Saturday, May 2, 2009

How The Red Wings Play to Win

NHL teams play to the score, especially in the playoffs. Late in the game, the leading team will take fewer chances and try to choke off the game, while the losing team will press for an equalizer and usually put as many pucks on net as they can.

The Detroit-Columbus series was an interesting showcase for how teams play based on the score, as Columbus didn't spend a single second of the series in the lead. That makes it easy to analyze the numbers, since we don't have to go through and remove any time that the Blue Jackets spent in the lead.

One thing that was striking in both my numbers and those at Hockey Numbers was that the expected goals totals for the Blue Jackets were very close to that of the Red Wings. This seems to indicate that the play was much closer than what observers were reporting. Often I tend to side with numerical evidence over eyewitness accounts, but on this one I think it is pretty clear that the eyeballs were seeing something that the math wasn't taking into account, namely that the Detroit Red Wings playing to win.

It has been well demonstrated that trailing teams tend to shoot more, and that teams with a large lead tend to shut it down somewhat to help preserve their lead late in games. Therefore, we would expect that Columbus' shot totals and expected goal totals to be overstated, and for Detroit's numbers to be understated. This turns out to be the case.

Let's break it down further. Detroit had a large outshooting edge in the first 2 periods (104-73), and shots were essentially even in the third (37-36). That is the kind of result we would expect to see for a talented team like the Red Wings against an overmatched opponent - take control early, score a few goals, and then kill the clock.

As I observed in my last post, point shots may be somewhat of an indicator of possession. Over the first two periods, Detroit had a 19-8 edge in point shots. In the third period, they took just 7 to Columbus' 5. That likely indicates that Detroit spent much of the first two periods in Columbus' zone, allowing their defencemen to get set up and fire at the net, but gave up that possession advantage in the final period.

Perimeter shots are similar in that they indicate possession, although they are more up to the shooter's discretion since they come from less well-defended areas and are therefore probably more likely to end up on net. In first periods throughout the series Detroit threw everything at the net, with 25 perimeter shots on Mason. In the second that dropped to 11, and in the third just 14. Osgood faced few perimeter shots the whole way, just 7 in the first and 10 in each of the second and third periods.

I would have expected Columbus to take more long shots when trailing late. I'm pretty sure that the Red Wings would show an increased perimeter shot rate when trailing late in games. Maybe the Red Wings' forwards were doing a great job of preventing or blocking shots. Or perhaps the Blue Jackets have different offensive tactics, and prefer to try to generate more high-quality scoring chances. They were actually fairly successful at getting high-quality scoring chances against the Red Wings, so if that's the case it might not have been a bad strategy, Osgood just managed to hold up better than expected in net.

In the third period Detroit seemed to be trying extra hard to limit those high-quality chances, and they were pretty successful in doing so. In the first two periods combined, Osgood faced 20 shots from the crease area while Mason faced 13. In the third periods, Osgood faced just 3 while Mason only had to deal with 2. This shows the tradeoff for the leading team: They reduce their chances to score as well as the opposition's. When up a goal or two in the third, however, this appears to be a favourable strategy.

Perhaps the biggest sign that Detroit wasn't trying as hard to score was that in the third period they very clearly stopped trying to shoot high. In the first two periods, 30% of the shots on Mason were high. In the third, just 17% were high. Osgood's numbers went from 14% to 11%.

These numbers show that both goalies were facing much easier shots in the third period. Taking into account both shot location and shot height, my estimate is that over the first two periods the expected save percentages for the two goalies were .915 for Osgood and .929 for Mason. In the third, those numbers rose to .926 and .945 respectively (.921 was the average).

Four games is a small sample size, so some of these results may have been random chance. However, they all seem to support playing to the score effects that have been observed in larger samples, so there is likely something significant there.

In summary, the shot charts make it look like Detroit and Columbus played a fairly close series, and that Mason played poorly while Osgood played well. The underlying numbers, however, suggest that the Red Wings were clearly the better team, and shut down their offence when they had the game in hand. If Columbus was able to keep the games closer, or if the Red Wings had decided to keep trying to score all the way up to the final whistle, the expected goals would have been much more tilted in favour of the defending champions.

11 comments:

JLikens said...

Yeah, I've been noticing the same thing -- that the expected goals figures tend to reflect more favorably upon the team that trailed for the majority of the game (relative to my perception of how things went).

I know that his expected goals figures from last year incorporated game score as one of the variables in his model.

As you're likely aware, this tended to produce some strange results whereby the team that played with the lead for the majority of the game would invariably have an -- often sizable -- advantage in expected goals (see PHI-WSH Game 2, PHI-MTL Game 1, DAL-DET Game 5, BOS-MTL Game 2, etc).

This year he's eliminated that variable from his model, perhaps for the above reasons. Of course, now it seems that the bias is working in the opposite direction.

I should note that I'm in no way knocking JG's work in the area -- his expected goals numbers are excellent and very informative when viewed in the proper context. It just seems as though his model may require further refinement, which is to be expected given the complexity of the subject and all.

Vic Ferrari said...

I only saw the first period of the first game of the CBJ-DET series, and the Jackets were dominant, a 2-0 lead at the intermission would have been a fair score.

Osgood was solid, but lucky too, a few shots that he didn't have a hope in hell on ... they just hit him. A couple were banged straight into his glove from point blank range, he didn't have time to even start to react, then he did the big windmill follow through thing. I freaking hate that.

Looks like the Jackets only managed 6 more shots on goal the rest of the game though, so I imagine the reports that DET dominated the rest of the way are probably fair.

Looking at Tyler's recent post o the subject, every team that is in the hunt for a playoff spot seems to average a similar 2.5% drop in territorial advantage (as measured by corsi+/corsi_total). I think anyways, he had it in jpg format, and raw numbers ... so very much 'back of an envelope' stuff.

In any case it's a pretty significant drop in territorial advantage, and appeared similar for
most teams.

In the short term a lot of anomalous stuff happens. Near the end of this year the Oilers, the team I follow, they managed to get early leads on superior teams a bunch of times. And in parallel universes those leads were short lived ... but in this one Roloson was hot, and they hung on for ages in spite of having their asses owned the rest of the way.

The local radio guy here is convinced that the Oilers play differently with the lead than other teams, I think he's dead wrong.

Like every other team in the league, the Oilers try to avoid giving up odd man rushes when holding on to a lead. So safer plays at the blue lines and the third forward always stays higher than the low defending forward of the opponent (usually the center, of course).

The Oilers have more young players so you're more likely to see a bad decision on the ice at any time. But there isn't anything peculiar about the way they play to my eye.

Marc Crawford does colour commentating for the CBC in Canada, and if you can get past the huge personal bias he has for and against some players, and the pitch of his voice, he's an interesting guy who makes a lot of good points. Or at least I think so.

Anyhow, he talks about staying higher than the puck (as opposed to the low opposing forward) when leading in the 3rd period. Fine brushstrokes I know, still, that's pretty aggressive hockey, but again looking at MC79s's list from a few weeks ago, VAN was about the same as every other (generally playoff-bound) team.

Maybe if the Vigneault years and Crawford years were separated you'd see a difference in the 'playing to the score' numbers (i.e. Crawford teams did seem more fun to watch, and I expect that they had a bit more possession with the lead and gave up a touch more odd man rushes with the lead as well). But I don't know.

Vic Ferrari said...

Also, CG, I don't know if you follow any Oiler blogs, but a guy has been tracking scoring chances over there and it's fascinating stuff.

Over the short term it's all over the place, but as the sample grows larger scoring chance (SC) and shots-direct-at net plus and minus find each other.

So that the SC% (% of scoring on-ice chances that happened at the good end of the rink) and Corsi% (same reasoning) ... they mesh with r=.90 by season's end. r=.95 for those players with 65 or more games, r=.7 for players with 30 or fewer games. Just by memory.

And there probably is some difference in scoring chance quality against, though if it exists it is too small to see with this team this season.

I mean if a line dominates a shift territorially, and gets a hard shot onto goal with traffic in front ... if the teammate in front whiffs on the deflection or tips it over the net, it's pretty close to a non-event in termsof expected goals. If he tips it high and on goal, then there is a good chance it will miss the goalie and hit mesh, so the expected goal ranking gets a big bump.

And I'm sure we can all think of a whack of similar scenarios. A lot of that stuff is in the hands of the gods in any one game. And the guys that domninated the shift, they're probably good enough to do it again several times through the series. Eventually they'll probably be rewarded.

Just generally I think that possession is the overwhelming measure of team quality at 5v5. And that odd man rushes are the other element. And I suspect they are spread out more evenly amongst teams than we think, our memories hold onto goals more than chances, or at least mine does.

Second periods always have more odd man rushes, due to the bench positions. And a useful shot quality metric should be capturing that.

Jlikens quoted a guy from Ryder's site whose algorithm did attempt to grab that. Unfortunately he only ran them for a couple of seasons and didn't account for home scoring bias re shot location.

with correction for that by JLikens, I think that his stuff repeated the best and correlated with save% the best. Though I think the std deviation for his stuff was only .003. So most teams had a strikingly similar shot quality. And if you removed short-handed icetime, esp 5v3s ... well their can't be much left.

And intuitively EVsave% does seem to follow goalies around a lot, so that makes sense.

Anonymous said...

Contrarian, even though it is clear and undeniable that Steve Mason underperformed woefully and that Osgood way exceeded expectations (as he continues to, for the most part, against Anaheim), as you yourself would say you have to consider the teams in front of each. Columbus played their hearts out, but they could not compete with four battle-hardened superstar lines over 240 minutes.

Also do factor in that Detroit received ENORMOUS help from the zebras: not only were a great many of their goals on the PP, but the time spent killing repeated penalties exhausts a team and makes them less likely to score even when their man is out of the box.

seventieslord said...

I would say Detroit helped themselves by taking significantly fewer penalties than their opponents. Unless you think the refs are biased.

The Contrarian Goaltender said...

Vic: I follow most of the stats blogs, which are of course mostly Oiler guys, so yes I have been looking at the scoring chance stuff and I look forward to seeing what you manage to tease out of the data.

The relationship between Corsi and scoring chances was certainly compelling in the Oilers sample. I agree with your basic premise that the team with the superior territorial play is going to end up getting more shots and chances and is more likely to score goals. For average teams I think that the shots numbers probably do a pretty good job of estimating the balance of play and expected goals for/against.

I think there must be some shot quality effect, however, even if it is just significant for a few outlying teams, to explain something like the persistently high save percentages of the Minnesota Wild goalies. Much of that is probably because of facing fewer power plays, which wouldn't affect 5-on-5. I do think style of play must have some effect, however, because we see some coaches have similar shot prevention/save percentage results across different teams.

I think there are a few variables that are affecting shot quality measures and save percentages that we still need to figure out.

We have done some work on playing to the score, that is obviously one of them, the outshooting effect is well-demonstrated but it isn't really clear exactly how the shot distribution changes for both the leading and trailing team compared to in a tie game.

It also makes sense that odd-man rushes would have an effect on scoring chance quality and goal scoring/prevention. That's why I have been intrigued by some of the shot charts that record shots as high or low. I would expect that high shots come disproportionately from shooters with the time and space to have an open look at net. As a goalie, I know that low shots from the slot are fairly harmless but a shooter that aims high and makes his shot has a very high scoring percentage. I think including that factor would make up for some of that difference, if there is indeed a variance between teams. Unfortunately right now the inconsistent tracking limits the usefulness of this data.

Another factor that I expect is at least somewhat significant is blocked shots. I have noticed a negative correlation between blocked shots and save percentage, and I think JLikens found similar results. This might be one of the style issues that is affecting some of the metrics like shot quality.

It could be that shots at net determine nearly everything, and these variables are simply affecting teams' percentages differently without having much overall impact on actual goal prevention. Or, I think more likely, they are resulting in some shot quality difference between teams, which is now being partially captured by shot quality metrics but surely can be further refined. For most of them the effect is probably not large, but I think we need to work on it to make sure we level the field for the Backstroms and Brodeurs of the world.

Anonymous said...

"seventieslord said...
I would say Detroit helped themselves by taking significantly fewer penalties than their opponents. Unless you think the refs are biased."

I will confess that I did not watch any games in the DET-CBJ series (just looked at boxscores), but considering what has been allowed to go on so far in the DET-ANA series (i.e Franzen getting a goal by throwing Hiller into the net, Kronwall breaking Ryan Carter's nose with an elbow, and the disallowed tying goal in Game 1 with 3 seconds remaining, for just three incidents), it sure does make you wonder, doesn't it? I seem to recall that Columbus took 11 penalties in one game, and that Detroit had 24 power plays to 13 for the Jackets.

Vic Ferrari said...

I think that the variance of goalie EVsave%, relative to their career average EVsave%, is about exactly what chance would dictate that it should be.

Not so for PPsave% though, where it looks like about 1/4 to 1/3 of the results come from something other than the goalie's historic ability to stop pucks.

So for Minny, they historically take fewer penalties and have a very high PPsave% to boot. So that would skew up the overall save% numbers a bunch. Plus it affects the shot quality numbers, because Ryder's colleague accounted for game situation (5v4 shots were given a higher value, 5v3 shots even moreso).

Are Manny Fernandez's EVsave% numbers in Minny different from his DAL and BOS numbers? How about Roloson? I don't know, but I suspect they are close.

And the PPsave% numbers are probably a fair bit lower elsewhere, though I don't know how good the PKs were on their other teams (aside from EDM, which on the whole has been average in Rolo's time here imo)

In short, when valuing goalies I just look at EVsave%, accept they if the sample os small there is a luck compenent that can be pretty big. You can use timeonice.com/binomial.html to calculate that btw.

And maybe credit a guy .003 or so for playing his career in MIN or N.J and up it .003 or so for playing most of his time in COL or T.B.

How does EV shot quality correlate with EVsave% in the league, anyways? Has anyone checked that?

The Contrarian Goaltender said...

Re: Roloson and Fernandez:

I have ES save percentage stats since 1998-99 from NHL.com. The non-Minnesota sample for Fernandez is pretty limited, but here you go:

Dallas: 488 SA, .920
Minnesota: 5534 SA, .924
Boston: 702 SA, .920

Roloson:

Buffalo: 582 SA, .916
Minnesota: 3054 SA, .928
2005-06: 941 SA, .911
Edmonton: 3929 SA, .916

2005-06 was of course split between the two teams. Most of that would have been Minnesota, so that likely narrows the gap somewhat. If I include that with the Wild sample it drops to .924. We'd also expect some age effects there, of course.

That's just two guys, but it makes it seem like there is some significant 5 on 5 save percentage effect from playing for Lemaire. I'd be very surprised if your .003 is a proper adjustment for playing in Minnesota, even for 5 on 5 play.

overpass said...

Regarding Vic's hypothesis that there are no team effects on EV SV%:

1. I'm not sure it's possible to find statistically significant variations in goalie EV stats by team to disprove this, especially considering that team quality can change quickly and make multi-year samples difficult to work with(see Boston).

2. That said, if there are apparent effects that aren't statistically significant and that match what we might expect without looking at the stats (Minnesota), that's reasonable evidence in favour of team effects.

3. If in fact there are no team effects on EV SV% large enough to detect using our methods, we can only generalize that as far as NHL teams in this league using the strategies they currently use. A league with less parity would likely allow us to find EV SV% team effects, and an NHL team that radically changed their style of play and assumed that EV SV% would remain constant would likely be in for a rude surprise.

In that sense it's probably not any kind of hockey truth, simply that the statistical methods we use make it difficult to find team effects on EV SV% in a league with a lot of parity and personnel turnover.

I do think it's likely that at least 20 out of 30 teams have no team effect on EV SV%, including Edmonton, but it's the outliers that are in question. It's hard to believe that Minnesota, Detroit, and the Islanders all have the same EV shot quality allowed. However, assuming no team effect on SV% at EV is probably fine for most analysis.

The Contrarian Goaltender said...

"I do think it's likely that at least 20 out of 30 teams have no team effect on EV SV%, including Edmonton, but it's the outliers that are in question. It's hard to believe that Minnesota, Detroit, and the Islanders all have the same EV shot quality allowed. However, assuming no team effect on SV% at EV is probably fine for most analysis."

I'll agree with this. I think it has been demonstrated from the scoring chance data that for a team like Edmonton shot quality doesn't make much of difference. Then again, they have been right around average in terms of shot quality over the last few seasons, and when I did a shot quality prediction for this season based on past averages I had Edmonton at 1.003, so in that sense the Oilers results aren't really surprising.

I would agree that most of the league is in the same boat. From the shot quality numbers I've seen, there has consistently been about 20 out of 30 teams who are +/- 5% from average in shot quality.

"A league with less parity would likely allow us to find EV SV% team effects, and an NHL team that radically changed their style of play and assumed that EV SV% would remain constant would likely be in for a rude surprise."

I've looked at numbers from historical seasons, international tournaments and junior hockey, and that certainly seems to bear this out. If all we are concerned about is evaluating performance at the NHL level going forward, then those other results don't really matter and if the NHL results seem to indicate little effect then that's good enough. The assumption wouldn't hold for evaluating past players, however, or players in different leagues.

I've said it before, but I think we have to be careful about drawing conclusions from evidence that might be more indicative of the parity at the NHL level rather than fundamentally describing how the game is played. On the other hand, most of the stuff Vic has demonstrated so well (that possession drives results, luck is a big factor, etc.) is undoubtedly also true at other levels of hockey.