Spreadsheet with adjusted and situational stats

[overpass]

A few years ago I put together a spreadsheet with situational and adjusted stats for NHL players. The stats begin in 1967-68 when plus-minus data was first officially published by the NHL. Some members of the ATD community have found it useful so I thought I might as well post it for general use and discussion.

Link is here. Stats are included for individual seasons and careers. If you use pivot tables, there's one included so you can easily filter and calculate the stats over a particular time frame.

Full file through 2020

Career numbers only

There's a glossary explaining the different terms included in the download. Here are a few of the commonly used ones.

$ESG Even strength goals, adjusted to a league scoring level of 180 ESG per team.
$ESA Even strength assists, adjusted to a league scoring level of 180 ESG per team.
$ESP Even strength points, adjusted to a league scoring level of 180 ESG per team.
$ESGF Even strength goals for (on-ice), adjusted to a scoring level of 180 ESG per team.
$ESGA Even strength goals against (on-ice), adjusted to a scoring level of 180 ESG per team.
R-ON Even strength GF/GA ratio for the player's team while the player was on the ice.
R-OFF Even strength GF/GA ratio for the player's team while the player was off the ice.
AEV+/- Adjusted even-strength plus-minus - plus minus adjusted for scoring level and strength of team. Calculated with (EV+/-) - (XEV+/-). Does not adjust for strength of linemates or opposition.
EV% Percentage of team's even strength goals, both for and against, for which the player was on the ice (per game). Calculation is (ESGF+ESGA)/(TmESGF+TmESGA)/SFrac. If you don't want it on a per-game basis, simply multiply by SFrac.
TmEV+ Strength of the player's team at even strength. 1 is average, higher is better. Calculated with TmESGF/TmESGA.
$PPG Power play goals, adjusted to a league scoring level of 60 PPG per team and a league-average number of power play opportunities.
$PPA Power play assists, adjusted to a league scoring level of 60 PPG per team and a league-average number of power play opportunities.
$PPP Power play points, adjusted to a league scoring level of 60 PPG per team and a league-average number of power play opportunities.
PP% Percentage of team's power play goals for which the player was on the ice (per game). Calculation is PGF/TmPPGF/SFrac. If you don't want it on a per-game basis, simply multiply by SFrac.
TmPP+ Strength of the player's team's power play. 1 is average, higher is better. Calculated with ((TmPPGF-TmSHGA)/TmPPO)/((LgPPGF-LgSHGA)/LgPPO).
$SHG Shorthanded goals, adjusted to a league scoring level of 6 SHG per team and a league-average number of times shorthanded.
$SHA Shorthanded assists, adjusted to a league scoring level of 6 SHG per team and a league-average number of times shorthanded.
$SHP Shorthanded points, adjusted to a league scoring level of 6 SHG per team and a league-average number of times shorthanded.
SH% Percentage of team's power play goals against which the player was on the ice (per game). Calculation is PGA/TmPPGA/SFrac. If you don't want it on a per-game basis, simply multiply by SFrac.
TmSH+ Strength of the player's team's penalty kill. 1 is average, lower is better. Calculated with ((TmPPGA-TmSHGF)/TmTSH)/((LgPPGF-LgSHGA)/LgPPO)

Click to read more...

 

[jarek]

Is this the one that goes up to 2015?

 

[Iceman]

Is this the one that goes up to 2015?

It goes up to 11.

 

[jarek]

It goes up to 11.

The download says 2015. It's probably the same one I got a while ago.

Anyways, thanks for sharing it overpass!

 

[Rob Scuderi]

Awesome, thanks overpass!

It is in fact including 2015, I've been using 2013 for awhile

If anyone in the drafts see people post % of penalties killed this would be where that data comes from. Not to mention a whole host of other useful information.

 

[TheDevilMadeMe]

Updated through 2015? Nice

 

[BenchBrawl]

Thanks overpass.

 

[Iceman]

The download says 2015. It's probably the same one I got a while ago.

Anyways, thanks for sharing it overpass!

 

[jarek]

How would I use the pivot table to, for example, calculate Suter's PK% while eliminating the first 3 years of his career?

 

[overpass]

How would I use the pivot table to, for example, calculate Suter's PK% while eliminating the first 3 years of his career?

Use the filters at the top of the sheet. Click on Year and deselect all seasons except the ones you want to view. Then find Suter's numbers listed in the table for those seasons.

 

[BraveCanadian]

Great work as always overpass

I know a lot of people rely on this sheet

 

[overpass]

To be honest, I have reason to believe the PK% stat is a little dubious. The formula for it is (player power play goals against / team power play goals against) * sfrac (I assume this is the ratio of player games played to team games played).

Basically this formula rewards people who perform poorly on the PK, as far as I can tell. Take Ed Westfall for example:

He had a 69% career PK usage. That means he was on the ice for 69% of his team's total goals against on the PK. Either he played a lot more minutes than just about every other forward for his teams on the PK (quite possible), or he just wasn't a very good PK'er. A more recent example:

2016: Ryan Kesler (10 PGA) vs Jakob Silfverberg (13 PGA) against the team's 28 PGA. The formula would have Silfverberg at 46% usage to Kesler's 36% usage.. despite the fact that Kesler leads Silfverberg in PK TOI/G by 22 seconds (2:53 to 2:31). Kesler is simply a significantly more effective PK'er than Silfverberg, but PK% wouldn't tell you that.

It is a bit of a flawed stat, I think, but unfortunately, all we have for before TOI was tracked.

I won't claim the PK% stat is 100% accurate. I originally created it just to see what the numbers looked like, and found that the career numbers were a very good match for the best penalty killers by reputation. For example, Ed Westfall had a reputation as an excellent penalty killer, and he also has a very high PK%.

Furthermore, the vast majority of the players who have done the most penalty killing according to this stat also had above-average team results. So it's extremely unlikely that the fact that they were on the ice for a lot of goals against means that they were bad penalty killers.

Looking at career numbers or multi-season numbers means the coaching filter becomes very strong. We wouldn't expect players to play on the penalty kill unless the coach thought they were doing well there. When the same player continues to get used in this role over multiple seasons and through multiple coaches, at some point it becomes extremely unlikely that they were a negative in this role.

I wouldn't put too much stock in the stat for single seasons as there is a fair bit of random fluctuation for goals-based stats in a single season. And it's a short enough time period that a player might perform poorly on the PK and still get used in that role due to a lack of internal options. The stat is best used over a career or over multiple seasons. The Kesler/Silfverberg example is a relevant criticsm of the stat on a single season level. I would add the example of Patrice Bergeron in the shortened 2013 season - only 4 PPGA despite taking a regular PK shift. Those are examples of why I only trust the stat so far for a single season, but they don't really speak to the career numbers. Over a career, top penalty killers will continue to get used on the penalty kill, and if they are really so effective that they significantly decrease goals against, their team will have very good results and this will show up in the TmPK+ numbers.

There is another reason that the PK% stat may be better than you think. Simply put, quality of competition is extremely important on the penalty kill. Shorthanded ice time can be time defending a 5-on-3, time defending a 5-on-4 against the opposing first PP unit, and time defending a 5-on-4 against the opposing second PP unit. The three situations are very different. The coach will probably use the best defensive forward for the 5-on-3 - and he will have a high rate of goals against in that 5-on-3 ice time. Similarly, first PP units in the NHL are significantly more dangerous than second PP units in the NHL. The talent concentration on the first unit is a major factor, and it also helps that they get to start their shift in the offensive zone rather than on the fly. So players on the first PK unit, regardless of skill, will almost always have a higher rate of goals against than the players on the second unit. So a guy like Martin St. Louis looks good if you look at his goals-against/minute on the penalty kill over his career, and also at his shorthanded scoring - but he has basically never been on the ice against the oppositions first unit, so he gets relatively "easy" shorthanded minutes.

 

[jarek]

I won't claim the PK% stat is 100% accurate. I originally created it just to see what the numbers looked like, and found that the career numbers were a very good match for the best penalty killers by reputation. For example, Ed Westfall had a reputation as an excellent penalty killer, and he also has a very high PK%.

Furthermore, the vast majority of the players who have done the most penalty killing according to this stat also had above-average team results. So it's extremely unlikely that the fact that they were on the ice for a lot of goals against means that they were bad penalty killers.

Looking at career numbers or multi-season numbers means the coaching filter becomes very strong. We wouldn't expect players to play on the penalty kill unless the coach thought they were doing well there. When the same player continues to get used in this role over multiple seasons and through multiple coaches, at some point it becomes extremely unlikely that they were a negative in this role.

I wouldn't put too much stock in the stat for single seasons as there is a fair bit of random fluctuation for goals-based stats in a single season. And it's a short enough time period that a player might perform poorly on the PK and still get used in that role due to a lack of internal options. The stat is best used over a career or over multiple seasons. The Kesler/Silfverberg example is a relevant criticsm of the stat on a single season level. I would add the example of Patrice Bergeron in the shortened 2013 season - only 4 PPGA despite taking a regular PK shift. Those are examples of why I only trust the stat so far for a single season, but they don't really speak to the career numbers. Over a career, top penalty killers will continue to get used on the penalty kill, and if they are really so effective that they significantly decrease goals against, their team will have very good results and this will show up in the TmPK+ numbers.

There is another reason that the PK% stat may be better than you think. Simply put, quality of competition is extremely important on the penalty kill. Shorthanded ice time can be time defending a 5-on-3, time defending a 5-on-4 against the opposing first PP unit, and time defending a 5-on-4 against the opposing second PP unit. The three situations are very different. The coach will probably use the best defensive forward for the 5-on-3 - and he will have a high rate of goals against in that 5-on-3 ice time. Similarly, first PP units in the NHL are significantly more dangerous than second PP units in the NHL. The talent concentration on the first unit is a major factor, and it also helps that they get to start their shift in the offensive zone rather than on the fly. So players on the first PK unit, regardless of skill, will almost always have a higher rate of goals against than the players on the second unit. So a guy like Martin St. Louis looks good if you look at his goals-against/minute on the penalty kill over his career, and also at his shorthanded scoring - but he has basically never been on the ice against the oppositions first unit, so he gets relatively "easy" shorthanded minutes.

Really good points here. I mostly came to the same conclusion - that the PK% is useful over a long stretch but not for a single season due to the reasons you mentioned. I agree that a guy getting used very often on the PK generally means he's pretty good at it. I would find it rather unlikely over a stretch of, say, 10 years, that a guy getting as much usage as Westfall got was due to the team having no better options.

I do think, though, that there are probably some players who get unfairly dinged by the PK% stat more often than others. That's just how it is when you do an analysis like this - some guys will look better than they actually are, and some will look worse then they actually are. I think SH TOI/G, when available, should probably supersede PK%. However, for older players, I think PK% alone just isn't quite enough. Having some contemporary evidence about a guy's PK'ing ability to go along with the stat would be very much preferred I think. I don't think this is generally too hard to find for the best guys, but I think there's some room for improvement for the guys who weren't quite as prolific in that role.

 

[seventieslord]

I wouldn't worry about it too much, jarek. I think all your concerns in your last post were already addressed in the post it was quoting.

It's always nice to have quotes about any aspect of a player's game, but if he was consistently used on the penalty kill, that tells us that a player's coaches thought he was useful in that role.

Is there "room for improvement" in how much info we have on pre-TOI players' penalty kill abilities? sure. Is there room for improvement in what overpass did here? no, it is what it is. I don't see how the scant official numbers that exist can be better used to approximate PK usage aside from the TOI numbers and overpass' excellent shorthand.

 

[tinyzombies]

Is there an updated version of this?

 

[overpass]

Updated numbers including the 2019-20 season are available here. The links go to the file on dropbox and you should be able to download them.

Full file through 2020

Just the career sheet (much smaller)

I've added career $ESG/82 and $PPG/82 to the sheet.

I endorse the use of even-strength-VsX over the adjusted even strength scoring numbers in my sheet for first line scorers. Even-strength usage and opportunity to score for first line players has varied over time in a way that isn't captured by a scoring level adjustment. For example, the time period from 1982-1992 was a time when lines were used more evenly than at any other time. Prior to 1982 there were only 16-17 skaters dressed, so a full fourth line was not used. From 1993-on, longer TV timeouts meant that coaches could play their top lines more, usually at the expense of their fourth line, without tiring them out. So my adjusted even strength scoring numbers tend to be lower for players from 1982-1992 compared to players from earlier and later years*. I've chosen not to adjust for this factor because I report the numbers for everyone, including the third and fourth lines, and I'm not sure how to apply an adjustment that could apply to all players. Just something to be aware of if you are using my adjusted even strength scoring numbers.

Also, not sure if anyone has noticed, but nhl.com now has PP% and SH% stats based on ice time since 1997-98 available on their stats site. I expect them to differ a bit from my numbers that are only goal-based, mostly because first unit power plays tend to score more than second unit power plays.

NHL.com Stats

NHL.com Stats

*Wayne Gretzky is a bit of an exception here -- he was able to manage very high levels of ice time in the 80s compared to his contemporary star forwards, even without the benefit of longer TV timeouts, due to his exceptional capacity for endurance.

 

[tinyzombies]

Thanks!

 

[overpass]

Just realized there was an issue with the career-only sheet I uploaded. Positions for some players were incorrect -- for example, Bobby Hull was listed as D. I've fixed it, re-uploaded the file, and updated the link in post #16. Here's the link again.

If you aren't sure if you have the right sheet, check Bobby Hull's position. It should be listed as L.

The position I've listed on the career sheet is usually the position first played by the player, so it's not necessarily the position the player played most over their career.

 

[BM67]

Just looking around the file, I found Ryan McDonagh has a 0% SH% for 2017, despite being on for 24 of 45 PGA.

 

[overpass]

Just looking around the file, I found Ryan McDonagh has a 0% SH% for 2017, despite being on for 24 of 45 PGA.

Thanks for pointing it out. Looks like a there's a problem with some of the 2017 stats for SH%. I'll fix and upload soon.

 

[overpass]

Just looking around the file, I found Ryan McDonagh has a 0% SH% for 2017, despite being on for 24 of 45 PGA.

I've redone the calculations and updated the file. It should all be good now.

Full file

Career only

Ryan McDonagh has a 57% SH% for 2016-17 and a 55% SH% for his career. nhl.com has his SH% based on ice time as 54.7.

There is one data issue I'm aware of. Hockey-reference.com is my data source for the 2017-18 season. They show Jaccob Slavin with 43 power play goals against. Carolina as a team allowed 43 PPGA for the 2017-18 season, and it seems unlikely that Slavin was on the ice for every single goal against. In fact, I checked with @tarheelhockey to see if it was remotely possible, and he confirmed there were at least 3 power play goals allowed by Carolina where Slavin was not on the ice--probably more. But I haven't found another source with correct data, so the sheet shows Slavin with a 100% SH% for 2017-18. He does play a lot of shorthanded minutes (69% for 2017-18 according to nhl.com), but not that much.

 

[Hockey Outsider]

@overpass - thanks again for posting this. By any chance, do you also have playoff numbers?

 

[Hockey Outsider]

One more question/favour - do you also have TOI data in this format ("LAST NAME, FIRST NAME")? I used this spreadsheet for a project I've been meaning to do for a while and now that I'm trying to cross-refer it to the NHL's ice time data, I can't find any site that has the player names in this format. Thanks again.

 

[overpass]

@overpass - thanks again for posting this. By any chance, do you also have playoff numbers?

I have the 1960-67 numbers somewhere, I’ll try to dig them up. Unfortunately I never compiled later years, and now the plus-minus component stats are no longer on nhl.com

 

[overpass]

One more question/favour - do you also have TOI data in this format ("LAST NAME, FIRST NAME")? I used this spreadsheet for a project I've been meaning to do for a while and now that I'm trying to cross-refer it to the NHL's ice time data, I can't find any site that has the player names in this format. Thanks again.

I’ll see if I have this one as well.