TheStatican
Registered User
- Mar 14, 2012
- 1,731
- 1,512
I was looking for a way to come up with an estimate of a players special team ice time, both powerplay & shorthanded in order to make comparisons between players before TOI figures were taken. As we know the NHL did not start keeping track of a player's ice time until 1997-98 but there are some lesser known or used stats available to give us some approximation of how much a given player's PP & SH TOI was.
These stats include the following;
PGF - Power play goals for while player was on the ice(tracked since 1959-60)
PGA - Power play goals against while player was on the ice(tracked since 1959-60)
PP(labeled team PGF in my chart) - Total number of Powerplay goals scored by a team in a season.
PPA(labeled team PGA in my chart) - Total number of Powerplay goals scored by a the opposition in a season. For accuracy only power play goals scored in games the player played in are included for PP and PPA.
% PGF - Percentage of team powerplay goals player was on the ice for
% PGA - Percentage of opposition powerplay goals player was on the ice for
PPO - Powerplay opportunities for.
PPOA - Powerplay opportunities against. For accuracy only PPO's in games the player played in are included for PPO and PPOA.
PP TOI - Powerplay time on ice(tracked since 1997-98)
SH TOI - Shorthanded time on ice(tracked since 1997-98)
SPT TOI - Special teams time on ice(simply PP TOI+SH TOI)
The calculation I used are as such;
To estimate a players PP TOI; PP opportunities per game X % of team powerplay goals on ice for X 90 seconds
To estimate a players SH TOI; Opponent PP opportunities per game X % of opponent powerplay goals on ice for X 100 seconds
For some reason 90 seconds yielded more accurate PP figures, likely because when star players are on the ice more scoring happens and powerplays are shorted vs when less offensively gifted players are on the ice i.e. a teams second powerplay unit.
Here are some control calculations to see how accurate this method is using known TOI figures. I used these players because they averaged a lot of specail team time one ice(over 8 mins) and they played all 82 games, meaning I could use their team stats without subtracting totals from any missed games.
The PP accuracy is pretty good, penalty-killing not so much. Seems like individual performance makes a much more dramatic difference there. Inadvertently what the numbers also do is shows the difference between a players expected goals for on a powerplay and expected goals against shorthand verse their actual goals for and against in each situation.
For instance Ray Whitney was on the ice for 11 of the 60 power play goals scored against his team in 2002-03. Based on the fact that his team averaged 4.99 PPO against and assuming a 100 second powerplay length for each of those powerplays that means the average player would have been on the ice for 1:22 mins(82sec) of SH time per game. However Whitney actually played 2:03 mins(123sec) of SH TOI per game, exactly 50% more. The average player would've had 16 or 17 goals(16.5 is 50% more) powerplay goals scored against him. With him on the ice opposing teams scored far less powerplay goals than expected against the Jackets.
The one flaw in this calculation is that we don't know exactly how long each powerplay was. However powerplay length varies much less than any other factor and so it would change the results less than any of the other figures. The difference between a powerplay operating at a 28.57% efficiency and one operating at 11.48% efficiency relatively small. Also as we see below a team can have exactly the same powerplay efficiency and yet the powerplay length varies by as much as 3.2% and likely more.
Some confirmed numbers;
We could make additional adjustments factoring in powerplay and penalty killing percentages but without knowing exactly how long a the average powerplay and penalty kill actually was for a team these figures would never be completely accurate. It adds a lot of additional calculations for a marginal increase in accuracy when it comes to determining how effective a player was in these special team circumstances but seemingly no increase in accuracy for determining ice times.
These were the actual powerplay & penalty kill %
Based on the fact that more efficient powerplays trend to have shorter overall powerplay length times and an inverse effect on better penalty killing units I assigned each player an estimated PP & SH team length. It yielded the following calculations;
These results are far more accurate for SH TOI but are somehow less accurate overall than the results not incorporating estimated variances in PP & PK length;
But it does likely give us an even better approximation of a players expected vs actual goals for and against on the PP & PK;
The better than expected figure is in comparison with the average efficiency for a player on the same team at the player, not league-wide. The increase or decrease in goals is probably the more valuable metric as it gives credit for better performance over a larger number of opportunities. Also all these players played in 82 games so it's an equal "playing field" so to speak.
This could be seen as a sort of plus/minus stat for powerplay or penalty killing. But that also means it is flawed in similar way, in that even if you were on the ice for a higher than expected % of PP goals that could very well have been more so due to the performance of the other players around you on the ice at the time. Though it would be more accurate for penalty killing since there's only 3 other players around to effect a rating rather than 4. The stat also requires a lot of metrics that haven't been accurate tailed or tracked in the past such needing to know exactly how many power play opportunities(both for/against) occurred in games X player played in. And for absolute accuracy we'd need to know the length of those powerplays and penalty kills. At least we do have accurate TOI stats for all players now.
Useful or Nay... but still at least slightly interesting?
Edit* Another realization I just had was that in order to be a correct +/- figure of sort for the powerplay and penalty kill you would also have to include shorthanded goals scored for and against while on the ice and I'm not sure where I'd be able to find these stats though I'm sure it's something the NHL has tracked as it's easy enough to do so.
These stats include the following;
PGF - Power play goals for while player was on the ice(tracked since 1959-60)
PGA - Power play goals against while player was on the ice(tracked since 1959-60)
PP(labeled team PGF in my chart) - Total number of Powerplay goals scored by a team in a season.
PPA(labeled team PGA in my chart) - Total number of Powerplay goals scored by a the opposition in a season. For accuracy only power play goals scored in games the player played in are included for PP and PPA.
% PGF - Percentage of team powerplay goals player was on the ice for
% PGA - Percentage of opposition powerplay goals player was on the ice for
PPO - Powerplay opportunities for.
PPOA - Powerplay opportunities against. For accuracy only PPO's in games the player played in are included for PPO and PPOA.
PP TOI - Powerplay time on ice(tracked since 1997-98)
SH TOI - Shorthanded time on ice(tracked since 1997-98)
SPT TOI - Special teams time on ice(simply PP TOI+SH TOI)
The calculation I used are as such;
To estimate a players PP TOI; PP opportunities per game X % of team powerplay goals on ice for X 90 seconds
To estimate a players SH TOI; Opponent PP opportunities per game X % of opponent powerplay goals on ice for X 100 seconds
For some reason 90 seconds yielded more accurate PP figures, likely because when star players are on the ice more scoring happens and powerplays are shorted vs when less offensively gifted players are on the ice i.e. a teams second powerplay unit.
Here are some control calculations to see how accurate this method is using known TOI figures. I used these players because they averaged a lot of specail team time one ice(over 8 mins) and they played all 82 games, meaning I could use their team stats without subtracting totals from any missed games.
Player | Season | Gm | player PGF | team PGF | % PGF | player PGA | team PGA | % PGA | PPO | PPO/Gm | PPOA | PPOA/Gm | Actual PP TOI | Calculated PP TOI | Accuracy | Actual SH TOI | Calculated SH TOI | Accuracy | Actual SPT TOI | Calculated SPT TOI | Accuracy |
Richards | 05-06 | 82 | 60 | 81 | 74.1% | 23 | 72 | 31.9% | 485 | 5.91 | 391 | 4.77 | 6:34 | 6:34 | 100.0% | 2:18 | 2:17 | 99.3% | 8:52 | 8:51 | 99.8% |
Kariya | 98-99 | 82 | 79 | 83 | 95.2% | 24 | 60 | 40.0% | 378 | 4.61 | 387 | 4.72 | 5:44 | 6:35 | 87.1% | 2:53 | 2:50 | 98.3% | 8:37 | 9:25 | 91.5% |
Rolston | 05-06 | 82 | 48 | 77 | 62.3% | 16 | 55 | 29.1% | 454 | 5.54 | 436 | 5.32 | 5:01 | 5:11 | 96.8% | 3:16 | 2:19 | 70.9% | 8:17 | 7:30 | 90.5% |
Iginla | 05-06 | 82 | 60 | 87 | 69.0% | 24 | 80 | 30.0% | 478 | 5.83 | 508 | 6.20 | 5:24 | 6:02 | 89.5% | 2:50 | 2:47 | 98.2% | 8:14 | 8:49 | 93.4% |
Williams | 05-06 | 82 | 45 | 95 | 47.4% | 21 | 81 | 25.9% | 531 | 6.48 | 445 | 5.43 | 4:38 | 4:36 | 99.3% | 3:33 | 2:07 | 59.6% | 8:11 | 6:43 | 82.1% |
Whitney | 02-03 | 82 | 60 | 71 | 84.5% | 11 | 60 | 18.3% | 410 | 5.00 | 409 | 4.99 | 6:07 | 6:20 | 96.6% | 2:03 | 1:22 | 66.7% | 8:10 | 7:42 | 94.3% |
| Average | 5:35 | 5:53 | 94.9% | 2:48 | 2:17 | 81.5% | 8:25 | 8:10 | 97.4% |
The PP accuracy is pretty good, penalty-killing not so much. Seems like individual performance makes a much more dramatic difference there. Inadvertently what the numbers also do is shows the difference between a players expected goals for on a powerplay and expected goals against shorthand verse their actual goals for and against in each situation.
For instance Ray Whitney was on the ice for 11 of the 60 power play goals scored against his team in 2002-03. Based on the fact that his team averaged 4.99 PPO against and assuming a 100 second powerplay length for each of those powerplays that means the average player would have been on the ice for 1:22 mins(82sec) of SH time per game. However Whitney actually played 2:03 mins(123sec) of SH TOI per game, exactly 50% more. The average player would've had 16 or 17 goals(16.5 is 50% more) powerplay goals scored against him. With him on the ice opposing teams scored far less powerplay goals than expected against the Jackets.
The one flaw in this calculation is that we don't know exactly how long each powerplay was. However powerplay length varies much less than any other factor and so it would change the results less than any of the other figures. The difference between a powerplay operating at a 28.57% efficiency and one operating at 11.48% efficiency relatively small. Also as we see below a team can have exactly the same powerplay efficiency and yet the powerplay length varies by as much as 3.2% and likely more.
Some confirmed numbers;
| efficiency % | PP time in sec | Efficiency variance from 20% | Time variance from 100 s |
Pen's 95-96 with/ML | 28.41% | 93.7 | 42.1% | 6.3% |
20.00% | 98.4 | 0.0% | 1.6% | |
20.00% | 99.5 | 0.0% | 0.5% | |
20.00% | 101.6 | 0.0% | 1.6% | |
Pen's 95-96 without/ML | 11.48% | 108.8 | 42.6% | 8.8% |
We could make additional adjustments factoring in powerplay and penalty killing percentages but without knowing exactly how long a the average powerplay and penalty kill actually was for a team these figures would never be completely accurate. It adds a lot of additional calculations for a marginal increase in accuracy when it comes to determining how effective a player was in these special team circumstances but seemingly no increase in accuracy for determining ice times.
These were the actual powerplay & penalty kill %
| | PP % | PK % | Estimated PP length | Estimated PK length |
05-06 | Richards | 16.7 | 81.6(18.4) | 103.5 | 102 |
98-99 | Kariya | 22.0 | 84.5(15.5) | 96 | 104.5 |
05-06 | Rolston | 17.0 | 87.4(12.6) | 103 | 108 |
05-06 | Iginla | 18.2 | 84.3(15.8) | 102 | 104 |
05-06 | Williams | 17.9 | 81.8(18.2) | 102 | 102 |
02-03 | Whitney | 17.3 | 85.3(14.7) | 102.5 | 105.5 |
Based on the fact that more efficient powerplays trend to have shorter overall powerplay length times and an inverse effect on better penalty killing units I assigned each player an estimated PP & SH team length. It yielded the following calculations;
| | Actual PP TOI | Calculated PP | Accuracy | Actual SH TOI | Calculated SH | Accuracy | Actual SPT TOI | Calculated SPT | Accuracy |
05-06 | Richards | 06:34 | 06:48 | 96.6% | 02:18 | 02:35 | 89.0% | 08:52 | 09:23 | 94.5% |
98-99 | Kariya | 05:44 | 06:19 | 90.8% | 02:53 | 03:29 | 82.8% | 08:37 | 09:48 | 87.9% |
05-06 | Rolston | 05:01 | 05:20 | 94.1% | 03:16 | 02:47 | 85.2% | 08:17 | 08:07 | 98.0% |
05-06 | Iginla | 05:24 | 06:09 | 87.8% | 02:50 | 03:13 | 88.1% | 08:14 | 09:22 | 87.9% |
05-06 | Williams | 04:38 | 04:42 | 98.6% | 03:33 | 02:24 | 67.6% | 08:11 | 07:06 | 86.8% |
02-03 | Whitney | 06:07 | 06:30 | 94.1% | 02:03 | 01:36 | 78.0% | 08:10 | 08:06 | 99.2% |
| Average | 05:35 | 05:58 | 93.6% | 02:48 | 02:41 | 95.8% | 08:23 | 08:39 | 96.9% |
These results are far more accurate for SH TOI but are somehow less accurate overall than the results not incorporating estimated variances in PP & PK length;
| PP TOI average | Accuracy | SH TOI average | Accuracy | SPT TOI average | Accuracy |
Actual TOI | 05:35:00 | | 02:48:00 | | 08:23:00 | |
Calculated | 05:53:00 | 94.9% | 02:17:00 | 81.5% | 08:10:00 | 97.4% |
with PP/PK length | 05:58:00 | 93.6% | 02:41:00 | 95.8% | 08:39:00 | 96.9% |
But it does likely give us an even better approximation of a players expected vs actual goals for and against on the PP & PK;
Power Play | | Actual PGF | Team PGF | Expected PGF based on actual PP TOI | # of goals more than expected scored on PP | % better than expected |
05-06 | Richards | 60 | 81 | 57.9 | 2.1 | 3.4% |
98-99 | Kariya | 79 | 83 | 71.7 | 7.3 | 9.2% |
05-06 | Rolston | 48 | 77 | 45.2 | 2.9 | 5.9% |
05-06 | Iginla | 60 | 87 | 52.7 | 7.3 | 12.2% |
05-06 | Williams | 45 | 95 | 44.4 | 0.6 | 1.4% |
02-03 | Whitney | 60 | 71 | 56.5 | 3.5 | 5.9% |
Penalty Kill | | Actual PGA | Team PGA | Expected PGA based on actual SH TOI | # of goals less than expected scored against on PK | % better than expected | "Special Team +/-" |
05-06 | Richards | 23 | 72 | 25.8 | 2.8 | 12.3% | 4.9 |
98-99 | Kariya | 24 | 60 | 29.0 | 5.0 | 20.8% | 12.3 |
05-06 | Rolston | 16 | 55 | 18.8 | 2.8 | 17.4% | 5.7 |
05-06 | Iginla | 24 | 80 | 27.2 | 3.2 | 13.5% | 10.5 |
05-06 | Williams | 21 | 81 | 31.1 | 10.1 | 47.9% | 10.7 |
02-03 | Whitney | 11 | 60 | 14.1 | 3.1 | 28.1% | 6.6 |
The better than expected figure is in comparison with the average efficiency for a player on the same team at the player, not league-wide. The increase or decrease in goals is probably the more valuable metric as it gives credit for better performance over a larger number of opportunities. Also all these players played in 82 games so it's an equal "playing field" so to speak.
This could be seen as a sort of plus/minus stat for powerplay or penalty killing. But that also means it is flawed in similar way, in that even if you were on the ice for a higher than expected % of PP goals that could very well have been more so due to the performance of the other players around you on the ice at the time. Though it would be more accurate for penalty killing since there's only 3 other players around to effect a rating rather than 4. The stat also requires a lot of metrics that haven't been accurate tailed or tracked in the past such needing to know exactly how many power play opportunities(both for/against) occurred in games X player played in. And for absolute accuracy we'd need to know the length of those powerplays and penalty kills. At least we do have accurate TOI stats for all players now.
Useful or Nay... but still at least slightly interesting?
Edit* Another realization I just had was that in order to be a correct +/- figure of sort for the powerplay and penalty kill you would also have to include shorthanded goals scored for and against while on the ice and I'm not sure where I'd be able to find these stats though I'm sure it's something the NHL has tracked as it's easy enough to do so.
Last edited:
