The Pronger Effect: How much impact can one player have?
Eric T.
May 17 2012 07:36AM
One of the most eye-popping things I have seen in non-traditional statistics was the difference between how the Flyers performed last year with Chris Pronger and without him. The plot below shows what fraction of the shot attempts the Flyers got (their Corsi score), which correlates extremely well with puck possession, zone time, and scoring chances.
When Pronger was in the lineup, the Flyers were getting 53.2% of the chances -- a typical number for a top-tier team. When he was out of the lineup, they were below average, a trend that continued into this season.
Certainly, Pronger is an important player, but could one player really make that big of a difference? It seemed impossible -- the difference between being a 48% Corsi team and a 53% Corsi team is huge. Teams in the neighborhood of 53% Corsi this year included the Red Wings, Kings, Blackhawks, and Bruins; teams in the neighborhood of 48% included the Hurricanes, Ducks, Sabres, Oilers, and Blue Jackets.
I don't believe that if the Blue Jackets added Pronger they would suddenly be a championship contender, so simple variance had to be playing a role here -- over a small sample size, random chance gives you some results that are more extreme than what the true effect will show in the long run.
But how big is the effect really? To answer that, I started making a list of very good players who missed somewhere between 20 and 60 games in a season recently, giving us a sample of at least 20 games with them in the lineup and at least 20 with them out of the lineup. For each of those players, I pulled the team's Corsi score with and without the player in the lineup. To eliminate score effects (the tendency of teams to go into a defensive shell to protect a lead), I used Corsi close, which only counts the shots when the score is within one goal in the first two periods or tied in the third.
Here's what I found:
Obviously this is not a comprehensive list of injuries; I was just skimming through a list of games played and looking for the names that popped out at me as being important players.
Still, the trend seems clear: there is some variance over a ~20-40 game span and Pronger's impact probably was indeed overstated, but adding a top-tier player to the lineup does seem to result in a roughly 2% boost to a team's Corsi score -- roughly the difference between the Sharks and the Avalanche or between the Capitals and the Canadiens.
An elite talent may not be an absolute necessity in hockey the way it is in basketball, but the upper-tier players still have a big impact on the team's results.
Follow @BSH_EricT and @NHLNumbers on Twitter.
Recently by Eric T.
Great stuff. Given how it seemed Pronger dragged the Oilers to the the final, have always wondered if his impact was bigger than some of the rest. Nice to see some hard numbers.
I find the drop off with Koivu IN the lineup to be quite surprising. He was in the lineup for much of the time when they enjoyed their much talked about unsustainable success and out of the lineup for chunks of the end of the season when they just flat out weren't that good all around.
@Derek Jedamski
Goal differential definitely tilted more against them with him out of the lineup, for a reason I didn't list:
PDO with him was 1012 (93.9% Sv%, 7.28% Sh%) PDO without him was 975 (92.9% Sv%, 4.58% Sh%)
(The above numbers are specifically for the close situations, like what's used for Corsi close in the article.)
@Phil S
Yeah, I think we probably underestimate the impact of the cascade effect (add a #1C and now your old #1 gets to play #2 competition, your old #2 gets to play #3 competition, your old #3 gets to play #4 competition, and everyone's numbers improve).
I came away from this wondering if top players are actually underpaid.
It seems like a player who can single-handedly boost your team from bad to average or from average to very good might be worth more than 10% of your cap.
I think they improved towards the end of the season because they just couldn't stay that low, right?
I think the cascade effect has something to do with it.
A team built without a superstar C will still try to have 4 Cs who can play in the NHL.
A team with a superstar C will also have 4 Cs who can play in the NHL.
Losing that superstar means replacing him with an AHLer while a team built without him replaced him with an experienced bottom-6 guy.
So basically, what you get there is the effect of losing a superstar to injury. You wouldn’t have the opposite effect by signing one as a free agent since the other option (cheap vet) is better than the AHL replacement (usually).
The Pronger effect is easy for the average fan to see even without advanced statistics.
Every place he's been became noticeably better upon his arrival, both in penalty killing and over-all results, and noticeably worse upon his departure.
Either the guy has great timing or playing a big role in that himself.
How were the Flyers with him in the game but off the ice compared to out of the lineup? That would be complicated by his partners being top-pair guys which would make the out-of-game sample and not the on-the-bench, but it would be interesting to get some idea how much is him giving his teammates easier ice time.
@Jared
In Anaheim, the Ducks were a 50.7% Corsi team with Pronger on the ice and 48.6% with him off the ice but in the lineup. In Philadelphia (prior to this season), the Flyers were 52.3% with Pronger on the ice and 50.7% with him off the ice but in the lineup. That seems to square nicely with Eric's conclusion that elite talents have a ~2% effect on possession.
@Robert Vollman
Without advanced stats? I'd like to see that information. I don't doubt that he helps a team improve (and vice versa), but I find it hard to believe that one person can go into *multiple* organizations and have such an obvious impact.
Well I'm guessing he meant that non-fancystats people would draw conclusions about his value based on the fact that Edmonton, Anaheim and Philadelphia all went to the Stanley Cup Finals the year Pronger joined those teams and the Blues, Oilers and Ducks were all terrible after he left town.
Perhaps I wasn't clear.
I'm wondering what the effect is of shifting the d-men up a slot and giving everyone else (save his partner) tougher ice time. So how did guys Pronger never played with have their numbers change when he was in the lineup vs. scratched? Obviously this is impacted by depth, that would give a 7th d-man 3rd-pair minutes and someone off the top pairing top-pair minutes.
In other words, I think there is an on-ice effect, which you reference, but also an off-ice-in-lineup effect. I'd like to get at the latter.
Pronger's ability to move the puck out of the D-zone is remarkable. One of the best, if not the best, in the league at it. Combine that with a Chelios like ability to get away with penalties and you've got a major impact player.
@Jared Lunsford
In the 50 games that Pronger played in '10-11, the Flyers' Corsi close was 53.8% when he was on the ice and 52.8% when he was off the ice.
In the 32 games that he didn't play, their Corsi close was 48.0%.
So at least in that one instance, the majority of the effect was the cascading impact on other players when he was in the lineup.
I'll look at how universal this is in a follow-up article.
Re: Koivu - are there any effects from the Gilbert for Schultz trade there? With Koivu in they probably couldn't get much worse and they did trade for a guy who is pretty good territoriality IIRC.
haha. I'm pretty sure they wrote an exemption into the rulebook for him after the lockout.
That Pacioretty number is shocking, but how much of a difference did Gomez make this year?
@PopsTwitTar
Looking at team points-per-game in the seasons immediately before or after having Pronger.
The team penalty killing percentage is relative to league average.
Before Pronger arrived: 1.09 PTS/GP, PK% -0.69
First season with Pronger: 1.06 PTS/GP, PK%: +0.81
... ok Pronger didn't help out his teams point-wise, but apparently helped the penalty kill. Philly's really the one that hurt his average.
Pronger's last year: 1.07 PTS/GP, PK% -0.20
Year after Pronger: 0.90 PTS/GP, PK%: -0.48
... but it looks like teams really plummet after he's gone. And the PK gets worse.
In fact in every single one of the four cases the team got worse after he left, in both areas except Edmonton's PK was slightly stronger in 2006-07 than 2005-06.
Of course these are small sample sizes with a lot of other variables. So I guess my comment is essnetially worthless but since I was interested enough to check, maybe others would be too. :)
@Eric T.
I figured that was the case with the PDO. I guess I didn't realize they were actually better in possession numbers as the season wore on and it was simply (mostly) the PDO correction that caused them to fall.
Wow, that's interesting stuff. The Toews one is pretty odd.
I've been looking to find 'team' corsi stats, can't find them on Behind the Net. Where did you find them?
Cheers,
@Derek Zona
It becomes more believeable when you see how massive his possession rates were that year (lead leading). A big factor is also that he completed Montreal's 2nd line which was typically Moen/Pouliot-Gomez-Gionta for the previous year and a half. So with him playing the Habs had a break-even+ hardest minutes line (Kostitsyn-Plekanec-Cammelleri) and a massive possession 2nd line. Without him they basically broke even at best at the top of the lineup.
Gomez doesn't correlate all that strongly with good and bad possession play in Montreal this season but a large reason for that was that he was a brutal player under Cunneyworth while good on possession with Martin.
The magic player for Montreal a possesion this year was Gionta. He was there for the first 20 games were Montreal were even strength monsters running a very effective 3 line strategy. His absence largely co-incides with periods where Montreal was running only 2 good lines at best, post-Cammalleri trade it devolved into one line and Plekanec.