What advanced stats should i use to best portray a player?

[Pominville Knows]

And where do i find the formula or maybe even allready done players or calculators?

Click to read more...

 

[soireeculturelle]

And where do i find the formula or maybe even allready done players or calculators?

for guys who score goals for a living, I like shots per game over the years. its just a small part of quantifying a player, but it does a better job of predicting performance than goals totals and shot %

 

[hatterson]

That's a very open ended question.

What are you looking to portray? Offensive impact? Defensive impact? How he affected the powerplay? Just an overall player rating?

There's been some attempts at overall ratings (somewhat equivalent of win shares or wins above replacement in other spots) but all have noted holes. If you want to look at them, HARO, HARD, and HART on stats.hockey-analysis.com are one attempt. hockey-reference calculates offensive and defensive pointshares and there's several others.

I'd advise looking at those and refining exactly what you want to say about a player. That'll give you a good idea of how to say it.

 

[tarheelhockey]

There's been some attempts at overall ratings (somewhat equivalent of win shares or wins above replacement in other spots) but all have noted holes.

Yeah, I'd be really wary of using the overall ratings that have bubbled up so far. Given that they all have significant flaws, you're probably better off narrowing down to a specific aspect of the game that you want to evaluate.

The exception might be goalies, who seem to be somewhat easier to quantify than skaters.

 

[hatterson]

Yeah, I'd be really wary of using the overall ratings that have bubbled up so far. Given that they all have significant flaws, you're probably better off narrowing down to a specific aspect of the game that you want to evaluate.

The exception might be goalies, who seem to be somewhat easier to quantify than skaters.

And Tarheel was never heard from again after Doc No finished with him

You're absolutely right on the holes though.

Take HART for example. This season, of those with >500 minutes played, Tyler Toffoli is #2, Vladimir Tarasenko is #4, Dustin Penner is #7. I'm not saying these are bad players, but they're certainly not in the top 10. Even looking just at the offensive side, based on HARO Crosby was 25th this year, which seems odd given a 17 point Art Ross win.

Looking at pointshares on hockey-reference has similar issues. Crosby and Jagr had the 3rd and 4th most *defensive* point shares among forwards, Kunitz is at 5th. All ahead of Bergeron which seems dubious at best

 

[TheDevilMadeMe]

Looking at pointshares on hockey-reference has similar issues. Crosby and Jagr had the 3rd and 4th most *defensive* point shares among forwards, Kunitz is at 5th. All ahead of Bergeron which seems dubious at best

Hockey reference's point shares are just a disaster. Jagr's high "defensive" rating is largely a result of his plus/minus, and the fact that he never kills penalties. Yes, H-R counts goals against per minute of ice time without distinguishing between PP, PK, and ES time, so players can only hurt their "defensive point shares" by killing penalties. Basically, you're best just ignoring hockey reference's point-shares. Even compared to other all-in-one stats, they are terrible.

 

[Bear of Bad News]

And Tarheel was never heard from again after Doc No finished with him

On some very important level, goaltenders are responsible for less than skaters, so we should be easier to quantify.

One problem, then, is similar to comparing NFL quarterbacks by their "passer rating" (beyond flaws in the passer rating calculation itself). Quarterbacks do so many things beyond passing.

NHL goaltenders can be rated by how well they stop the puck (and of course there are flaws with doing that). Beyond that, they stickhandle, they prevent shots by positioning (that don't even up in the denominator for save percentage), they are largely responsible for communication and positioning in the defensive zone, they direct traffic, and they are de facto team leaders (letter or no letter).

 

[GuineaPig]

I find that GVT usually produces a reasonable-seeming result.

 

[hatterson]

I find that GVT usually produces a reasonable-seeming result.

Yea it's one of the better ones, but still has issues.

Ovechkin may not have had an elite season, but it was certainly better than 95th

 

[TheDevilMadeMe]

Yea it's one of the better ones, but still has issues.

Ovechkin may not have had an elite season, but it was certainly better than 95th

The top 4 players all goalies.... I see he hasn't fixed the overrating of the effect of goaltenders.

 

[mkwong268]

The top 4 players all goalies.... I see he hasn't fixed the overrating of the effect of goaltenders.

I always felt that GVT underrates defensemen, but I'm probably biased about the importance of the position I play.

 

[TheDevilMadeMe]

I always felt that GVT underrates defensemen, but I'm probably biased about the importance of the position I play.

When the top ranked defenseman is 18th.... yes, I agree with you.

It goes to the weakness of hockey stats. We have stats to measure offensive production, and goaltender stats are pretty good (save percentage with a few adjustments). But there really aren't any good stats to measure defensive performance (though there have been improvements in that area).

 

[mkwong268]

When the top ranked defenseman is 18th.... yes, I agree with you.

It goes to the weakness of hockey stats. We have stats to measure offensive production, and goaltender stats are pretty good (save percentage with a few adjustments). But there really aren't any good stats to measure defensive performance (though there have been improvements in that area).

What are some of the current defensive performance metrics that you think are decent? I'd like to take a look at some.

 

[GuineaPig]

The top 4 players all goalies.... I see he hasn't fixed the overrating of the effect of goaltenders.

I think good goalies are inherently more valuable than good skaters. But GVT certainly underrates defencemen.

 

[mkwong268]

I think good goalies are inherently more valuable than good skaters. But GVT certainly underrates defencemen.

Especially if you compare threshold goalies to good goalies. Average shots this regular season is 2463 per team. All other things being equal a goalie with 1% better save percentage should improve goal differential by 12.3 if he plays in just half the games.

 

[TheDevilMadeMe]

I think good goalies are inherently more valuable than good skaters. But GVT certainly underrates defencemen.

I just find it hard to believe that 4 goalies provide more value than Sidney Crosby is a season when Crosby won the Art Ross by a large margin

 

[TheDevilMadeMe]

What are some of the current defensive performance metrics that you think are decent? I'd like to take a look at some.

There really aren't any.

Corsi combined with zone starts is useful, though I think it overrates puck moving defensemen compared to defensive minded guys who aren't so good at clearing the zone, but are better at keeping scorers to the outside.

Likewise, some teams track zone exits for defensemen, which is good for (get this) figuring out who is best at clearing the zone.

Over a large sample size (minimum several seasons), I think tracking goals against per 60 minutes is very good for top defensemen, though it should be used in combination with the goals against when that player is off the ice (to adjust for team effects).

 

[wgknestrick]

I just find it hard to believe that 4 goalies provide more value than Sidney Crosby is a season when Crosby won the Art Ross by a large margin

GVT is great for forwards, and I don't think it needs much improvement there.

Goalies get about 20% more value than I think they should and that value comes at the expense of the Defensemen GVT.

In an ideal world, the GVT distribution/bell curve of every player position (F,D,G) should be identical. They are not.

Normal distribution for forwards, Goaltenders have HEAVY tails (either supermen, or "replacement" level fodder), and defensemen just don't have much that separates them.

Best D-man GVT is about 16 GVT every year, but are the elite D men only worth 6 goals more than a decent defender? Elite goal tenders are magically worth about 2 elite Dmen. We don't see salaries follow this ranking (not that salaries are perfect).

 

[Bear of Bad News]

In an ideal world, the GVT distribution/bell curve of every player position (F,D,G) should be identical. They are not.

Normal distribution for forwards, Goaltenders have HEAVY tails (either supermen, or "replacement" level fodder), and defensemen just don't have much that separates them.

I wouldn't expect a bell curve to represent *any* varity of professional sports at the highest level.

Define an "average" level of major league performance (however you want to define it). Now, how many people can perform three standard deviations better than that average? Very few.

On the other hand, lots of people can perform three standard deviations below that average. (On the other hand - and that's three hands if you're counting - very few get the opportunity to do so, which may be why you're seeing what you're seeing in the data).