Most useless or inaccurate stat in hockey?

[Jumptheshark]

Is cups won before the year you were born a stat?

Flames fans tend to disagree

a few fan bases put the same emphasis on cups won when the Pony express was still running to cup won in the last 20 years

 

[GuineaPig]

See these posts from this thread

The article that calls shootouts a crapshoot is from 2010. Since then we've had 3 more seasons of sample size.

I think luck is a big component of the shootout, but I just find it incredibly hard to believe that it's the only component as is claimed by a "crapshoot." In general, I'm highly skeptical when someone makes the claim that a result is based off nothing but "luck."

One of the reason the "shootout is a crapshoot" argument is so predominant is that even if you take all the shootouts for a team or player over a season, you're only coming up with 20 or so max attempts for a shooter, and triple that for a goalie. That's a tiny sample, and you're obviously going to see wild variation within it. And given that TV broadcasts typically present the stats only for that season, that's the impression that's going to be made on a fan.

Hell, even if you take all the shootout attempts Lundqvist has faced in 8 seasons, you've only got a sample of 287 shots. Shootouts might not be a complete crapshoot, but the sample sizes they operate on make it very difficult to separate luck and skill, to the extent that one is almost justified in disregarding them.

 

[Cunneen]

One of the reason the "shootout is a crapshoot" argument is so predominant is that even if you take all the shootouts for a team or player over a season, you're only coming up with 20 or so max attempts for a shooter, and triple that for a goalie. That's a tiny sample, and you're obviously going to see wild variation within it. And given that TV broadcasts typically present the stats only for that season, that's the impression that's going to be made on a fan.

Hell, even if you take all the shootout attempts Lundqvist has faced in 8 seasons, you've only got a sample of 287 shots. Shootouts might not be a complete crapshoot, but the sample sizes they operate on make it very difficult to separate luck and skill, to the extent that one is almost justified in disregarding them.

Yes, perhaps I should have worded it differently when I introduced those two articles. There is not a large enough sample size to differentiate luck vs skill.

 

[GuineaPig]

Yes, perhaps I should have worded it differently when I introduced those two articles. There is not a large enough sample size to differentiate luck vs skill.

I'd say that there isn't for a lot of players. But I think it's fairly well-proven that a guy like Lundqvist or Nielsen are better than their peers.

It is important to note that sample size issues are ameliorated a fair amount because of how specific a situation the shootout is. 3000 shots is usually the minimum sample stats guys like to use for evaluating a goalie; but when all the shots are (if not the same type) the same situation, the requirements loosen considerably.

 

[TheDevilMadeMe]

Yes, perhaps I should have worded it differently when I introduced those two articles. There is not a large enough sample size to differentiate luck vs skill.

Oh yeah, definitely true on a single season basis, especially for skaters. I do think that after 8 years, we've managed to get a pretty decent sample though, at least for goalies.

I'd say that there isn't for a lot of players. But I think it's fairly well-proven that a guy like Lundqvist or Nielsen are better than their peers.

It is important to note that sample size issues are ameliorated a fair amount because of how specific a situation the shootout is. 3000 shots is usually the minimum sample stats guys like to use for evaluating a goalie; but when all the shots are (if not the same type) the same situation, the requirements loosen considerably.

Always nice to have the guy I quoted from another thread come in so fast.

To add to your point that the "sample size doesn't need to be so high," you are also removing the team-based variables in the shootout.

Speaking of which, do you know how shootout save percentage is calculated? Does it have to be a shot on goal? Because that still would hurt goalies who like to poke check in the shootout or intimidate shooters into missing the net. If it's calculated simply as % of opponent's attempts on which they fail to score, it (unlike in-game save %) would be a perfect measure of a goalie's effectiveness... over a large enough sample size, of course.

 

[Bear of Bad News]

Speaking of which, do you know how shootout save percentage is calculated? Does it have to be a shot on goal? Because that still would hurt goalies who like to poke check in the shootout or intimidate shooters into missing the net. If it's calculated simply as % of opponent's attempts on which they fail to score, it (unlike in-game save %) would be a perfect measure of a goalie's effectiveness... over a large enough sample size, of course.

The NHL (correctly in my opinion) tracks this as the percentage of shooter attempts where they do not convert. It's inconsistent with "true" save percentage in this regard, of course.

 

[matnor]

Oh yeah, definitely true on a single season basis, especially for skaters. I do think that after 8 years, we've managed to get a pretty decent sample though, at least for goalies.

This is what confuses me. The second article uses 7 years worth of data and is still not able to detect that skill differences among goalies. So I can't really conclude anything other than either i) there is no skill difference among goalies in the shootout. Or ii), we do not have enough data to detect the skill difference. Personally, I think the second explanation is more likely but for that it definitely, based on the data in the articles, seems that the sample size is not large enough yet.

 

[skillhockey]

GAA for goalies totally useless, but it could be used to evaluate team defense.

 

[Cunneen]

This is what confuses me. The second article uses 7 years worth of data and is still not able to detect that skill differences among goalies. So I can't really conclude anything other than either i) there is no skill difference among goalies in the shootout. Or ii), we do not have enough data to detect the skill difference. Personally, I think the second explanation is more likely but for that it definitely, based on the data in the articles, seems that the sample size is not large enough yet.

Think of it this way. For shooters at least, they are only getting at max like 10 shootout attempts per year. Multiply by 8, and you have 80 attempts. Could you determine a player's true SH% after 80 shots. Hell no. Same thing with shootout attempts.

Now for goalies, they are facing a few more attempts. But remember that even a season's worth of goalie data (SV%) is not very accurate in predicating future outcomes. And a goalie will face way more shots in one season then they will in shootouts over their entire career. There just is not a large enough sample size to differentiate skill from luck in the shootout

 

[TheDevilMadeMe]

Now for goalies, they are facing a few more attempts. But remember that even a season's worth of goalie data (SV%) is not very accurate in predicating future outcomes. And a goalie will face way more shots in one season then they will in shootouts over their entire career. There just is not a large enough sample size to differentiate skill from luck in the shootout

I hate this interpretation. Being unable to predict future results does not mean that past results are caused by "luck."

 

[matnor]

Think of it this way. For shooters at least, they are only getting at max like 10 shootout attempts per year. Multiply by 8, and you have 80 attempts. Could you determine a player's true SH% after 80 shots. Hell no. Same thing with shootout attempts.

Now for goalies, they are facing a few more attempts. But remember that even a season's worth of goalie data (SV%) is not very accurate in predicating future outcomes. And a goalie will face way more shots in one season then they will in shootouts over their entire career. There just is not a large enough sample size to differentiate skill from luck in the shootout

No, I get that. What confused me was the statement that there is enough data to detect skill differences.

 

[GuineaPig]

The NHL (correctly in my opinion) tracks this as the percentage of shooter attempts where they do not convert. It's inconsistent with "true" save percentage in this regard, of course.

Yeah, once again it comes down to sample size. Over the course of a single season, or two or three hypothetically, one could imagine a goalie getting his fair share of rolling pucks and whiffed shots. But once you've faced over 150+ shootout attempts it'll even out and better reflect true skill.

In a weird sort of way, one could argue it's a greater measure of goalie skill than save percentage, just because it ends up giving the goalie credit for good positioning/stick work/whatever.

 

[mrhockey193195]

My vote goes for Attendance stats.

It gets ranked on how high your yearly attendance average is, when it should be a percentage based stat. A team like Winnipeg who has a smaller arena will always be in the bottom half even if they sell out. Which makes no sense to me.

Or the Canucks who have been selling out their arena since 2003 is 10th in the league. Makes zero sense.

Haven't read the whole thread, but to respond to the original post...your point makes sense, when dealing with sellouts (if the Jets sell out and have 15000 fans in the arena, there's no telling if there were 15000 people who wanted to go to the game, or 50000).

But if a team like the Jets don't sell out, it's hypothetically indicative of how many fans wanted to go to the game (hypothetically, because obviously there are a tons of factors involves in attendance that aren't related to level of fan support)...in that case, the raw numbers should be used.

EDIT: A team filling 4000 seats in a 5000 seat arena is not as impressive as a team filling 16000 seats in a 20000 seat arena, even though the percentages are the same.

 

[Cunneen]

I hate this interpretation. Being unable to predict future results does not mean that past results are caused by "luck."

As you can see if you read that post, I said the sample size is not large enough to differentiate between skill and luck. I didn't say in that post you quoted that it was luck.

 

[Cunneen]

No, I get that. What confused me was the statement that there is enough data to detect skill differences.

I don't think there is. At least no one has ever showed conclusive evidence towards this.

 

[Gregsky99]

Can I just say bravo to everyone posting in this thread. This is the first thread (and I've read a lot of them) where the topic has had meaningful discussion with sources and reasoning behind the posts. Bravo HF

 

[tucker3434]

GAA for goalies totally useless, but it could be used to evaluate team defense.

Yep, completely agree. If anything, it seems to me like goalies should be rewarded, not punished, for maintaining the same save percentage over a larger sample size.

 

[Bear of Bad News]

If anything, it seems to me like goalies should be rewarded, not punished, for maintaining the same save percentage over a larger sample size.

Although I agree with your thesis, how are goaltenders currently punished for maintaining the same save percentage over a larger sample size?

 

[Blue Blooded]

Yep, completely agree. If anything, it seems to me like goalies should be rewarded, not punished, for maintaining the same save percentage over a larger sample size.

Although I agree with your thesis, how are goaltenders currently punished for maintaining the same save percentage over a larger sample size?

My interpretation of quote 1:

Two goalies playing the same amount of games both posting 92.0% SV%, one with a GAA of 2.00, the other with 2.50.

Goalie A would seem superior due to his lower GAA, but goalie B actually faced more shots and therefore held up his 92.0 over a larger sample.

And they would be "punished" by old school thinkers who still consider GAA a relevant stat, if only the slightest.

 

[Bear of Bad News]

Ah yes - I agree with that completely (and agree that that's probably the interpretation offered). Thanks!

Klein and Reif offered up the "Lucy Ricardo in the chocolate factory" example (in their 1986 Hockey Compendium), where a goaltender's performance suffers when they get overwhelmed. Later analysis seems to show that it doesn't happen at NHL-level workloads (for instance, the difference between 30 shots/game and 40 shots/game), although it sure seems to at lower levels (where a high shot total is probably more reflective of the defense involved). I've been collecting data for a larger style analysis of this, but so far, I'm enjoying the collection aspects more than the analysis aspects.

Regardless, I agree that a goalie shouldn't be penalized for having a higher workload (as GAA does).

 

[FlyingKostitsyn]

Wins/Loss for a goalie.

 

[Hammer Time]

+/-, goalie wins, and (increasingly) Corsi are commonly misunderstood stats. Understood properly and with proper context, they can be useful.

Giveaways and takeaways, are so subjective that they might as well be useless.

 

[JoelWarlord]

I'd say the most useless stats are the ones that are so wildly inconsistent in the way they're tracked. Hits and give/takeaways come to mind for me, they're just so insanely different from building to building. It's not that those stats are entirely useless on their own, but it's pretty tough to use them effectively with how inconsistently they're tracked.

Find it a bit rich to see people saying Corsi is a useless stat. Even more absurd to say it's the most useless. No, it's not the be all and end all of hockey analysis, but to call it the most useless stat is patently false.

 

[GuineaPig]

I'd say the most useless stats are the ones that are so wildly inconsistent in the way they're tracked. Hits and give/takeaways come to mind for me, they're just so insanely different from building to building. It's not that those stats are entirely useless on their own, but it's pretty tough to use them effectively with how inconsistently they're tracked.

Just using Away totals helps a lot with this, though.

 

[pigpen65]

Hit stats, while i think would be very useful if accurate, have to be the most inaccurate stat in hockey. You don't have to look past the fact that almost every single home arena heavily skews numbers for their team to see something's not right.