many of these games that the shot clock was being driven by the scoreboard and not vice versa; the trailing team was really pushing the play.
Adjusting out shootouts by declaring them de facto ties moderates the effect but hardly eliminates it from this particular sample. It was a pretty unusual stretch.
It seemed like every game’s outcome was running contrary to the shot clock. It was clear in many of these games that the shot clock was being driven by the scoreboard and not vice versa; the trailing team was really pushing the play.
I’ve come up with some good evidence that this is the case leaguewide, at evens anyway. You know my thoughts on mashing the game states together. I want to finish with 06-07 and 07-08 before I post it but there does seem to be a very real effect in terms of teams doing worse shotswise when they’re leading than otherwise.
For reasons I’ll explain when I post on it, I don’t think that this undermines the idea that outshooting your opposition is a good thing, only that we might need to apply a little more context.
I’m also suddenly curious about Penner’s TOI when the Oilers are up by 1 or 2 versus his TOI when they’re down 1 or 2. Time will tell whether this is applicable to individuals or whether coaches shift their player mix but it might provide some defence for Moreau and explain away some of Penners great outshooting numbers.
Vic: Oddly enough, I had that link up and ready to go just as I flipped back there - I found it via prosportsdaily.com
As has been said here a little bit before, it looks like Lowe moved upstairs one year too late;)
I’ve come up with some good evidence that this is the case leaguewide, at evens anyway.
It stands to reason.
You know my thoughts on mashing the game states together.
… and you may know mine on disentangling them. Just different approaches, some validity to each I’d say. Holism v. reductionism and all that.
For reasons I’ll explain when I post on it, I don’t think that this undermines the idea that outshooting your opposition is a good thing, only that we might need to apply a little more context.
Context being game score at the time shots are taken, I hope. I’ve been interested in that for some time but don’t have the programming chops to extract them (to put it mildly). I expect the results would be teams generally outshoot to get the lead, but are prone to being outshot while protecting it. That was certainly the case in that 20-game closing stretch in ‘07-08, and a more general observation of mine over many years which I will be interested to see under the spotlight of rigorous analysis. I’ll look forward to your results, Tyler.
I’m also suddenly curious about Penner’s TOI when the Oilers are up by 1 or 2 versus his TOI when they’re down 1 or 2.
A very interesting question. I’m curious too. Lots of players who might see more or less ice time depending on leading or trailing.
That was covered on a post here months ago, no, Tyler? Playing to the score, and all that.
How it applies to players seems like a cool twist, I wouldn’t have thought player roles changed that much with the score.
And of course the Oilers weren’t nearly as bad by the territorial advantage numbers when the score was tied during that stretch.
http://www.timeonice.com/playershots0708tied.php?team=EDM&first=20945&last=21230
But they were still awful. Still, a bunch of games they were leading despite being outplayed, then they were hanging on for dear life in the third periods, so it probably looked worse than it seemed for the Oilers during that lucky stretch.
@Vic: Amazing what a 12.3% ES shooting percentage will do for your. You’re right about the discussion, although it was inferred more than covered. I have a touch of OCD and wanted to see if it checked out. It looks like it does, although I haven’t looked at 06-07 or 07-08 yet.
As far as player roles, what I’m wondering is if guys like Penner play more in situations where the opposition is going to be hanging back. If you think of true checkers, like the Zero Line in Anaheim in 06-07, I’d have no problem believing that part of the reason there numbers were so bad was that they played virtually no time when the opposition was sitting back and lots of minutes when they were pressing.
As well, Tyler, just generally I think things would be clearer if you got away from shots, corsi is better. Better yet would be a collection of every freaking thing they record by zone, then rationalize for things that double up, and weight things that are more indicative of meaningful possession in the zone than others. Then form it into an extraordinarily complicated equation that discourages people from looking at the elements, even those who were born to study the minutae. Then give it a cool name and wiki it in as a clear measure of terriorial advantage in a hockey game at even strength.
If nothing else, it should end the endless and senseless “so you’re saying blocking shots is a bad thing?!” conversations.
The zero line was high event the year before and the year after, no? Just coincidence, I imagine.
Along those lines, recently JLikens had some stunning stuff at his blog regarding EV shooting% at the team level, looking at the general population of teams and the spread of results.
And if you look at Dennis’ scoring chance stuff, just go to Bruce’s last summary where he subtracted the icetimes for the missing games. The total EV scoring chances (for and against) for a player correlate extremely well with his EV icetime. r=.99 or so I think, you can cut and paste yourself to double check.
The puck has to be somewhere, and intuitively I really doubt that there is much difference between high-event and low-event players in this league.
Yeah, I agree with that Vic. Here’s the thing though. My two years of data have the ratios of SF/SA at something like this:
+2: 0.81
+1: 0.87
0: 1.0
-1: 1.15
-2: 1.24
Now, as of yet, we don’t know why that is. Playing to the score is the obvious one. If player TOI is slanted towards one or the other end, I’d expect that their SF/SA ratios to be similarly slanted. I’m not referring to the Zero Line’s low event season, only that they are, I would think, more likely to be on the ice when the Ducks are playing passively and the other team aggressively than vice versa.
You’re right about the discussion, although it was inferred more than covered.
It was a great discussion as far as it went. As I recall — and I’m such an old fart that my memory is fading by the week — Vic did a shit ton of great work extracting the shots results for when the score was tied, which was very informative. (Tyler: Do you recall the name/date of that thread or can you link it?)
However, I don’t believe we ever got to shots data when one or the other team held a lead, let alone a lead of a specific margin. I suggested at the time that such information on leading/trailing might be both productive and important, but as mentioned many times I don’t have the programming chops to do it myself nor the right to make requests. I just hoped that sooner or later somebody would post something along that line.
Your numbers posted immediately above, Tyler, are the first I’ve seen to that end. Frankly they are along the lines (or at least the slope) of what I expected, but it’s very exciting to see the trend so clearly delineated.
The other expectation/educated guess I have is that the trailing team will have a relatively lousy shooting percentage on those higher number of shots, and that the leading team will have a good conversion rate.
A further filter which might be useful is time of game. A superior team will play differently with a one-goal lead in the first period than they will in the third. I have read some of the Contrarian Goaltenders’ work where he simply has worked with game score after 40 minutes and resultant third period shots, which is cruder cuz the score of course can change during that third period, but his work really advanced the discussion.
The idea of extending the concept to effects on individual players is intriguing. Role players in both directions might be negatively/positively affected in their shot count, and some strain on the relationship between shots and goals. In general the guys who play one special team and not the other might be expected to play more or less depending on leading or trailing.
I’ll look forward to your further results with great interest.
A further filter which might be useful is time of game. A superior team will play differently with a one-goal lead in the first period than they will in the third.
I had a spreadsheet crash and loss partway throug doing this a few days ago, so I haven’t recreated everything yet but the data seemed to support that. The effect is more pronounced in the third than the first.
For certain, if a player is starting more in his own end he’ll have worse underlying numbers, and ultimately worse EV +/- unless the hockey gods intervene.
Myself, Scott, slipper, Matt and probably several others have taken a shot at ‘how much’ before.
The issue of some guys getting more icetime when the team is leading is interesting, I can’t say that I’ve noticed it by eye though.
Something has to explain Penner’s stunning numbers though. I hadn’t checked the numbers in a while, or read blog comments in a while, but in a thread on here a couple of days ago Bruce put up some Penner results that show his underlying numbers are as strong as his counting EV numbers, so it’s not fluke. It’s compelling information, it moved Ron MacLean to use on the on-ice shots-for/shots-against numbers to defend Penner against Hrudey and Stock, and I’m sure that he won the argument in the minds of rational After Hours watchers coast to coast.
Still, it just doesn’t feel quite right to me. I remember the T.B game, that’s really a one line team and if MacT can go power vs power with Horc-Hemmer that sets the game up really well. And that’s what MacT did, and it went pretty well, and the shifts often ended well too. But still, they should have been better, the play was dying too much with Penner in all zones. And it wasn’t because he was trying to make plays that would have created scoring chances … the play was just dying for no reason too often.
By memory (always dangerous) MacT lost patience and pulled him off late in the game. So I go back and look at Dennis’ scoring chances in the game and there is Dustin outperforming 89 and 10. WTH. So we need to capture that.
Or maybe it’s just my personal bias shining through, I dunno.
At some point I’ll go through the shift chart and the play-by-play for that game and try to suss it out.
Some of that is explained by him being the best option more than being good. But I think there’s more to it.
It’s funny, this notion that Beezbo and Brownlee are the same guy - although I hope it was made tongue in cheek.
I read Brownlee’s comment in his article - and while I tend to think that the results cause the looser attitude rather than the other way around - I took it the way he seems to have meant for it to be taken: the Oilers put up wins like they were the 1957 Montreal Canadiens.
And they did have a nice tear - albeit one that was heavily based on luck and things working out, rather than genuinely outplaying their opposition. I don’t think that Brownlee has suggested it was sustainable; simply that a team with nothing to lose tends to play above their heads.
Beezbo’s argument, in contrast, was basically that if Rob Tychowski, Terry Jones and Joanne Ireland agree on something, there’s no point in using the numbers because they’re obviously right.
And in fairness, I really didn’t think that Tyler was attacking Brownlee so much as exposing something that we should all know - that the team last year got a bunch of breaks (in particular the Kid Line) and put up points well beyond their actual level of play. Which is important, because the “we were relaxed, so we caught fire down the stretch” narrative gets repeated way too much.
And don’t look now, but that lovely luck cake that Vic posted earlier shows some nice things about Marc Pouliot and Ladislav Smid. Surprisingly nice things, considering how little press I remember them getting down the stretch.
And Dennis, thanks for that Glencross quote. OF course, you and I both know that outplaying the other team isn’t as important as creating “momentum”, right
JW: it’s all about the energy you provide; and it’s also nice to be able to kill off certain penalties that are easier to kill than others.
“Any argument that their looseness was benefitting them in the standings has to explain not only how “looseness” produces results but how the Oilers managed to spread out their goals so perfectly.”
OK. I’ll give it a shot.
If I understand it correctly, the fact that the team was able to pull off something so statistically unlikely is being written off as “luck” because there seems to be no quantitative basis for labeling it as anything else.
The thing is, this “looseness” that you’re so casually writing off is the root cause of the luck. As I’m sure you’re aware in sport, the higher the caliber of competition, the more critical the mental aspect of the game becomes.
When an athelte or team is “loose”, and by loose I mean exhibiting the ability to play way above their heads for an extended period of time, they make their own luck. Things start to happen that are otherwise unexplainable. Things like perfect “seeing eye” passes, out of this world glove saves and a whole plethora of shots that wouldn’t normally go in do just that.
High caliber athletes deal with two major emotions - fear and confidence. The fear I’m talking about is fear of losing, and the consequences of such. You go one way (too much fear) and you have Shawn “Whiffcoff” 2009. You go the other way (lots of confidence) and you get the kid line 2008. Generally speaking, the player’s emotional state is determined by how much he can control the fear/confidence ratio. If something extraordinary happens that causes a radical reduction of the fear input (say the fact that you’ve come to the point of “nothing to lose”), then confidence escalates and what we term “luck” is more likely to happen. (As an aside, even most world-class athletes can’t get themselves to a hyper-positive mental state for very long and not nearly as often as they would like.)
Except that alot of it isn’t luck. It’s just the type of athletic performance that happens when the human mind has been released of doubt and anxiety. The brain functions far beyond normal constraints and athletic performance is enhanced as a result. Those few athletes that are physically gifted and mentally tough enough to control their cognitive state to such a positive degree on a regular basis we call phenoms. Or more accurately, Tiger Woods, Wayne Gretzky, Roger Federer, Micheal Jordan, Roger Phelps et al. These guys play “loose” almost all the time. Do you call them “lucky”?
So we have a unique, almost unrepeatable situation where a bunch of world-caliber athletes have pretty much all their negative mental barriers removed from their usual playing state. Coupled with the fact that their opposing teams (who are in contention) are playing “tight” (Oilers 2009) and it’s just possible you get something amazing, almost inexplicable.
If we assume the margin of difference between winning and losing in pro sport is alot smaller than we think, confidence and general mental well-being is sometimes all that separates winners from losers. That small GD proves the point. As you have correctly asserted, they weren’t playing so much better than their opposition. But their superior mental state was enough to cause the ice to tilt in their direction more often than not.
I’m sorry you can’t quantify an athletes (or team’s for that matter) mental state in an excel spreadsheet, but for you to pass off a good run of exceptional performances as “luck” is rather one-dimensional. Granted, there’s always some luck involved, but it’s highly unlikely to carry over 20 games. Your own stats prove that point.
David S.: Fun question - why does that “looseness” only manifest itself as a hgiher percentage of shots going in and saves being made rather than more shots for and fewer shots against?
To put it another way: Alexander Ovechkin might be a phenom because he plays “loose” but he puts up points and helps win games because his line puts a hell of a lot more shots on net than their opposition. I’ve yet to see a player make a career out of getting brutally outshot every season but having a much higher On-ice SH% and SV% than the other guys.
Jonathan - “Looseness” in Ovechkin’s case (or his line’s for that matter) might be exactly as you describe. Difference between him and normal NHL players is that he can replicate his ultra high level performances almost on a nightly basis. Thus he might also be classed as a phenom. His elevated mental state might allow him to shoot more than “normal” NHL players because he can get to more open places, thus the positive shot differential.
Brutal shot differentials might mean that the goalie was playing out of his head too, being part of the “loose” frame of mind. The shots that ARE generated on the bad end of the SD are loose too. Maybe their quality is higher. They just seem to find the net more frrequently than normal.
Like Tyler correctly stated, it’s not that the team played so much better last year (and that was the mistake MacT and so many others made). It’s that they played looser, which as I tried to explain is a recognized cognitive state in high level athletics. Shots that normally wouldn’t go in suddenly do exactly that.
Looseness is not a quantitatively measurable variable, but it does exist and it is a factor in abnormal athletic performance.
You go one way (too much fear) and you have Shawn “Whiffcoff” 2009. You go the other way (lots of confidence) and you get the kid line 2008.
I’ll point out that too much confidence is just as bad as too much fear — you need a good balance of both — but in general, I can get behind what you’re saying here.
As for JW’s question, I believe that particular effect has to do with both accuracy and power being a function of muscular coordination, which improves as an athlete reaches his or her optimal level of physiological arousal (not that kind of arousal, you perverts) and focus. “Looseness” can bring you closer to this level or further away, depending on your initial state; in 2008, it’s certainly plausible that “looseness” did precisely that for many of the Oilers. Ovechkin’s success is more to do with better pattern recognition and game awareness/understanding than the average NHLer, though I would imagine this can also be affected by arousal and focus (it’s been a while now since my last sports psych course, so I’m not sure).
Thanks Doogie.
If I remember correctly, the “loose” state of mind in athletics does indeed provide optimum focus. The advantage being that focus is achieved on a subconscious level, where brain activity is at it’s highest. Something along the lines of “the 70% of your brain you don’t use” sorta thing.
Either way, the loose state in athletics is recognized, although seldom sustainable except in the very few cases I mentioned above and as it appears, Ovechkin as well. Thus the crash to earth of the 2009 Oilers.
@David: As with a lot of this stuff, it’s not an exact science, and very individual. My understanding is that you can be too loose or too tense, but in general, most athletes’ optimal state of mind is closer to “loose” than “tense,” and that the inverted-U hypothesis applies from that point.
Doogie: I wouldn’t say that “looseness” is invalid as an explanation of good results. I think the point here is just that that narrative as applied to the 07-08 Oilers is a bit of a stretch.
The point, I think, is that the total results they were getting in that time span, in terms of goal differential and shot differential, were not terribly far off from what they were getting all season. They just managed to transalte that into a statistically unlikely total of points. Now, the question isn’t “Can being loose help you win hockey games?”, which I tend to think is probably true, but “Can being loose help you win hockey games without improving your teams overall ability to score or prevent goals over the same timespan?”
I don’t know. Maybe there’s a context-specific explanation for why the totals end up the way they do. Some of this tied-leading-trailing context Bruce, Vic, and Tyler are talking about may provide something.
This is a fun argument though, both because it’s gotten legitimately interesting and because the debate is more framed by the methods MC used to argue than the argument itself. Rejection of stats is generally based on the idea that the only way to know what really happened is to watch it. But, from the media quotes from the time that MC or Vic (I think) put up somewhere (LT’s comment threads?) I don’t think many people who saw anything more than their record thought they were playing great hockey at the time. But a year later, no one asks how, they just ask how many and fill in how with the best story available.
Whoa. I turn away for a minute and the discussion gets downright metaphysical.
The point is moot, IMO.
If you choose to think of this in terms of sports psychology, then consider Tyler’s data in this way: The probability of an NHL team collectively achieving a higher mental state that allows them to win in spite of certain conventions about team performance is at least equally as rare as pure puck luck.
Either way, the results still indicate that stars were aligned. I suppose you could pick your poison as to whether it was puck luck or collective positive thinking. I know which one I would choose.
The probability of an NHL team collectively achieving a higher mental state that allows them to win in spite of certain conventions about team performance is at least equally as rare as pure puck luck.
That makes a ton of sense to me.
Consider: In the positive thinking scenario, the same positive thinking that results in making more shots would also result in making more passes, yes? More passes would certainly show up as a territorial advantage - i.e Corsi.
The same positive thinking would result in fewer miscues, yes? Fewer miscues would also translate into a territorial advantage, which we would see by measuring Corsi.
I really have trouble seeing an argument that makes the benefits of “looseness” apply only to making your shots but not anywhere else in the game, because the benefits of confidence would seem to apply all over. Passing and shooting are, after all, related skills.
Just to be clear, there was obviously some luck involved, I’m just agreeing with David S that “looseness” is probably a valid explanation of where some of it came from.
Exactly. I’m not disputing Tyler’s findings as much as the casual insistence that the large variance from the norm last year was due entirely to luck.
A mental state of looseness did not cause Robert Nilsson or Sam Gagner to increase their physical output as we can assume most NHL players are exerting to their limits a majority of the time (Dustin Penner excluded). What it did do is make the existing efforts exerted more efficient as a result of superior focus. Thus more shots go in.
At the end of the day, no doubt pure luck had a part in all this. I would say a loose frame of mind creating what looked like luck was as much of a contributing factor, if not more.
You guys are just fixated on the word “luck”.
Luck is random variation. Nothing more, nothing less.
Your proposed explanation in terms of the players’ frame of mind is just a contributing factor to that random variation. Along with how the pucks bounced and how physically well the players felt and how many times they tripped on ruts and how often they had Taco Bell within 60min of game time and…blah blah.
It’s all a little pointless because it’s really all luck and regardless it sure appears to be unsustainable which was the point in the first place.
On the contrary. “Luck” in this case is the generic term ascribed to the happenstance of anything some people cannot quantify numerically.
While random variation certainly played a part, it is not an adequate description of the events, as Tyler’s analysis concedes. Luck combined with heightened performance as a result of playing in the “loose” state would be more accurate. The players admit they experienced this state, along with first-hand observers who witnessed it. Just because you can’t chart the effect doesn’t mean it doesn’t exist. In fact, any coach of high performance sport would tell you that the state does in fact exist.
While guys like Robin might be writing for effect (they are writers after all), it seems rather narrow-minded to dismiss their point of view simply because you neither understand it nor are able to allow for it in your analysis. The plausibility of the state of looseness as it affects high-level athletic performance gives their description of the event credence.
David S
I won’t speak for anyone else, but when I say luck I mean chance variation. And I mean precisely that, at the exclusion of the items that you are talking about.
In fairness, it hasn’t been laid out above, though it seems intuitively obvious.
JLikens post on team EV shooting% calculates the expected level of chance variation and plots it against actual. The difference between these distributions is the expression of the ability distribution of the teams, and ‘looseness’, clutchness, etc live in that distribution. Of course those are very nearly binomial distributions so the math is fairly simple.
More commonly the ability distribution is nothing close to a gaussian distribution, although this is implicitly assumed in most of the baseball stuff I’ve read. That seems dubious to me, but those cats love their linear regressions, and it makes the math about 100 times easier.
Bill James’ stochastic models ar really sensible I think. Things like randomly shuffling all of Andre Dawson’s plate appearances for 1980 (or whatever) and checking to see if the random Andre’s are more or less streaky than the real Andre, and so on.
Also interesting to note that Dr Jim Albert, the mathematician who has done the most to crack the ‘chance in sport’ nut, he will use the term ‘chance variation’ in articles published in formal journals, and in the draft articles on his personal site he uses the term ‘luck’ far more often, in fact usually.
So in a Luck vs Ability plot for player OBP, batting average, SO rate, IP-HR-AVG or whatever … the clutchness, looseness, injury status, etc, that stuff is all contained in the ability distribution, NOT in the luck distribution (which is of course binomial or multinomial, depending on the model).
Simple enough, no?
@Jonathan: I read Brownlee’s comment in his article - and while I tend to think that the results cause the looser attitude rather than the other way around - I took it the way he seems to have meant for it to be taken: the Oilers put up wins like they were the 1957 Montreal Canadiens.
One of the many things I do at work involves interpreting insurance contracts. There’s a principle that insurance policies are interpreted against the insurer - if there’s an ambiguity in the policy, the interpretation that’s most favourable to the insured prevails. I’m used to reading things with an eye towards that and I honestly don’t know how you can get that from what Brownlee said. Like I said, the point of this was not to show that the Oil weren’t the ‘57 Habs but his piece of writing simply doesn’t say that they played like the ‘57 Habs.
On the contrary. “Luck” in this case is the generic term ascribed to the happenstance of anything some people cannot quantify numerically.
What Vic said. Question for the looseness people: can you show me any on-point academic stuff? I have a hell of a time understanding how this is measured. I mean, were Cogs and the boys saying that they felt loose when the streak started? Or was it when they were 10 games in and on a run that that’s how they explained? I’d like to see the research explaining this phenomenon.
While random variation certainly played a part, it is not an adequate description of the events, as Tyler’s analysis concedes.
I actually think that it is an adequate description of the events. The Oilers were at the right end of the tail, sure, but it’s not like they put up 35 points or anything. 29 points is a ton but not impossible.
Major fallacy is using goal differential to decide whether or not a team was ‘playing well’ or getting ‘full value’. I mean really how can you possibly statistically illustrate JUST using goal differential the quality of the teams play over a stretch of 20 games?
If you want to accurately assess team performance, then you need to look at a lot more numbers then merely goal differential. Playing well would mean for instance higher shots/60 rates versus players previous rates in the season, higher points+goals/60, better ga/60 versus the previous 60 etc.
As far as I can tell, the only relationship you can accurately determine with that single piece of information is that the oilers managed to win some close games.
I’m just going to say subjectively that I think for the record they did play well for that stretch of games, and mainly for two reasons, the play of the kids (who seemed to really mesh and offensively dominate B/C pairing defencemen) and even more telling the ease with which our 4th line was scoring goals during that period. Teams were getting beaten up so badly by stortini/glencross/brodziak down low (they seemed to be scoring every game for a while there) that they started drawing heat off of the kids (all of the forward lines actually), who responded by cashing in a good percentage of their chances. I’m curious if the stats would bear out my subjective perception of those games.
@Tyler: I’ve emailed my Sports Psych professor. I’ll let you know what comes of it.
The thing about shot differential is that maybe its just the system. Until you can split up the shots to show shots within 30 feet, 10 feet or 5 feet, it doesnt really matter to me. Early in the season the Oilers were getting outshot on a regular basis and I remember thinking, ‘just for one game can they out shoot the opposition’. They finally did for a couple games and still lost.
All I’m saying is that maybe its part of the system to let oppostion shoot as much perimeter shots as they want, the defense just clear the rebounds. Other teams like Minny, dont let you shoot from anywhere.
There are too many variables in hockey to just look at shots/goal differential to determine how a team is playing.
Given that a team like the Oilers, and most teams, directs about 4 shots in the general direction of the net for every one that actually arrives shots on goal is hardly a viable stat anyways. As well, it seems to me many of the best saves are actually on shots which wouldn’t count as shots if the goalie hadn’t actually made the save. Anyway, in any unrigged sport, all things tend towards 500, whether goals per shot, wins per game played or whatever. It would be more meaningful if save percentage on shots actually having a chance to go in could be calculated but how to be objective is the problem. Off hand, on nothing but experience, I would guess less than 5% of shots have even a reasonable expectation of going in.
Vic: Nice job on the Andre Dawson reference:)
Right now I’m reading “Baseball By The Numbers” and it’s a BP - Baseball Prospectus - work and it’s edited by a guy named Jonah Keri. Cosh would know this guy as well - we all used to post on the Expos portion of baseballboards.com - and later Fanhome - and Keri was what we’d call a goody-goody.
BTW, there were a lot of smart cats on that board. There was a guy named Scott who was a bit older than me and was from New Brunswick and he happened to be an Oilers fan as well. He would fit in great on the sphere because he was ahead of his time in terms of fans using stats. There was another guy named Sanders who lived in SoCal and he was a bright fucker as well. Another guy named Matt from Ottawa who was doing an Economics degree and once again, really bright.
Anyway, I was in my early to mid 20’s back then and Much more confrontational than I am now. Back then I would just bully and berate the hell out of fellows who didn’t believe in numbers and instead stuck to old axioms.
Anyway, Keri was a numbers guy as well but he was super nice and wouldn’t say shit if his mouth was full of it. Come to think of it, he got through to more people with his manner than I did;) which is why now I’ll still throw out liberal doses of sarcasm but I’m not as “in your face” as I used to be. Anyway, there’s still some reason why I can’t really trust or fully like someone who likes everything or everybody;)
Anyway, I’m going through the BP book and there’s a story about how Billy Martin once picked the names out of a hat for batting order and how much difference it made and what would be the optimum batting order. Anyway, I grew up on the Expos and that meant pitching and small ball and it’s the NL so you’re well used to the bunt. Later on, I started reading more and more and I became a fellow who lives and dies with the numbers. One of the things they speak out in the BP book is what would make for the best lineup and it’s all about putting guys with power and OBP at the top. I run my softball team and I believe it was back in ‘02 when we picked up a guy off another team. He could hit for average and he hit for the most power in the league; we play in the best league on the Island and I’d say we’re “C” calibre nationally. I’d say he was probably our third best hitter for OBP but I batted him leadoff all summer because of all the extra bases his homers provided. Well, guys just about went off their head because it wasn’t conventional. I had to explain it to my own fellows but opposing teams couldn’t get over it.
And last year I tried something else.
We play cash tournaments that are Super Series and runs-differential is the second tie-breaker; head-to-head being the first. So, what I started doing was having us batting in the top of the inning even when we won the toss because this would always allow us seven innings of hitting No Matter What.
Now, on my team there are two engineers, a doctor and two other guys who have three degrees apiece. And do you think I could convince these guys this was the best approach?
Not a chance.
Everyone liked the feeling of having “last bats” and was of the opinion that if you didn’t score in the top of the first, the other team had the game’s advantage because they had the last swings. I tried to explain to them that when there isn’t a home park per se - we always play at the same complex - there could be no decided home-field advantage and 21 outs were 21 outs and we’d have a better chance of scoring more runs and being better off come tie-breaker time if we chose the option of always having 21 outs.
Anyway, the point is that you just can’t convince some people of some things.
Even extraordinarily smart ones:)
Question for the looseness people: can you show me any on-point academic stuff? I have a hell of a time understanding how this is measured. I mean, were Cogs and the boys saying that they felt loose when the streak started? Or was it when they were 10 games in and on a run that that’s how they explained? I’d like to see the research explaining this phenomenon.
I think you touch on a great point here in that post-event rationalization often has absolutely zero basis in reality. That said, I want to frame my own perspective on “looseness” and especially “confidence” because I would suggest the latter is more easily measurable but represents the same family of success factors.
To me, anyone who suggests confidence doesn’t impact success in hockey is either made of stone or has never so much as picked up a stick and shot a tennis ball with it. That said, imagine a metric (call it CM) whereby “confidence” can be measured accurately among all NHL players and repeated throughout the year to give fluid results. What do you suppose the distribution of CM would be across teams? How much variation would you expect there to be from team to team and from time to time on the same team? If you plotted your expectations against actual approved psychological CM measures, do you figure there would be a whole hell of a lot of difference?
I don’t know what the consensus is but it is my own belief that confidence is randomly distributed throughout the NHL to the point such that its net effect on a whole season of NHL games is zero. I’m not trying to say it doesn’t affect how a player or team plays for a short stretch but with all the other factors that go into a hockey game, does anyone think that having a relative CM of 1.1 vs. a series of teams playing at 1.0 is going to translate to 14-5-1 with any consistency?
And yes, someone with a psych degree just claimed that psychological factors are probably negligible. No wonder I switched programs, eh?
I won’t speak for anyone else, but when I say luck I mean chance variation.
This has always been my interpretation of the word “luck” as used here on the net.
JLikens post on team EV shooting% calculates the expected level of chance variation and plots it against actual.
I understand the point you make to follow this but my question is how is “expected variation” calculated? Can two different events have two different levels of expected variation on a large enough sample size? Does the difference between binomial and multinomial outcomes for an event equate to more variation?
Bill James’ stochastic models ar really sensible I think. Things like randomly shuffling all of Andre Dawson’s plate appearances for 1980 (or whatever) and checking to see if the random Andre’s are more or less streaky than the real Andre, and so on.
This is the sort of thing that entertains me. To tie this back to Tyler’s post that I quoted to start, you just know that any and all streaks are going to be rationalized and in some batshit crazy ways.
Great read, both the article and the comments.
Although I appreciate the statistical point of view, it should be noted that when you compute 20 game segments like you have (overlapping, 1-20, 2-21, ect.) the outcomes are not independent events and so your probabilities are not correct. Not to say that it is not interesting, because it certainly is, but to say its a 1/40 chance based on this method is literally incorrect.