How Good is Your Power-Play?
by Hawerchuk on Sep 7, 2010 9:00 AM EDT
Email
PrintSkill-testing question: if a team has an above-average power-play efficiency, what's the likelihood that they're actually an above-average power-play?
Let's take an actual example from 2009-10 - the league-leading Washington Capitals (25.2% PP) vs the Detroit Red Wings, who were #2 in the league two seasons ago, but closer to average last year:
| PP% | |
| WSH | 25.2 |
| DET | 19.2 |
Washington had 313 power-play opportunities, while Detroit had 307. If we make no assumptions about underlying talent distributions, then the likelihood of Washington's PP being legitimately better than Detroit's is 96%. Do we really believe that?
We already know that you can't assume that an observed result reflects true talent. So we need to regress to the mean. If we simulate the last 22 years of team power-play performance - assuming every team is equally talented - we get a standard deviation of 2.01%. The observed standard deviation is 2.57%, so skill makes up 38% of the total observed performance:
| PP% | Stdev | Luck | Skill |
| 17.9 | 2.57 | 2.01 | 1.59 |
| 62% | 38% |
We can estimate Washington's and Detroit's true talent levels by two different methods - we can simply regress them to the mean by 62%, or we can start from a realistic underlying talent distribution (17.9% average PP%, 1.59% standard deviation) and update the distribution using the observed PP% from 2009-10 via Bayes rule. The results:
| PP% | Actual | Estimate |
| WSH | 25.2 | 20.3 |
| DET | 19.2 | 18.4 |
Now we estimate that there's only an 80% likelihood that Washington's PP was better than Detroit's. The regressed difference between the teams is eight goals vs 20 before the regression. That seems a bit more reasonable, especially since Washington kept their PP unit together over the last two seasons while Detroit lost Marian Hossa and Mikael Samuelsson.
Of course, we know more about these two power-play units than one year's worth of PP%. Detroit played in a tougher division than Washington, and faced tougher competition generally. It's possible that Hossa and Samuelsson were worth one extra win on the PP alone, but Washington's PP is definitely not eight goals better than Detroit's given the same opponents. The key thing to keep in mind is that even what appears to be the best PP in the entire league might not be much above average.
22 comments
|
0 recs |
Do you like this story?
Comments
this is a phenomenon i observed last season when new jersey’s power play was inexplicably #2 overall in the league in january.
i guess my next question is, how great is the true skill gap in a given season from best PP to worst PP, regressing both to the mean? the implication above is that the skill gap between best and worst is far smaller than the %s that separate them.
Bayes
Since you mentioned Bayes, I imagine you’re conceptually picturing two huge bags of marbles, one labelled Washington and the other Detroit. You’ve pulled 313 marbles out of the first, and 307 out of the other, and have pulled 79 black marbles out of the Washington bag and 59 out of Detroit’s, and now you’re trying to calculate the odds that the Washington bag actually has a higher proportion of black marbles than the Detroit bag, correct?
If so, why are you looking at other bags of marbles – can’t you just look at these bags, and look at how the marbles were drawn throughout the season?
Jackson – by simulation. You assume every team is equal and run simulations to determine the spread of PP% by luck alone. Then you look at the difference between the observed stdev and the luck-only stdev.
If teams can only expect a spread of nine goals from best to worst, do you really find Hossa and Samuelsson being worth 6 goals combined plausible? That would value the rest of Detroit’s PP at 3 if they belong at the very top end, and that low figure seems very unlikely (when compared to the supposed value of Hossa and Samuelsson). It seems like either the spread would need to be larger, or individual performers valued somewhat less.
On a related question, how much do you think teams overpay for PP performance? If the best PP in the league is only worth 9 goals more than the worst, it seems to me that most players wouldn’t have a big enough share for it to be worth paying them. For instance, if John Tavares and Evander Kane both needed new contracts, I have little doubt that Tavares’ 0.66 PPG would earn him a better salary than Kane’s 0.39. And yet, all of that difference (and more) came on the PP. At EV, Kane was actually a very similar scorer (a bit better actually, but his on-ice sh% was a bit better too). Would you say, then, that any extra money the Islanders pay Tavares is wasted, since it’s for performance that has only minimal impact?
Sorry, I’m smoking crack. Top-to-bottom is 13 goals. WSH to DET was ~8 goals. Two years ago, WSH was +13, along with DET. Last year, they were +13 and +5, respectively. Is having Samuelsson and Hossa on the 2nd PP unit worth 6 goals? I guess I ignored Detroit’s injuries too. So maybe Hossa’s worth half a win, and Samuelsson 1/4.
I don’t think a team should ever spend specifically for the PP. Forget that I’m not convinced that there are offensively-minded players which a given production at EV who are better able to drive the PP. The bigger issue is that it’s a rare organization that can tell the difference between results and talent over the course of half a season’s worth of PP opportunities. (The number of requests from NHL teams for 4-on-4 and 5-on-3 data that land in my email is evidence thereof.)
I’m guessing those requests aren’t plentiful! I suppose what I’m getting at is that teams would seem to be much better off not giving their RFA’s PP time, even if they think they’re the best options available, since there are likely other players on the team who can deliver similar PP performance, and if not, it doesn’t cost you that much in terms of goal differential (i.e., even if you’re awful, you’re still only about 6.5 goals below average). It seems like you could more than make up for the cost by limiting the money given to your RFA’s, who are under team control.
For instance, say Tavares scores 45 EV points and 30 PP points in his third year. The Islanders believe he’s the best option available, they play him, and he succeeds. Everybody wins! The Thrashers, meanwhile, think Kane would be a good option on their PP, but decide it would be prudent to use Rich Peverley instead, since he’ll be a UFA come contract time anyway. They then play Kane the same amount as Tavares at EV, and give him PK instead of PP. For sake of argument, say he scores 45 EV points and 5 SH/PP points. It seems to me that the Thrashers will end up way ahead of the Isles come contract time for a pretty similar EV player.
by Scott Reynolds on Sep 7, 2010 9:08 PM EDT up reply actions
Ah. I see what you’re getting at.
My experience is that a coach is going to do what he wants – if he isn’t allowed to use his personnel as he sees fit, he’s going to quit or you’re going to have to fire him. You could couch this argument in a “give veterans PP time because they’re more reliable” angle, but I can’t see a competing team knowingly scoring 1 goal less to save money in the future.
Better bet is to sign a guy when his PDO is really low!
I don't think it's just a coach thing
Would Tavares be likely to resign with the Isles if he was benched all year from the power play for someone inferior to him? After all, it’s only a change in behavior if the person getting the ice time is inferior to the guy you’re benching.
I’m guessing you get stuck with situations like Kessel where he refused to re-sign with the Bruins and demanded to be traded (depending on whose version of the story you believe). However, since you demonstrated, through your ice time allocation, that the player was unsuitable to the power play, you just lost yourself a chunk of the value of the player.
I believe this is one of those suggestions that may look good on paper, but ignores the fact that there are actual people involved who will respond negatively to this type of situation.
Atlanta got away with it last year without (to my knowledge) upsetting young Evander Kane (and why would he complain? they did, after all, bring him to the NHL is his first year of eligibility). I do see where you’re coming from though. The team would need to create a policy wherein “proven” veterans are given priority over younger players for PP time on a consistent basis, and young guys who show offensive talent are brought on to the PP slowly over the course of their first (and maybe second) contract. For instance, if Kane goes from about 20 minutes last year, to 50 minutes this year, and about 100 minutes in his last, I doubt he’ll be up in arms.
The real struggle, I think, is the coaching problem. You’d need someone who is on board with the idea of each season being in service to another in which you have a better chance to win. Coaches seem fine with this when it comes to “development” so I don’t think it would be impossible to get them on board. You’d just need to explain it as trading few goals now for more goals later, rather than trading few goals now for saving money later. Coaches trade some goals now for more later (as in, subsequent seasons) all the time.
by Scott Reynolds on Sep 8, 2010 11:38 AM EDT up reply actions
I assume that you put this together to see how much you were giving up with a power play unit of Ryder, Vrbata and Horcoff?
I also assume the spread of 2 wins between first and last is so small that the real key remains drawing power plays.
Editor of The Copper & Blue, and leader of The Cult Of Hartikainen.
Yes, penalty differential is the important point. I basically assume 100% regression to the mean for team PP performance even if (as we see here) that’s excessive.
Skill vs luck % for 5 on 5 play?
Hawerchuck;
Great post as always!
I understand the Math but don’t have simulation software.
Could you give me an approximation for skill vs luck on even strength play (over the last 22) years?
It would be interesting to compare to see if skill plays more of a factor 5 on5 or on the pp.
thanks Dan
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Certain photos copyright © 2012 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
