Power play percentage - how important?

I see a couple other “we appear to have time on our hands” threads. Here’s another: Seeing that roughly half of Princeton’s, RPI’s and (almost 60% of) Harvard’s goals this season are off the power play makes you wonder given their places in the constellation of college hockey power: Is power-play percentage all it’s cracked up to be? And are there better ways to measure the power play than the percentage you convert on? Right now, the three leaders in power play percentage are not exactly USCHO Top Fifteen mainstays:

1. Princeton, 27.9%, but 4-5-1 and quite likely due to be 4-7-1 by Sunday
2. UMass-Lowell, 27.0%, 5-5-1 and winless in Hockey East
3. Harvard, 25.9%, 5-3-1
(RPI is #7 21.6%.)
(Cornell is #18 at 19.6%)

I wonder if a good power play percentage might mask an otherwise poor offense. That matters because sooner or later you’ll come up against a team that’s good on penalty kill, your power play gets shut down, and if you don’t have a good six-on-six offense, you’re toast. Perhaps the effectiveness of your power play should also (not instead of) be measured against what percentage you have scoring in any non-PP two-minute stretch of the game. If you can put up 3 non-PP goals in a 60-minute game with no penalties, then your baseline scoring percentage for any 2:00 segment is 10.0%. Assuming I did the math right:

Princeton is 17x61=27.9% on power play, with 18 non-power play goals in the ~239 two-minute segments when there was no power play, a 7.5% nonPP effectiveness. That’s 1.8 nonPPG, 1.7PPG average per game. (This simplified calculation assumes 60 minute games or 30 2:00-segments per game less the time on power plays, assumes power plays run 2:00, and doesn’t yet factor in segments spent killing penalties. This is just a start. It’s simple and needs refinement, like I said.)

UMass Lowell is 20x74=27.0% on power play, 18x256=7.0% nonPP, 1.6 nonPPG/game, 1.8PPG/game.

Fair Harvard is 14x54=25.9% on power play, 10x216=4.5% nonPP, 1.1nonPPG/game, 1.6PPG/game. In other words, if Cantab opponents don't get penalized, it’s going to be ~54 minutes into the game until Harvard lights the lamp.

Cornell for comparison is 9x46=19.6% on power play, 21x224=9.4% nonPP, 2.3nonPPG/game, 1.0PPG/game.

Just as baseball has evolved to more sophisticated calculations than batting average and ERA, I wonder if there are more useful or meaningful stats for hockey. Power play percentage is obviously useful when the first line hops over the boards to have at it on a man up situation: You know odds are 1 in 5 Cornell’s going to score (although with our luck recently, if doesn’t mean if Cornell gets 5 PP opportunities we’ll see a goal). But maybe there’s something more sophisticated such as power play percentage beyond even-strength percentage. And maybe you want to focus only on your effectiveness against .500 or better teams, because it doesn’t really matter how you do against the patsies. You’re going to beat the sub-.500 teams (well, supposed to) anyway.

Or maybe it’s not PP percentage but number of PP goals in a game. A team that draws 10 PP opportunities a game and converts 15% does just as well as a team that draws 5 PP opportunities and has a dynamite 30% conversion rate. Or maybe not because when a team has a 30% PP conversion rate, maybe the opponent alters its style of play to cut its penalties from 10 to 5 but also stops checking hard.

I wonder if there’s a correlation of special team effectiveness on PP and man-short defense. About a third of the top 20 teams are on both. I suspect that would be hard to show. In comparison, about half the total offense leaders are also PP% top 20 leaders. At quick glance there are more Top Fifteen Poll teams in the PK leaders than in the PP leaders.

As game stats become computerized it’s going to be possible to slice and dice stats many more ways. Baseball has the advantage because there are more stats you can compile. But for hockey, maybe you can draw correlations over time as to the importance of power play, penalty kill, effect a PPG or MSG has on the other team’s ability to score for the next 10 minutes or remainder or game. The danger of course is that meaningless stats get tossed around all too readily. (McKee for instance has the worse save percentage of Cornell’s two playing goalies, doesn’t he?) The one I dislike is the one that says a team scoring first in the world series, or that holds a one- or more-goal advantage going into the final period, wins the majority of the time. That’s kind of self-proving because a team that leads at any point is more likely to win than the one that’s behind. Unless it’s the Red Sox Nation.

You may be onto something here. I like the idea of seeing how much a team's scoring rate (goals/minute) increases between man up and even strength situtations.

You can ask similar questions about goal tending statistics. Should't we be tallying the save percentage and goals against separately for even strength and power play situations?

Indeed, there's a lot you can do with statistics, although for hockey and even more for soccer, there's less than you can do for baseball, football, or golf, where there are individual plays and situations.

For the goalie you could break out stats for man-down goals vs. even-up. Maybe you could compare how the goalie does period by period. Or how he fares when he makes fewer than 15 saves in a game vs. 25 or more. You could break out how he does against breakaways (a judgment call as to what's a breakaway, but still trackable).

For the team (and for individual players), time on offense, time on defense, times they were called offsides, times they were on the ice when their team was penalized (the forwards are slow getting back so Charlie Cook gets called for a hold as he slows down the breakaway - the forwards should be charged for being involved in that mixup).

Maybe there's a distant future of automatic stats creation where a rafters camera tracks each player's position (hmm, RF ID tags in the helmets?), and tracking software notes who he passes to, who he gets a pass from, if he's offsides (called or not), etcetera. But if you really love stats, I guess you need to wait for baseball season. LGR-Sox.