Tuesday, November 15, 2011

Can Rinne Earn All That Money?

The Pekka Rinne contract extension a couple weeks back remains the most interesting off-ice move of the season. It touches on a lot of things we have written about and discussed behind the scenes, most notably the role of luck and skill in results and whether it's better to build a team by spending big on goaltending or leave that money for the skaters - see Matt's thought-provoking post on the subject. The Predators are also in a peculiar spot as a team that has spent near the floor in the past but now will perhaps change gears and spend closer to the cap. This is all happening, and almost certainly related to, with Weber (RFA) and Suter (UFA) coming up as free agents this next offseason. Chase covered the cap/budget impact quite well last week. There is a lot to unpack here and we'll be revisiting this deal, looking at Nashville's situation and trying to answer more general questions like how much top goaltenders should be paid over the course of the season.

In this article, I will look at how his contract compares to those of other top goaltenders, at least those paid like one. The question at hand is how much he has to contribute for his $7M a year to be reasonable compared to other big-money goaltender deals. I am ignoring several key things like regression to the mean that might impact the overall value of goalies. In other words, I'm not looking at more general things like whether goaltenders as a whole are overpaid, or even if Rinne will be but rather what he has to do for us to say he's not overpaid relative to other top-dollar goaltenders.

Everywhere Is WAR

The nice thing about analyzing goalies is that we have a large number of discrete events and while there are teammate, opponent and rink/scorekeeper effects, the strength of a goaltender's performance pretty much boils down to how well they stopped the puck in different situations. Goaltender analysis is more similar to hitters in baseball than it is to skaters in hockey. The metric I'll use is WAR, Wins Above Replacement, which is very similar to the baseball stat of the same name. The idea is to look at how many goals the goalie in question gave up and compare that to what a typical replacement-level goalie (think free agent paid the minimum salary) would have allowed on the same number and type (ES, PK, PP) of shots. We can translate this number into wins to see how many wins each goaltender gave their team over what a replacement-level goalie would have.

I'm far from the first to use this method. As far as I understand it, GVT follows a similar approach for goalies. For a few other examples, Gabe Desjardins did something very similar two years ago over at Puck Prospectus and there was a fanpost on the subject by DoctorMyBrainHurts at Gabe's usual home, Arctic Ice Hockey. Philadelphia's goalie issues and the signing of Bryzgalov motivated some similar work by our friends Kent Wilson and Geoff Detweiler. With skaters it's rather more complicated, but for goalie analysis this approach is pretty clearly the way to go.

This is already one of the longest intros of all time so I won't go too far into detail about exactly how I calculated this. For replacement level, I took took the combined results of goalies that were not in the top 60 in games started for each season after the lockout. Another difference between my work and the others linked above is that I use 5.52 goals per win instead of the usual 6.0, which I feel is more accurate based on regressing league points on non-empty-net goal differential since the lockout. This warrants an article of its own, which I'll post later this week.

What does $4+ million buy these days?

Capgeek only goes back a couple years and nhlnumbers, which I used, only stretches back to 2007-2008. In those four seasons, we have a sample of 60 in which a goaltender had an annualized cap hit of $4M or greater. Here is a scatter plot showing the relationship between goaltender wins-above-replcement and the cap hit minus the minimum player salary. Something to note is that there isn't a very strong relationship between a goaltender's cap hit and how much value in wins he turned out to provide to his team. This is a sign that maybe high-price goaltenders as a group are overpaid, but I'll leave that for future work as it's outside the scope of this article.


The regression equation you see tells us what we should expect a goaltender to produce for a given cap hit over the minimum salary. The last two seasons the minimum has been half a million dollars. Going by that, here is what we should expect out of goaltenders in this range:

Cap Hit ($M)WAR
43.42
4.54.28
55.14
5.56
66.86
6.57.71
78.57

So 4 million dollars buys you about three and a half wins.

How productive must Rinne be?

And finally we are ready to answer the question at hand. How well does Rinne need to play for his contract to compare well to other high-dollar goalie contracts? Looking at the last row of the table, we see that producing a WAR of about 8.57 a year is about right. Here are his last 3 seasons, which comprise 167 of his 168 career starts:

SeasonStartsWAR
2009494.75
2010543.76
20116411.17

This averages out to a WAR of 6.56 per season, far below the expected 8.57. However, there might be two reasons to be optimistic. His number of starts per year has gone up each season and even on a per-start basis his WAR was substantially higher last year than the first two. While our gut instinct may chalk the latter up to random variance, the starts going up each year is obviously important since it's hard to provide value from the bench or IR. This raises two issues, how much he needs to play to get his WAR up to the 8.57 range and/or how much his save percentage might need to improve to do so. Let's consider those separately.

Rinne has faced an average of 24.35 shots at even strength, 4.32 shots on the PK and 0.67 shots with the Preds up a man per start in his career up to the current season. Based on his career save percentages in these spots (0.928/0.877/0.903) and the replacement group's (0.907/0.845/0.908) he has a WAR of 0.118 per start. To get to 8.57 for the season, he would have to start about 72 games a year! Keep in mind that this would be starting almost every game without seeing any dropoff in save percentage from his career average. Only Lundqvist has gotten close to that many starts and based on what Rinne has done the last three years I think we can all agree that this isn't realistic.

So, then, it would appear that he has to improve on his already high save percentage that most of us would guess is over expectation. How much improvement? Let's take his 64 starts last year as the jumping-off point. If he faces the same number and type of shots per start as he has in his career thus far that would be about 1,877 shots a season. A replacement-level goaltender would allow almost exactly 3 goals per start, or 192 goals for the season. At 5.52 goals per win, Rinne would need to concede about 47 fewer goals to be worth 8.57 wins above replacement. This translates to a save percentage of 0.923. Here is a table with all goaltenders with a career save percentage of 0.923 or above, minimum of 500 games played:

PlayerStartsCareer Save %




Just off the list is Dominik Hasek with a career save percentage of 0.922.

For his contract to be about about equal in value to other goaltenders making $4M a year or above, all Rinne has to do is play at the Hasek level for 65+ games a year for 7 years.

11 comments:

  1. Haha...love the touch of including the empty table.

    ReplyDelete
  2. Oh wow, ahaha that last line is astonishing. Great work.

    ReplyDelete
  3. I think that WAR/$ is really interesting, but WAR/(% of cap) would be more fair... I think that proportion of cap hit should be the currency of NHL post lock-out.
    Thank you for all the hard work!
    That was a great article!

    ReplyDelete
  4. Great post, I think it's clear to see that he's not going to be a bargain at that cap hit. But it wouldn't be all that horrendous if he just performs to the level of King Henrik over the length of the contract, right? I mean he'll be providing more value than Gomez, Lecavalier, Bouwmeester, Heatley, Campbell?

    ReplyDelete
  5. I feel like this is being looked at wrong on the 5.52 number. Yes, I understand that you can calculate the number of goals it takes for a win by dividing, but I think there is a much more accurate way to analyze this. Look at goal differential. Based on their styles of play, Phoenix and Nashville both require a tip-top goalie. They play grind-it-out hockey that results in 1-goal games a majority of the time (stripping out ENGs of course). This means for a goalie like Rinne the vast majority of the time it only takes 1 goal to equal getting at least a point in the standing, two to get two. If most of Nashville's losses are by one goal with an average minimum salary goalie, then Pekka's 0.5-1% better save percentage is astronomical in terms of team performance.

    ReplyDelete
  6. The Belgian Canucks Fan - I like your idea a lot. Cap% seems like a much better way to go, particularly with the player minimum staying flat, it's gone up 25K every other year, while the salary cap has gone up. One of the benefits of any long deal is that as time goes on it's better as the cap goes up and that's not reflected at all in cap hit above replacement. This definitely warrants further work.

    ReplyDelete
  7. To Anonymous:

    Is Mike Smith a tip-top goalie? Because he's winning in Phoenix.

    But just for fun, Nashville played 39 one goal games, ranking 18th in the NHL. Their winning percentage - 46.2% - ranked 22nd in the league.

    They were 26th in the league in number of games decided by 2 goals or less.

    So they played a below average number of one goal games and they lost an above average number of one goal games. Rather than fix that problem, they've locked up a goalie and took money away from the team that clearly needs it more.

    ReplyDelete
  8. 5.52 goals to win a game? not the preds. You should look at our win percentage when we score 3 goals in a game. its crazy. I have see soooo many 3-2 wins for the preds in the last 5 years its crazy.

    ReplyDelete
  9. Great analysis Jared. Not to beat a dead horse, but given his 167 starts over the last 3 years, we would need to regress his EV Sv% about 50% to league average (0.915), which would be 0.9215.

    ReplyDelete
  10. Patrick - yeah, like I said in the article I left that out but it's a big factor.

    Based on the comments here and on twitter and some other thoughts I've had since, I'm planning several articles for the next week or so on goalies. They will touch on:

    1. Why I used 5.52 goals per win and whether it should be lower for defensive-minded teams.
    2. Usage/Shots against. In this article I made the assumption that Nashville would be able to keep Rinne's shots against per game at about the same rate going forward. If they become awful, for example if their two best defensemen leave in quick succession, his contract becomes more reasonable.
    3. Regression to the mean.
    4. Cap percentage instead of cap hit over replacement.
    5. Whether big-money goaltenders are good value from a $/win angle. The numbers above seem to imply that they are but I think that's deceiving.

    ReplyDelete
  11. http://www.broadstreethockey.com/2011/3/1/2018705/a-peak-behind-the-curtain-how-do-numbers-get-analyzed

    At the team level, an improvement in six goals in goal differential is worth about one win, and the deviation between teams is pretty small. So, if you add a player who adds six goals to your team (not just scores six goals) you can expect to add a win.

    Also, it's hard to consistently be winning close games
    http://www.coppernblue.com/2011/10/30/2523507/clear-victory-standings
    since there's so much randomness involved.

    ReplyDelete