Power play metrics

[Church Hill]

I think we need a refined powerplay statistic. Here's what I invision:

Create a statistic for measuring how much better a powerplay makes a team's offense. To do this, find first the team's rate of goal-scoring per 2 minutes of 5 on 5 play, then find the rate of scoring per 2 minutes of 5 on 4 play (note: 2 minutes, regardless of if one powerplay ends in 10 seconds, then the next powerplay counts until the 1:50 mark). This should give a standardized appraisal of how much better a team's 5on4 play is than its 5on5 play. This would be helpful for analyzing just how effective the system, coaching, and players are in this situation.

Sorry if this was a little unclear I'm on my way out but let me know your thoughts or if need clarification

Click to read more...

 

[seventieslord]

I think we need a refined powerplay statistic. Here's what I invision:

Create a statistic for measuring how much better a powerplay makes a team's offense. To do this, find first the team's rate of goal-scoring per 2 minutes of 5 on 5 play, then find the rate of scoring per 2 minutes of 5 on 4 play (note: 2 minutes, regardless of if one powerplay ends in 10 seconds, then the next powerplay counts until the 1:50 mark). This should give a standardized appraisal of how much better a team's 5on4 play is than its 5on5 play. This would be helpful for analyzing just how effective the system, coaching, and players are in this situation.

Sorry if this was a little unclear I'm on my way out but let me know your thoughts or if need clarification

I think that scoring is more or less the same at even strength as it is in odd-man situations, except that one team's chances of scoring are essentially doubled and the other's are essentially negated.

 

[TheDevilMadeMe]

I think that scoring is more or less the same at even strength as it is in odd-man situations, except that one team's chances of scoring are essentially doubled and the other's are essentially negated.

This seems wrong to me. Isn't an increase in # of PPs considered a major factor in the scoring spikes of 1995-96 and 2005-06?

 

[Czech Your Math]

I think that scoring is more or less the same at even strength as it is in odd-man situations, except that one team's chances of scoring are essentially doubled and the other's are essentially negated.

That's correct, but not what most of us would guess.

This seems wrong to me. Isn't an increase in # of PPs considered a major factor in the scoring spikes of 1995-96 and 2005-06?

It does seem wrong somehow, yet it isn't. It surprised me to see that it's roughly equal.

It's actually the enforcement of the rules that creates both the increase in PPs and the increase in league-wide scoring. The increase in PPs helps the top line players more, since they get the PP time, so PPs are a major factor for most of the leaders' points totals rising.

I'll try to be more careful about my phrasing in the future, as I know I've attributed scoring increases to PPs while mentally including rules enforcement under that umbrella.

 

[Coconuts]

The increase in PPs helps the top line players more, since they get the PP time, so PPs are a major factor for most of the leaders' points totals rising.

This is an excellent point. I don't know how much of the difference is accounted by this, but I'm sure someone has done the analysis.

I've calculated the situational averages from the 2010-2011 season up to January 25, 2011, in units of goals per 60 minutes.

5-on-5: 4.71
4-on-4: 5.97

5-on-4: 6.89
5-on-3: 25.65
4-on-3: 12.32

 

[Czech Your Math]

This is an excellent point. I don't know how much of the difference is accounted by this, but I'm sure someone has done the analysis.

I've calculated the situational averages from the 2010-2011 season, in units of goals per 60 minutes.

5-on-5: 4.71
4-on-4: 5.97

5-on-4: 6.89
5-on-3: 25.65
4-on-3: 12.32

Interesting data, thanks for sharing that. I'm surprised the 4 on 3 and 5 on 3 rates are that much higher.

The differences in scoring for top players depends on the player really. The study I did gives a general for top players as a whole, but you also have to look at the individual players' data. For instance, a player like Lemieux is helped disproportionately by an increase in PPs, because he scored a higher % of his points on the PP than most other superstars.

 

[Coconuts]

Interesting data, thanks for sharing that. I'm surprised the 4 on 3 and 5 on 3 rates are that much higher.

The differences in scoring for top players depends on the player really. The study I did gives a general for top players as a whole, but you also have to look at the individual players' data. For instance, a player like Lemieux is helped disproportionately by an increase in PPs, because he scored a higher % of his points on the PP than most other superstars.

Blanket PP numbers such as those that I posted may be slightly misleading for some applications because the best offensive players are on the ice in those situations whereas they get a relatively smaller piece of the even-strength pie. In other words, 5-on-5 and 4-on-4 numbers may be watered down by 3rd and 4th line offensive talent. On the other hand, penalty-killing teams send out their best defenders, focus on playing defense almost exclusively, and are able to ice the puck without penalty.

A more informative assessment might require numbers on a player-by-player basis. I am sure the G/60 gap would be smaller in that case.

 

[TheDevilMadeMe]

This is an excellent point. I don't know how much of the difference is accounted by this, but I'm sure someone has done the analysis.

I've calculated the situational averages from the 2010-2011 season, in units of goals per 60 minutes.

5-on-5: 4.71
4-on-4: 5.97

5-on-4: 6.89
5-on-3: 25.65
4-on-3: 12.32

So... More goals were scored per minute of 5 on 4 than minute of 5 on 5?

 

[seventieslord]

So... More goals were scored per minute of 5 on 4 than minute of 5 on 5?

yes, close to 50% more... which is odd because I could have sworn in the past week someone else who knows their stats posted what I repeated yesterday.

 

[Czech Your Math]

yes, close to 50% more... which is odd because I could have sworn in the past week someone else who knows their stats posted what I repeated yesterday.

I thought you were right, based on ES vs. special teams gpg and the number of PP/game. I was assuming 2 min/PP (less when team on PP scores, more if 5 min PP), but forgot how often there are penalties that shorten both the original PP and (after 4 on 4) the resulting one.

 

[TheDevilMadeMe]

yes, close to 50% more... which is odd because I could have sworn in the past week someone else who knows their stats posted what I repeated yesterday.

I thought you were right, based on ES vs. special teams gpg and the number of PP/game. I was assuming 2 min/PP (less when team on PP scores, more if 5 min PP), but forgot how often there are penalties that shorten both the original PP and (after 4 on 4) the resulting one.

So... Conventional wisdom was right on this one?

 

[Czech Your Math]

So... Conventional wisdom was right on this one?

Yep, looks like it! Conventional wisdom is often right, but that doesn't mean it isn't ever wrong, nor that it shouldn't be backed up with fact if possible.

 

[Church Hill]

I'm just thinking about how long it would take to do a study like this manually... For each team, you'd have to go game by game to calculate total minutes spent up one man (accounting for powerplays ending prematurely) and total minutes spent 5 on 5 (removing time played 4 on 4, 6 on 5, etc). Maybe this would take 30 seconds per box score, but that would take about 41 minutes at that pace to do just ONE team's entire season.

It would still be interesting to see, but maybe if someone built a script to bring the data in we could get some concrete figures to talk about. I'm most interested in seeing of the team with the best PP% (as traditionally calculated) has the largest leap in 5on4 efficiency vs 5on5 efficiency, which I suspect is not the case.

 

[MadLuke]

So... Conventional wisdom was right on this one?

And easy math, power play situation would give a goal to each 10 minutes (take 20% (a bit high but let 2 full minutes by power play event the one scored and they often are in the first minute), that would be about 6 goals a game in the nhl, and you did not count the Pk goals againts yet and almost no team have 3gpg or more in the nhl today.

 

[Iain Fyffe]

Yep, looks like it! Conventional wisdom is often right, but that doesn't mean it isn't ever wrong, nor that it shouldn't be backed up with fact if possible.

Much analysis is in fact dedicated to testing conventional wisdom, not disproving it. Skepticism is in asking "is that true?", not in saying "that's not true."

 

[Czech Your Math]

Much analysis is in fact dedicated to testing conventional wisdom, not disproving it. Skepticism is in asking "is that true?", not in saying "that's not true."

You can't say it isn't true without evidence one way or the other. Whether one thought it was true, false or was uncertain shouldn't matter much, unless one is using hypothesis testing. Except, if the evidence is unclear, one is likely to fall back to one's original view on the matter.

 

[Iain Fyffe]

You can't say it isn't true without evidence one way or the other. Whether one thought it was true, false or was uncertain shouldn't matter much, unless one is using hypothesis testing. Except, if the evidence is unclear, one is likely to fall back to one's original view on the matter.

Indeed. Just commenting on the fact that analysts are often accused of naysaying conventional wisdom for its own sake.

 

[Coconuts]

This is an excellent point. I don't know how much of the difference is accounted by this, but I'm sure someone has done the analysis.

I've calculated the situational averages from the 2010-2011 season, in units of goals per 60 minutes.

5-on-5: 4.71
4-on-4: 5.97

5-on-4: 6.89
5-on-3: 25.65
4-on-3: 12.32

I just discovered I only had the data up until late January, so that isn't for the full season. I will have to rebuild the data-retrieving program I used, so it will take a while to get the right numbers.

 

[Czech Your Math]

I just discovered I only had the data up until late January, so that isn't for the full season. I will have to rebuild the data-retrieving program I used, so it will take a while to get the right numbers.

For 2012, using data from NHL.com:

5 on 4- 5.91
5 on 3- 20.71
4 on 3- 11.76

 

[Rutabaga]

I just discovered I only had the data up until late January, so that isn't for the full season. I will have to rebuild the data-retrieving program I used, so it will take a while to get the right numbers.

Yeah, i was wondering if i was wrong, or if there was something weird with the data i have...

5-on-5 : 2.28
4-on-4 : 2.65
5-on-4 : 5.84
4-on-3 : 10.1
5-on-3 : 20.7

I'm just thinking about how long it would take to do a study like this manually... For each team, you'd have to go game by game to calculate total minutes spent up one man (accounting for powerplays ending prematurely) and total minutes spent 5 on 5 (removing time played 4 on 4, 6 on 5, etc). Maybe this would take 30 seconds per box score, but that would take about 41 minutes at that pace to do just ONE team's entire season.

It would still be interesting to see, but maybe if someone built a script to bring the data in we could get some concrete figures to talk about. I'm most interested in seeing of the team with the best PP% (as traditionally calculated) has the largest leap in 5on4 efficiency vs 5on5 efficiency, which I suspect is not the case.

Rank | Team | 5-on-5 goals/60 | Pos. in the Standings |
1 | Pittsburgh | 2.91 | 4 |
2 | Boston | 2.88 | 7 |
3 | Detroit | 2.77 | 9 |
4 | Philadelphia | 2.68 | 6 |
5 | Tampa Bay| 2.62 | 21 |
6 | Chicago | 2.57 | 10 |
7 | Ottawa | 2.56 | 16 |
8 | Vancouver | 2.47 | 1 |
9 | Winnipeg | 2.39 | 22 |
10 | Phoenix | 2.38 | 11 |
11 | Nashville | 2.33 | 5 |
12 | Toronto | 2.32 | 25 |
13 | Buffalo | 2.31 | 19 |
14 | Rangers | 2.28 | 2 |
15 | Dallas | 2.27 | 18 |
16 | Washington | 2.27 | 15 |
17 | Montréal | 2.25 | 28 |
18 | San José | 2.18 | 12 |
19 | Anaheim | 2.15 | 26 |
20 | Edmonton | 2.13 | 29 |
21 | Carolina | 2.12 | 23 |
22 | New Jersey | 2.10 | 8 |
23 | Saint-Louis | 2.10 | 3 |
24 | Colorado | 2.06 | 20 |
25 | Columbus | 2.05 | 30 |
26 | Calgary | 2.02 | 17 |
27 | Islanders | 1.96 | 27 |
28 | Florida | 1.90 | 14 |
29 | Los Angeles | 1.84 | 13 |
30 | Minnesota | 1.71 | 24 |

Rank | Team | 5-on-4 goals/60 | PP% Rank | PP% |
1 | San José | 7.20 | 2 | 21.1 |
2 | Nashville | 7.19 | 1 | 21.6 |
3 | Edmonton | 7.15 | 3 | 20.6 |
4 | Vancouver | 6.93 | 4 | 19.8 |
5 | Islanders | 6.82 | 8 | 18.5 |
6 | Philadelphia | 6.77 | 6 | 19.7 |
7 | Pittsburgh | 6.67 | 5 | 19.7 |
8 | Florida | 6.25 | 7 | 18.5 |
9 | Colorado | 6.23 | 9 | 18.4 |
10 | Toronto | 6.20 | 10 | 18.4 |
11 | Winnipeg | 6.19 | 12 | 17.9 |
12 | Ottawa | 6.15 | 11 | 18.2 |
13 | Boston | 6.14 | 15 | 17.2 |
14 | Calgary | 6.09 | 13 | 17.7 |
15 | Anaheim | 5.90 | 21 | 16.6 |
16 | Carolina | 5.87 | 20 | 16.7 |
17 | Saint-Louis | 5.80 | 19 | 16.7 |
18 | Detroit | 5.74 | 22 | 16.1 |
19 | Buffalo | 5.66 | 16 | 17.0 |
20 | Washington | 5.59 | 18 | 16.7 |
21 | New Jersey | 5.42 | 14 | 17.2 |
22 | Rangers | 5.21 | 23 | 15.7 |
23 | Tampa Bay | 5.17 | 25 | 15.2 |
24 | Los Angeles | 5.06 | 17 | 17.0 |
25 | Minnesota | 5.04 | 27 | 15.1 |
26 | Columbus | 4.86 | 24 | 15.5 |
27 | Chicago | 4.84 | 26 | 15.2 |
28 | Dallas | 4.68 | 30 | 13.5 |
29 | Phoenix | 4.42 | 29 | 13.6 |
30 | Montréal | 4.32 | 28 | 14.3 |

Is it clear enough ?
The difference (standings-wise) between PP% and the goals scored per 60 mins is not that big, i thought we could have seen some differences in the rankings, but it seems that its not the case.

San José and Edmonton are improving with the same jump : +5.03 goals/60.
Then you have Nashville and the Islanders at +4.86, and Vancouver at +4.45. Over +4.00 are also Florida, Colorado, Philadelphia and Calgary.
Under +3.00 are Detroit, Rangers, Columbus, Tampa, Dallas, Chicago, Montreal and Phoenix.

 

[Church Hill]

Wow that's actually really interesting and gives me a new perspective on the game. Thanks for putting those numbers together. I'll drop in a bit later and look closer to see if there's anything hidden there!

 

[Coconuts]

Yeah, i was wondering if i was wrong, or if there was something weird with the data i have...

5-on-5 : 2.28
4-on-4 : 2.65
5-on-4 : 5.84
4-on-3 : 10.1
5-on-3 : 20.7

Two things I should note:

1. My data did not include empty-net goals, as I prefer not to include them in most analyses unless necessary.

2. I was counting goals in total, not just GF. So 4-on-5 goals are included in my 5-on-4 total, for example. This is also why my even strength numbers are about half what yours are.

 

[Rutabaga]

Two things I should note:

1. My data did not include empty-net goals, as I prefer not to include them in most analyses unless necessary.

2. I was counting goals in total, not just GF. So 4-on-5 goals are included in my 5-on-4 total, for example. This is also why my even strength numbers are about half what yours are.

Ok, i see now where the difference is. That was bothering me.

 

[Jason MacIsaac]

Why look at goals when shots for is better at prdicting future rates. The best stat one can look at is sh/60. The best PP teams are often near the top. The same goes for on the PK.

This site has a long long way to go before it is comparable to any of the true Saber sites. Some of these topics are nhl.com stat worthy.

 

[Bear of Bad News]

This site has a long long way to go before it is comparable to any of the true Sabre sites. Some of these topics are nhl.com stat worthy.

I guess the last two weeks have been a failure, then - let's shut down the site.