Wins/ Points Above Replacement Results

[NHL WAR]

Some of you may remember my past postings about a NHL version of baseball`s WAR. At those times, I had only attempted to replicate it for forwards. I have finally got around to completing it for all positions, dating back to the 2007-08 season (the first year we have shot attempt data).

I really have no interest in advanced stats preaching, so if this type of thing isn`t your cup of tea, no hard feelings. There are some good games on today, countless trade speculation threads and the weather is getting decent.. you don`t have to spend your time here.

For those curious, I made four write ups going over the goals and process of creating these stats, plus a look at some of the most relevant results and a few tests to see how effective the numbers are:

Goals for Wins Above Replacement and Things It Can Assist
Part 2: The Stat
Part 3: Results
Part 4: Effectiveness

For those just wanting a snapshot, here are the top career performers:


Top 9 Forwards

	WAR	PAR	GP
SIDNEY.CROSBY	36.7	82	824
ALEX.OVECHKIN	31.5	70.3	989
EVGENI.MALKIN	30.9	69	829
PATRICK.KANE	30.2	67.4	973
PATRICE.BERGERON	27.5	61.4	860
JONATHAN.TOEWS	27.3	61	943
ANZE.KOPITAR	26.7	59.6	1001
JOE.THORNTON	26.7	59.6	964
PAVEL.DATSYUK	25	55.9	590

[TBODY] [/TBODY]

Top 6 Defenseman

		WAR	PAR	GP
ERIK.KARLSSON		24.2	54.1	736
KRIS.LETANG		21.8	48.6	801
DUNCAN.KEITH		21.6	48.3	975
MARK.GIORDANO		21	47	838
VICTOR.HEDMAN		20.3	45.3	762
RYAN.SUTER		20.1	44.9	989

[TBODY] [/TBODY]

Top 3 Goalies

	WAR	PAR	GP
HENRIK.LUNDQVIST	31.4	70	764
CAREY.PRICE	20.8	46.4	682
ROBERTO.LUONGO	20.2	45.1	627

[TBODY] [/TBODY]

Top performers as of Mar.14th (roughly halfway point) of this season:

	WAR	PAR
CONNOR.MCDAVID (EDM)	2	4.4
ANDREI.VASILEVSKIY (T.B)	1.9	4.3
AUSTON.MATTHEWS (TOR)	1.4	3.2
ALEKSANDER.BARKOV (FLA)	1.4	3.2
MARC-ANDRE.FLEURY (VGK)	1.4	3.2
MARK.STONE (VGK)	1.4	3.1
LEON.DRAISAITL (EDM)	1.4	3.1

[TBODY] [/TBODY]

A couple rough guides for how to interpret this data:

DOs	DONTs
-look at career totals when judging past draft classes, trades, etc. as a one size fits all, objective ranker	-take the results as gospel -piggybacking on this, don`t have too much faith in the decimal places. Player A having 0.2 more WAR than player B is far from conclusive evidence that player A is a more valuable player
-check out recent numbers to determine whose play/ reputation has the numbers to back it up, identify possible upgrades for your team, find the next undervalued asset	-fail to consider sample sizes
-take the current season`s numbers into account come award season	-lose the motivation to ask what other factors could be influencing a particular players numbers (if you have the time.. don`t miss puck drop)

[TBODY] [/TBODY]

WAR (per 82 games)	PAR	ROLE
3.5+	7.8+	Hart Finalist, Norris/ Vezina Winner
1.8+	4+	All-star
0.8+	1.8+	Top 6 forward, top 3 defenseman, 1A goalie
0.2+	0.4+	Contributing depth
0	0	Replacement level/ dressing room DJ/ decent player being asked to do too much

[TBODY] [/TBODY]

The complete results can be retrieved downloaded here: Smart Hockey Simple


I fully acknowledge that this is all amateur hour, but I think it can still be useful when utilized correctly.

Feel free to use any of the data, so long as you source it and send me a link if you do any analysis in the future... always interested in that sort of stuff. I personally plan to update this with posts looking at the relationships between WAR/ PAR and cap hits, draft choices, trades, age curves, etc.

Click to read more...

 

[TaLoN]

Not surprised to see Suter in good company career wise... nice write up!

 

[Sidney the Kidney]

I'll just say I enjoyed this because it shows irrefutably that Crosby was the better player than Ovechkin over the course of their careers.

 

[NHL WAR]

I'll just say I enjoyed this because it shows irrefutably that Crosby was the better player than Ovechkin over the course of their careers.

How it rates your favourite players is a big test for a stat. I was personally disappointed when James Reimer was considered less valuable than Carey Price

 

[silkyjohnson50]

Great work as always.

If we were to look at a per game basis, would we simply divide the total war/games played or is there another calculation? I can’t help but notice Datsyuk sticks out in terms of playing so few games compared to the rest of the list.

 

[Dominance]

Crosby significantly ahead of the field, as one would expect. Thanks for putting all the time into this!

 

[FunkySeeFunkyDo]

A couple surprises:
This stat likes Letang and Toews.(Not that those both aren’t excellent, but just how highly they are ranked.)

The narrative of the post-Datsyuk/pre-McDavid era is that there’s Crosby/Ovy then Malkin/Kane then everyone else. It really should be Crosby then Ovy/Malkin/Kane then everyone else.

Just thought of a third surprise. Didn’t see Jack Johnson on the bottom of the pack list despite the fact he has played a ton of minutes. Gonna guess that he had enough offense as a young player to keep him off.

 

[NHL WAR]

Great work as always.

If we were to look at a per game basis, would we simply divide the total war/games played or is there another calculation? I can’t help but notice Datsyuk sticks out in terms of playing so few games compared to the rest of the list.

Yep, WAR/ games played works. I usually then multiply it by 82 games to get a nicer scale. Crosby still leads the pack with 3.7 WAR per 82. Goalies benefit from the per game analysis, with several nipping at Crosbys heels (Andrew Hammond is runner up to Crosby, which is amusing). Just looking at skaters, the top 10 per game is:

Crosby
McDavid
Datsyuk
Panarin
Malkin
Makar
Kucherov
Matthews
Pastrnak
Fox

Little biased towards guys that haven't been in the league long enough to have down years, but quality players nonetheless

 

[Hockey Outsider]

Thanks for posting - very interesting set of articles. I also like that you included an interpretation guide.

A few questions (all intended to be constructive):

1. There's a perception that some players consistently out-perform their expected goals (ie Ovechkin). Would this formula therefore underrate elite goal-scorers? (Ovechkin is still way ahead of everyone in WAR from shooting goals).

2. Is defensive play on the penalty kill included here? Looking at the defense components, the Corsi against and expected goals against data appears to be based on 5-on-5 only.

3. The correlations between PAR and actual results (at the team level) are quite strong. But actual results, of course, include the results of shootouts. If you disregard shootouts (which aren't captured in your formula, as far as I can tell) - maybe by treating all shootout wins and losses as ties - I wonder if you'd get an even higher correlation.

4. The correlation between players' 1st and 2nd half of the year is quite a bit weaker. I wonder if that's simply due to sample sizes. Did you look at players' results from one season to the next? (EDIT - actually I see that you have this. A decent correlation, but not overly strong. But I wasn't sure if it's a straightforward calculation, or if you're doing adjustments for games played - since it's a GAR as opposed to a GAA type calculation, that could have a significant impact).

 

[NHL WAR]

A couple surprises:
This stat likes Letang and Toews.(Not that those both aren’t excellent, but just how highly they are ranked.)

The narrative of the post-Datsyuk/pre-McDavid era is that there’s Crosby/Ovy then Malkin/Kane then everyone else. It really should be Crosby then Ovy/Malkin/Kane then everyone else.

Just thought of a third surprise. Didn’t see Jack Johnson on the bottom of the pack list despite the fact he has played a ton of minutes. Gonna guess that he had enough offense as a young player to keep him off.

Letang definitely ranked higher than I expected.

Ovi has the two highest WAR seasons of any forward, but some one-dimensional years drag his career totals closer to Malkin and Kane.. still very good company.

The amount of minutes Johnson logged also benefits him when the time comes in the calculations to convert from above average to above replacement level (there is a set number of rep. goals each season and they are divvied up based on ice time). Still, his results weren't pretty. In the 07-08 season, he had -14 Defensive Goals Saved (next worst was -8) and was valued at -1.5 WAR/ -3.3 PAR. For his career, he again ranks last with -58 D.G.S and is at -0.7 WAR.

 

[gtrower]

I'll just say I enjoyed this because it shows irrefutably that Crosby was the better player than Ovechkin over the course of their careers.

Well when you normalize over games played this happens. The argument has always been (and likely will always be) Ovi was able to stay on the ice through injuries. Sid gets the benefit of extrapolating his healthiest games. Ovi gets his unhealthy games (when he still played) interpolated.

 

[Hockey4Lyfe]

Well when you normalize over games played this happens. The argument has always been (and likely will always be) Ovi was able to stay on the ice through injuries. Sid gets the benefit of extrapolating his healthiest games. Ovi gets his unhealthy games (when he still played) interpolated.

Damn. Talk about going to great lengths to try and downplay a stat.

Assuming that Crosby never played through any type of injury his entire career is not just one thing but then trying to discredit Ovechkins years where he was seemingly an average player, is entirely different.

Are you trying to say all those years he was just injured and playing through those so they should be thrown out when putting these stats together?

 

[Butch 19]

Some of you may remember my past postings about a NHL version of baseball`s WAR. At those times,

.

srsly?

 

[NHL WAR]

Thanks for posting - very interesting set of articles. I also like that you included an interpretation guide.

A few questions (all intended to be constructive):

1. There's a perception that some players consistently out-perform their expected goals (ie Ovechkin). Would this formula therefore underrate elite goal-scorers? (Ovechkin is still way ahead of everyone in WAR from shooting goals).

2. Is defensive play on the penalty kill included here? Looking at the defense components, the Corsi against and expected goals against data appears to be based on 5-on-5 only.

3. The correlations between PAR and actual results (at the team level) are quite strong. But actual results, of course, include the results of shootouts. If you disregard shootouts (which aren't captured in your formula, as far as I can tell) - maybe by treating all shootout wins and losses as ties - I wonder if you'd get an even higher correlation.

4. The correlation between players' 1st and 2nd half of the year is quite a bit weaker. I wonder if that's simply due to sample sizes. Did you look at players' results from one season to the next? (EDIT - actually I see that you have this. A decent correlation, but not overly strong. But I wasn't sure if it's a straightforward calculation, or if you're doing adjustments for games played - since it's a GAR as opposed to a GAA type calculation, that could have a significant impact).

1) Yeah I definitely agree that some guys will experience significant variance from their expected goals and would like to do a study on that one day. In this model, though, the expected goals for are more of a tiebreaker than the straw that stirs the drink. The iXG (and iCF) above average component is weighted in a way that it only ends up with about 1/20th the influence of actual goals. As a result, it doesnt have much of an impact on Ovi (who still rates above average in individual expected goals anyhow)

2) Defensive play on the PK is not included. It was incorporated into a very early draft and I'll probably look into it again. I think the ideal model would include it, but I was comfortable excluding it for now as I think it would be an area that would take a few seasons worth of data for a player's true talent level to differentiate from their environment.

3) Treating all shootouts as ties bumps it from 0.881 to 0.885. Biggest jump comes from simply removing Anaheim, that raises it to 0.921. It is interesting that they stray so far from the rest of the pack.

4) Yeah, I kept it pretty simple with the year to year correlation, I think only requirement was 40 games played each season. Will likely try again with some combination of:
-incorporating two or three past years to predict future
-regression to the mean
-age curve
-using a per 82 game rate
to try and build a more accurate forecast

 

[NHL WAR]

srsly?

Dfntly

 

[DearDiary]

Well when you normalize over games played this happens. The argument has always been (and likely will always be) Ovi was able to stay on the ice through injuries. Sid gets the benefit of extrapolating his healthiest games. Ovi gets his unhealthy games (when he still played) interpolated.

Your argument falls apart as Crosby's statistics drop with each game he plays. Because obviously it hurts your career numbers when your average includes almost all your games on the decline but so few during your peak.

e.g.
650 games played
300 before peak
50 peak
300 after peak

vs

800 games played
300 before peak
200 peak
300 after peak

Crosby's numbers are worse than they'd be if he got to play his peak games like Ovechkin.

Your healthy games argument makes no sense whatsoever. Are you implying Crosby never played injured? Last season Crosby played every single game with an injury or on rehab. He still had a much higher ppg than Ovechkin. This year he had wrist surgery and is still outplaying Ovechkin.

Let me know how many games Ovechkin has had to play through recovering from surgeries.

 

[Svencouver]

This lines up pretty well with my intuition, so as someone who likes having their beliefs confirmed I think this is great. Nice work

 

[gtrower]

Your argument falls apart as Crosby's statistics drop with each game he plays. Because obviously it hurts your career numbers when your average includes almost all your games on the decline but so few during your peak.

e.g.
650 games played
300 before peak
50 peak
300 after peak

vs

800 games played
300 before peak
200 peak
300 after peak

Crosby's numbers are worse than they'd be if he got to play his peak games like Ovechkin.

Your healthy games argument makes no sense whatsoever. Are you implying Crosby never played injured? Last season Crosby played every single game with an injury or on rehab. He still had a much higher ppg than Ovechkin. This year he had wrist surgery and is still outplaying Ovechkin.

Let me know how many games Ovechkin has had to play through recovering from surgeries.

Well it’s finally happened. Pens fans are know claiming Crosby as the guy who plays through more.

 

[DearDiary]

Well it’s finally happened. Pens fans are know claiming Crosby as the guy who plays through more.

Are you going to back up your argument or not? Explain how Crosby benefited missing so many games during his peak, lowering his average career stats and playing after so many injuries and surgeries. The worst being the high ankle sprain (2007?) that permanently affected his stride.

 

[Brown Cat]

Matthews and Barkov being so close this season is interesting. Can't really go wrong with either one then so far.

 

[nturn06]

If I understand this right, this suggest that the Oilers would still be 3rd in North without McD and without Drai.... Hmmm...

 

[Herregud]

Ovechkin and Crosby on the same list, so now of course it is an Ovechkin v Crosby thread.

I thought we'd moved on to Crosby v McDavid. Or McDavid v Draisaitl.

 

[NHL WAR]

If I understand this right, this suggest that the Oilers would still be 3rd in North without McD and without Drai.... Hmmm...

As of Mar.14th:

Leafs: GP: 30 P:40
Oilers: 30, 36
Jets: 27, 36
Habs: 27, 31
Flames: 28, 29
Canucks: 31, 28
Sens: 31, 21

So since McDavid and Drai combined for 7.5 PAR at that time, Oilers would drop down to Canucks level. I get that it seems weird that they wouldn't completely crater without McDrai, though.

 

[nturn06]

As of Mar.14th:

Leafs: GP: 30 P:40
Oilers: 30, 36
Jets: 27, 36
Habs: 27, 31
Flames: 28, 29
Canucks: 31, 28
Sens: 31, 21

So since McDavid and Drai combined for 7.5 PAR at that time, Oilers would drop down to Canucks level. I get that it seems weird that they wouldn't completely crater without McDrai, though.

OK, missread the table... But still, the table suggests that without McD and Drai they would right now be in a battle for the playoffs, that is more than odd...

And without just McD they would still be third in the North, hmmm....

 

[gtrower]

Are you going to back up your argument or not? Explain how Crosby benefited missing so many games during his peak, lowering his average career stats and playing after so many injuries and surgeries. The worst being the high ankle sprain (2007?) that permanently affected his stride.

Yes. After injuries. Not through injuries. Pretending Crosby is the tougher player here is just laughable. So many other legitimate arguments to be made on your end.