Sunday, November 8, 2009

Team Effects on Goalies

I've been playing around with my win threshold stat lately. Through the Hockey Summary Project and as a result of a new book by Sebastien Tremblay, we now have shots against data for every NHL season since 1954-55. That opens the door for a complete statistical analysis, both at the team level and at the individual goalie level.

Since the primary focus of this blog is team effects on goalies, that was the first area I looked at. Here are the correlation coefficients between each team's win threshold and save percentage (both figures adjusted for league scoring levels):

The Original Six Era (1955-1967): -0.545
The Expansion Era (1968-1979): -0.394
The Open Eighties (1980-1990): -0.306
The Talent Influx* (1991-1997): -0.133
The Dead Puck Era (1998-2004): -0.081
Post-Lockout NHL (2006-2009): 0.014

Over most of the NHL's history, the teams that scored the most goals and prevented the most shots have also tended to have the goalies with the best numbers. It is only over the last two decades that we have seen increasing independence between team results and goalie results. That is not to say that there is none of that in recent years, simply that when looking at the overall picture we would expect more of a team impact on the numbers of a goalie playing in the 1960s or 1970s than on a goalie playing today.

Part of this could have been that the better teams had more of a tendency to develop or acquire the top goaltenders in the past. However, the evidence to me suggests that shot quality effects are mostly determined by differences in skill rather than differences in style of play, and therefore shot quality effects are going to be largest in an unbalanced league with large differences in skill between the top and bottom teams. That is supported by the data above, since in today's salary-capped league we don't see the same degree of goalie/team stat dependence, whereas in leagues that were more unbalanced because of factors like expansion, territorial rights or management competence, the goalie and team numbers are far more intertwined.

I have a few posts coming up on the topic of win thresholds, including which goalies had the lowest and highest career numbers, the importance of goaltending through different periods of league history, and how well win threshold numbers predict playoff results.

(*-I'm planning to look at a few different metrics broken down by era, and I wasn't really sure what to call the transitional period between the high-scoring 1980s and the low-scoring late '90s/early '00s. I'm not aware of any term that is in common use to describe the period, but if there is hopefully somebody can let me know. In my opinion the most defining part of the early 1990s was the migration of new talent into the NHL that really broke the Canadian dominance of the league. More Americans started playing the game at the highest level, while the fall of communism and escalating player salaries also attracted more of the top European players. At the same time, there was an overall improvement in the level of goaltending around the league, with a new generation of goalies breaking in and mostly displacing the generation before them.)

21 comments:

Kent W. said...

low scoring period of the 90's-00's = "the dead puck era" I believe.

The Contrarian Goaltender said...

I meant the period before the "Dead Puck Era" from about 1990 to 1997, when scoring levels were still relatively high.

Anonymous said...

Here is a question: How many goals would Teemu Selanne in his rookie year (76 G; 1993) have scored in the high-flying eighties, the Dead Puck Era, and post-lockout?

My guess is 90+ goals in the 80s, 50+- in the DPE, and probably in the high fifties today.

overpass said...

Good stuff. It does appear that past save percentages were more dependent on the team.

Since win threshold is a combination of goals scored and shots against, have you run separate correlations for each of those components? Also, have you tried regression analysis to estimate the magnitude of the team effects?

The Contrarian Goaltender said...

Overpass: I haven't done much beyond compiling the data at this point, but those are good suggestions. I'm still trying to develop a good method for estimating EV SV% for historical results, and that will likely require some regression analysis.

I can give you the correlations for goals scored vs. save % and shots against vs. save %:

Goals scored/Save %:
1955-1967: 0.590
1968-1979: 0.405
1980-1990: 0.313
1991-1997: 0.199
1998-2004: 0.086
2006-2009: 0.070

Shots against/Save %:
1955-1967: -0.279
1968-1979: -0.265
1980-1990: -0.114
1991-1997: +0.025
1998-2004: -0.066
2006-2009: +0.120

The Contrarian Goaltender said...

Wow, I wasn't expecting a question about Teemu Selanne. I think there are a lot of other people who would be better at answering that question than me, but if you want my opinion I'd say he scores less than 76 in the early '80s and 50-something today.

That year (1992-93) was a weird season, the top players in the league all had inflated stats because the season went from 80 to 84 games played and there were a couple of very weak expansion teams. Selanne played on an offensive team in an offensive division with pretty good linemates and a top offensive defenceman (Phil Housley) behind him, and he probably got loads of ice time as well.

If all the situational factors were in place maybe Selanne scores 80+ in the early '80s, but if you throw him on some random team I figure he would probably end up in Mike Bossy range.

Anonymous said...

OK, let's say 72 or 74 goals instead of 76 to compensate for the two to four extra games played compared to other eras. I think he would score more in the '80s compared to early '90s if for no other reason than the greatly inferior goaltending of the '80s.

I saw a Youtube video of Gretzky arguing that Teemu and Paul Kariya would have outperformed him and Jari Kurri if they had played in his heyday. I will try to refind the link.

Anonymous said...

Here is that video I mentioned:

http://www.youtube.com/watch?v=q0CrARZLocc

Wayne Gretzky: "I know for a fact that if they (Kariya and Selanne) had been playing together and it was 1982, that they definitely would have broken the records."

Let's discuss what I consider Selanne's best year, 1999, in which he scored 47 goals and won the Rocket Richard. He missed 7 games that year, and had at least one goal that should have counted waived off (for standing in the crease against, ironically enough, Ed Belfour). Also Paul Kariya was not quite as good as in previous years owing to the Gary Suter hit. Assuming he played the full 82 games and has a totally healthy Paul Kariya, I can easily see Selanne scoring 55 goals in 1998-99, which would have been a remarkable achievement for the era (this was Hasek's peak year, for instance).

In comparison to the Gretzky heyday, don't forget that Selanne and Kariya were the only superstars on an otherwise very mediocre team. Their centerman (Steve Rucchin) was never more than a second-liner, for instance, and the other Mighty Ducks players were much worse. What happened with the Edmonton Oilers in the early/mid-'80s will never happen again (five superstars in Gretzky, Kurri, Messier, Anderson, and Coffey all playing together). If Gretzky had only played with Jari Kurri and no other star players, his point totals would have been a lot less than they were. Likewise, if Selanne and Kariya had had a real first-line centerman (even the long-past-prime Gretzky of the late '90s would have counted), they would have scored a lot more--let alone a real offensive quarterback defenseman.

Triumph said...

will never happen again? it's already happening in washington with 4 legitimate superstars - ovechkin, green, backstrom, and semin. i suspect we'd notice it a lot more if the league's talent disparity was where it was in the 80s (which is borne out by TCG's findings... keeping it relevant)

in 92-93, the league had 3 teams with a combined goal differential of -476. this essentially meant that non-expansion teams had a goal differential of 317 to 294. that's an advantage that cannot be overlooked. it is also worth noting that selanne has not only not matched the SOG/G of his rookie season, he's not even come close to matching it.

transplanting players from era to era and making wild claims about their ability isn't really a fruitful exercise. yes, selanne probably would've been the best scorer in the 80s, but even the worst NHL player would've been dominant in the 1920s. selanne's game is no doubt influenced by offensive changes that came about during the 1980s - to then throw him back in time and say he'd be the best isn't really a worthwhile argument.

Statman said...

For a simple goal-scoring across era estimation, just:

player goals per game that year, divided by NHL average goals per game that year, multiplied by NHL avg goals per game of the hypothetical season, multiplied by whatever games played you think the player would've played that hypothetical season.

Of course, my spreadsheets with these stats are on a different hard drive.

Statman said...

Very interesting column, by the way.

Anonymous said...

Triumph, you are correct--Washington is shaping up to be an Oiler-like team. So let's use the example of Ovechkin rather than Selanne. How many goals would Ovechkin in his best year so far (65G; 2008) have scored in the '80s or the Dead Puck Era?

If any hockey player today is of true Wayne Gretzky-in-his-heyday caliber, it's gotta be Ovechkin. I can see him scoring 100 goals/year in the 1980s.

The Contrarian Goaltender said...

Here's the problem I have with the claim that Selanne would have scored 90+ in the early '80s. Let's review the top 5 goalscorers in 1992-93:

1. Selanne 76
1. Mogilny 76
3. Lemieux 69
4. Robitaille 63
5. Bure 60

If you're projecting Selanne to 90+, that means you're projecting Bure to 71. Is it likely that 5 guys would top 70 in the same season in the early '80s? Jari Kurri's career high was 71 goals, playing on Gretzky's wing on the best offensive team in the league in a high-scoring era. Mike Bossy never scored 70 goals once in his entire career, and yet 5 guys would beat him, 4 of which would be universally considered to be inferior players relative to their eras?

I agree with Triumph, it was a situational thing that year. Look at Selanne's shots in 1992-93, he got over 100 more shots that year than any other season of his career. Something must have been going on, and whether it was that he was getting tons of ice time or the other teams were bad or his teammates were really feeding him the puck we can't necessarily assume it's going to translate to different eras in exactly the same way. Most of the time you can run some league average adjustment that is going to be a good approximation, but I think it breaks down somewhat for 1992-93.

Gretzky was probably talking about it in the literal sense, i.e. if you put Kariya and Selanne in a time machine and drop them off in 1981-82, then they would score more goals than Gretzky. Maybe they would have, but who cares since that's an artificial scenario and has more to do with overall changes in the game than the skill of either player. If you introduce any sort of era-independent baseline, whether that is natural talent or performance compared to average or number of high scoring finishes or whatever, I don't think either Kariya or Selanne beat Mike Bossy. So it doesn't make sense that they would have theoretically scored more goals than him back in the day, does it?

Statman said...

Maybe more PP's in 92-93 than most seasons prior to that?

Statman said...

Selanne was also a rookie in 92-93; sometimes rookies take the league by surprise. That probably has something to do with that season of his.

overpass said...

Thanks for the further numbers, CG. Interesting that goals for appears to be the larger factor.

Regarding 1992-93, I read an article from that season that quotes Scotty Bowman on why Lemieux was having such a great year.

"(Lemieux) plays a lot," Bowman said. "He averages between 30 and 35 minutes, and when you play that much, it's hard to come out the next day and run through practice."

Bowman has been able to give Lemieux breathers during games. The NHL changed its format for TV timeouts this season, using four, 70-second timeouts per period, instead of seven 30-second timeouts.

"You can play your top guys a lot," Bowman said.

The NHL's new rules also have helped Lemieux, who isn't being held, hooked or interfered with as much as he used to be.

"With the new rules, it's a lot better for great players, er, good players to go on the ice and do their jobs," Lemieux said. "There's not much interference, hooking and grabbing. It's better for the players and the fans."


The obstruction crackdown almost certainly played a part, and possibly the change in TV timeouts allowed top players to get more ice time.

Anonymous said...

So you are arguing that Mike Bossy was a better player than Selanne? Was he also better, era-adjusted, than Gretzky?

In any case, I already conceded to Triumph that the Ovechkin example is a better study tool for the eras than Selanne. Ovechkin is probably more talented than Teemu, even in his prime, and is the closest analogy today to Gretzky and his world-class line. If he scored 65 goals in 2008, a relatively low-scoring era, he would have certainly topped Teemu's 76 in 1993, and I would not be surprised to see him reach or top 100 in the 1980s.

seventieslord said...

Nice work.

It was I who posted those numbers over at hfboards.

I meant to email you becuase I knew you'd want to do something with them... but I forgot. Looks like you didn't take too long to find 'em.

Anonymous said...

Was Lemieux really averaging over 30 minutes a game? That is ridiculous. Have other star players averaged that much in the past? They certainly don't today, not even star defensemen. If not that might explain why Mario's career +/- isn't quite as good as you would expect from someone so dominate.

Anonymous said...

The word is DOMINANT, not "dominate".

overpass said...

Was Lemieux really averaging over 30 minutes a game? That is ridiculous. Have other star players averaged that much in the past? They certainly don't today, not even star defensemen. If not that might explain why Mario's career +/- isn't quite as good as you would expect from someone so dominate.

My sense of ice time quotes like this are that they are usually a little higher than the data suggests. Lemieux may have averaged 30 minutes a game, but not 35. That would include playing almost all available power play time and a spending a lot of time killing penalties.

Regarding Lemieux's career +/-, there are a few reasons that it isn't as high as you might expect.

1) High ice time, as you note, means that he has more goals against per game than almost any other forward. Gretzky's goals against are similar, for the same reason.

2) Much of his scoring came on the power play - at even strength he was closer to Jagr than Gretzky as a scorer.

3) He had very little support from his teammates for most of his career. His teammates GF/GA with him off the ice over his career was an average of 0.86 - only Marcel Dionne had less support among star forwards.

4) He played hurt a lot of the time and was less effective. See 89/90, 90/91 and 93/94 - he was playing hurt the whole time in all three seasons and was a combined -12.