Arguments Against Advanced Stats?

[ruxpin]

Here you go!

 

I see dead people.

[doom88]

I see dead people.

What did you do now?

[TropicalFruitGirl26]

I have no problem with regular stats, advanced stats, fruit flavored stats, ANY kind of stats that help anyone examine any given game/player/etc as closely as they like.

 

Some I care for, some I think are irrelevant, but ALL, like any set of numbers, whether in sports or business, can be manipulated in various ways to fit a given situation.

 

Are stats of any kind useful? Sure.

Are they be all, end all? Absolutely not.

 

For all the stats in the world to cover hockey, I have yet to see a stat that measures 'heart n soul'.....I think most understand, especially those who have played, that sometimes a player or team can simply WILL their way to a victory through sheer determination...a victory they otherwise probably don't get simply based on talent alone.

The Flyers vs Pens sets of games recently is a good example of this...stats say the Pens SHOULD beat the Flyers..and probably pretty easily too.

Stats also say that Pittsburgh has the better overall group of players....players that have been proven time and again in past games and seasons.

So why is it that the Flyers deal with the Pens so well then go and struggle against a "lesser" team?  Sheer will, determination, and an unquantifiable attitude to "beat those Penguins".

 

Irrational? Probably. But it happens. And that is just one example.

 

I've seen this whole 'advanced stats' vs 'regular stats' or 'eyeball test' argument for years in baseball, as I am avid follower of that as well.

Baseball is more of a numbers driven game than hockey is, though obviously, numbers play a big part in both.

Yet something like heart and determination play a bigger part in hockey than they do in baseball....much harder to 'will' your way to that game winning hit than it is to will yourself to driving that net, create space and jam a puck home for a goal, IMO.

 

Still though, just like baseball stats, hockey stats can be VERY useful....but don't, and never will, tell the whole story about a team, a player, or the direction of a franchise.

As someone already mentioned in this thread, they are tools....and as such, they will always have positives and negatives to them....and like I mentioned, can be manipulated in various fashions to suit what the stat looker is looking for.

This isn't necessarily a 'bad thing'...because like I said, businesses do this all the time as well....and most stats are most useful to those who actually work within the industry that the numbers are being used in.....not the casual fan or follower, though the casuals can certainly delve in as far as they  like in order to try to think along like the 'experts' do.

 

Which leads me to my last point:

Where I have a problem with advanced stats.

And it isn't with their usage, manipulation, or place in supporting arguments.

 

It's when you get someone who delves so far into them, that they deem THEMSELVES experts on the subject matter....then go on to criticize those who may not have delved as deep as they have into them.

Again, have seen this in baseball over and over.

A sect of readers/posters/fans will 'educate' themselves on stat usage to the point where they develop this elitist attitude and at some point, stop becoming actual fans, and more of a regurgitator of numbers, formulas, and probabilities.....then make matters worse by going on to put down others because others do not agree with their viewpoints, didn't put in the time to study numbers as much as they do, and because those others simply prefer a different method of evaluation...some of which may very well include 'going with their gut' or the ol fashioned eyeball test.....both of which can be equally as correct (more fun even), if not scientific.

 

And those casuals using those archaic methods would have their flaws in evaluations many times over.....oh wait, just like advanced stats users would...hmmm...

 

Bottom line:

Stats, advanced or otherwise, are a wonderful thing.

Great tools for aiding evaluations and they are constantly evolving.

 

However, for those that use them as be all end alls.....well, those aren't necessarily the smartest guys in the room they may fancy themselves to be.

Complicate that by name calling, looking down on others, or insulting the intelligence of those who don't care to examine stats as closely as the 'sabernerds', well, IMO, those lose MUCH credibility with me as far as listening to them and acknowledging how knowledgeable they really are on a given subject, no matter how many stats they quote.

[idahophilly]

Yes, games are won by goals, not shots. However, wouldn't it be agreed that usually more shots equals a better chance of winning? Usually, obviously not all the time.

 

How did the higher shot count work out for LA last night? One game but never the less it shows how numbers can be skewed to make any argument.

[idahophilly]

but there are also nearly just as many games where they are outshot and win, yesterday for example. However, it usually all balances out.

 

Your really into this idea it all balances out. how do you know that? Is there proof that it does. Is there a corsi like stat for how often corsi in wrong? And if it has to "even" out then is tells me the numbers are sqewed to begin with, probably because a calculator doesn't compute all the intangibles in a game.

[ruxpin]

Your really into this idea it all balances out. how do you know that? Is there proof that it does. Is there a corsi like stat for how often corsi in wrong? And if it has to "even" out then is tells me the numbers are sqewed to begin with, probably because a calculator doesn't compute all the intangibles in a game.

 

Just to play devils advocate (having already made it clear that my knowledge on this is extremely limited).  I'm guessing it's a little like batting average or ERA.  Overall, a batter has a .340 batting average but it doesn't guarantee he's going to go 1 for 3 today.  Over the course of an extended period, though, he's going to do better at getting on base than the guy batting .250    Similar with the pitcher and his ERA.

 

I think with those stats (particularly the pitcher since it's one every 5 days), I think I'd be hesitant to view them as tried and true for a 10 game stint, but for over the course of a season one can look back and see the stats measure up against the play.   On the otherhand, the .340 BA doesn't tell when he is getting those hits (RSP, RBI, etc) so other stats are needed to create a more full picture.

 

I'm thinking with corsi, that using it for specific games might create such a statistical anomaly and probably still need to be measured against other more traditional stats and the gut/eyeball.   It's worthwhile, I think, to be included in any discussion but it seems like it's not a be-all end all (necessarily) as to whether player X did well last night.  

 

At the same time, I think really good stats for last night should probably reflect a good game whereas I'm not so sure bad corsi stats necessarily reflect a bad game.   I don't know, but I'm willing to guess it's a mixed bag.

[idahophilly]

Mixed bag is all I have been preaching. Use it as a tool but the tool box better have more than one tool in it. A guy can have great corsi numbers but if he is a cancer in the locker room it doesn't show up in the stats at the end of the year but likely had a profound impact on the ice. Same as a 4th liner can be victimized and have a bad corsi number but is the quintessential mentor to the young players (work ethic and such), think Lappy... But still, it won't show up in the stats. The same can be said for the intangibles that left with Hartnell. It's showing up but not in the stats.

[RiskyBryzness]

Corsi correlates well to scoring chances, as well as goals.

 

This following chart looks at a teams' puck possession over time and how it impacts their playoff performance. As you can see, the cup winners typically have higher puck possession.

A better look into advanced stats, if you dare.

http://www.broadstreethockey.com/2013/3/13/4099158/kangzzz

http://www.broadstreethockey.com/2013/3/14/4104952/hockey-advanced-stats-corsi-intro-primer

[RiskyBryzness]

The impacts of different situations have very little impact over the course of a season and it has been researched and somewhat proven.

[RiskyBryzness]

Your really into this idea it all balances out. how do you know that? Is there proof that it does. Is there a corsi like stat for how often corsi in wrong? And if it has to "even" out then is tells me the numbers are sqewed to begin with, probably because a calculator doesn't compute all the intangibles in a game.

http://en.wikipedia.org/wiki/Law_of_averages

 

Just to play devils advocate (having already made it clear that my knowledge on this is extremely limited).  I'm guessing it's a little like batting average or ERA.  Overall, a batter has a .340 batting average but it doesn't guarantee he's going to go 1 for 3 today.  Over the course of an extended period, though, he's going to do better at getting on base than the guy batting .250    Similar with the pitcher and his ERA.

 

I think with those stats (particularly the pitcher since it's one every 5 days), I think I'd be hesitant to view them as tried and true for a 10 game stint, but for over the course of a season one can look back and see the stats measure up against the play.   On the otherhand, the .340 BA doesn't tell when he is getting those hits (RSP, RBI, etc) so other stats are needed to create a more full picture.

 

I'm thinking with corsi, that using it for specific games might create such a statistical anomaly and probably still need to be measured against other more traditional stats and the gut/eyeball.   It's worthwhile, I think, to be included in any discussion but it seems like it's not a be-all end all (necessarily) as to whether player X did well last night.  

 

At the same time, I think really good stats for last night should probably reflect a good game whereas I'm not so sure bad corsi stats necessarily reflect a bad game.   I don't know, but I'm willing to guess it's a mixed bag.

 

Great comparison.

Mixed bag is all I have been preaching. Use it as a tool but the tool box better have more than one tool in it. A guy can have great corsi numbers but if he is a cancer in the locker room it doesn't show up in the stats at the end of the year but likely had a profound impact on the ice. Same as a 4th liner can be victimized and have a bad corsi number but is the quintessential mentor to the young players (work ethic and such), think Lappy... But still, it won't show up in the stats. The same can be said for the intangibles that left with Hartnell. It's showing up but not in the stats.

But intangibles DO show up in the stats. If a player plays with better/more "intangibles" it should show up in the box score somewhere. If a player is a bad teammate, that will impact the on-ice performance somehow as well.

[aziz]

I'm genuinely curious to hear the main arguments against advanced stats.

 

you mentioned +/- as an agreed upon almost-useless stat.  +/- is a measure of how often a team scores when a given player is on the ice versus how often they are scored against.  

 

corsi is a measure of how often a team tries to take a shot on goal when a given player is on the ice versus how often there are shot attempts against.  

 

all of the reasons people look at +/- as vague and uninsightful apply to corsi.  both reward and punish players for things out of their control, both attempt to sketch a hazy outline of individual performance by measuring team stats while said individual is on the ice.  in both cases, a large enough sample will usually produce a trendline that roughly follows the player's actual contribution...but only roughly and only usually.  neither case reflects the play of the specific player clearly enough for anything more than a broad strokes comparison.

 

beyond all of that, the fact that corsi measures shot attempts themselves, not shots that actually become threats to score, is a problem to me.  i have seen plenty of players that just like to shoot.  they shoot from bad angles, they shoot into traffic, they shoot wide, but they shoot.  give them 3 feet of space in the offensive zone, and they'll fire away.  into shin pads, 10 feet wide, lollypops, whatever.  + corsi for each of those, but ineffective and pointless.  i've also seen players that hold the puck forever, but not shoot much.  forsberg was like that.  45 seconds of offensive possession resulting in a single shot on net.  corsi killer, but an effective hockey player.  

 

and finally, +/- reflects a thing that literally decides the winner and loser of a game.  if you are going to use a stat that is derived from the play of all 5 skaters on the ice and fails to directly measure anything about an individual player's performance or contribution...why would you decide to go with something that reflects an in-game occurrence that has no direct bearing on the team's success of failure?  isn't +/- really a far more telling metric than corsi, given that corsi is a measure of the things that hopefully turn into the events of which +/- is the measure?

 

i don't think +/- is particularly useful overall, but i find it to be a better indicator of player performance than corsi.  the thing being measured just matters almost infinitely more.

[RiskyBryzness]

you mentioned +/- as an agreed upon almost-useless stat.  +/- is a measure of how often a team scores when a given player is on the ice versus how often they are scored against.  

 

corsi is a measure of how often a team tries to take a shot on goal when a given player is on the ice versus how often there are shot attempts against.  

 

all of the reasons people look at +/- as vague and uninsightful apply to corsi.  both reward and punish players for things out of their control, both attempt to sketch a hazy outline of individual performance by measuring team stats while said individual is on the ice.  in both cases, a large enough sample will usually produce a trendline that roughly follows the player's actual contribution...but only roughly and only usually.  neither case reflects the play of the specific player clearly enough for anything more than a broad strokes comparison.

 

beyond all of that, the fact that corsi measures shot attempts themselves, not shots that actually become threats to score, is a problem to me.  i have seen plenty of players that just like to shoot.  they shoot from bad angles, they shoot into traffic, they shoot wide, but they shoot.  give them 3 feet of space in the offensive zone, and they'll fire away.  into shin pads, 10 feet wide, lollypops, whatever.  + corsi for each of those, but ineffective and pointless.  i've also seen players that hold the puck forever, but not shoot much.  forsberg was like that.  45 seconds of offensive possession resulting in a single shot on net.  corsi killer, but an effective hockey player.  

 

and finally, +/- reflects a thing that literally decides the winner and loser of a game.  if you are going to use a stat that is derived from the play of all 5 skaters on the ice and fails to directly measure anything about an individual player's performance or contribution...why would you decide to go with something that reflects an in-game occurrence that has no direct bearing on the team's success of failure?  isn't +/- really a far more telling metric than corsi, given that corsi is a measure of the things that hopefully turn into the events of which +/- is the measure?

 

i don't think +/- is particularly useful overall, but i find it to be a better indicator of player performance than corsi.  the thing being measured just matters almost infinitely more.

There is more randomness to regular +/- rather than Corsi, and that has been proven.

 

http://www.arcticicehockey.com/2014/6/5/5602668/why-plus-minus-is-the-worst-statistic-in-hockey

[aziz]

There is more randomness to regular +/- rather than Corsi, and that has been proven.

http://www.arcticicehockey.com/2014/6/5/5602668/why-plus-minus-is-the-worst-statistic-in-hockey

What do you mean "proven"?

Everything in that article can be applied equally well to corsi. Because it is collected in the exact same way as +/-, counts things equally outside the player's control as +/-, is every bit as" random" as +/-. It just measures something far less directly important to player or team success.

[RiskyBryzness]

What do you mean "proven"?

Everything in that article can be applied equally well to corsi. Because it is collected in the exact same way as +/-, counts things equally outside the player's control as +/-, is every bit as" random" as +/-. It just measures something far less directly important to player or team success.

http://www.hockeyforums.net/index.php/topic/63720-arguments-against-advanced-stats/?p=211807

[aziz]

@RiskyBryzness

 

ok, a post entitled "Corsi correlates well to scoring chances, as well as goals", which then goes on to demonstrate how taking shots relates reasonably well to getting scoring chances, which then in turn relates reasonably well to actually scoring goals.  with graphs.

 

now, imagine a post entitled, "+/- correlates well to scoring chances, as well as goals".  there'd be a graph of +/- events (i.e., goals) on one axis, with scoring chances on the other axis, and it there would be an all-but-straight line, indicating an almost-direct relationship between the two.  then, there would be a graph that showed +/- events (i.e., goals) on one axis, and goals on another axis, and it would be a straight line at 45º, a one-for-one relationship.

 

so, tell me again why corsi is informative but +/- isn't?

[idahophilly]

What do you mean "proven"?

 

I was waiting for you to show Aziz...

 

As always, you are the sites analytic master.

 

Besides, when it comes to subjective stats (at best), "proven" is in the eyes of the beholder. No rights or wrongs in the final analysis. The only thing that finalizes anything is the final buzzer. 

 

I guarantee you Bryz, if you had disposable money to place on final standings, teams and players, based solely on corsi you wouldn't make the bets. There would be all sorts of doubt and gut feelings (and eyeball tests) coming into play. You might not admit that but it's ok. If you like corsi I say go for it and enjoy. But, not everyone agrees with how it's applied. If they did then Richards and Carter probably would not have been traded. But, those ugly intangibles reared their ugly heads again.

 

Aziz, fascinating way to look at it (probably shewing what you meant but...). Corsi tries to predict the future while +/- is a result of the past... 

[RiskyBryzness]

I'm done. I tried my best. Enjoy the season guys.

[aziz]

Corsi tries to predict the future while +/- is a result of the past... 

 

well, corsi is a result of the past, too.  in the same way +/- is.  in the exact same way +/- is, really.  corsi measures the times a team *could have* scored.  +/- measures the times a team *did* score.  and they both use the same "things any player on the ice did during a particular player's shift" methodology.  which is to say, most of the things that increment someone's corsi number or their +/- will be because of someone else.  someone else scored, someone else lost their check and the other team scored, someone else squeezed off a desperation shot, someone else gave their check an extra two feet of space and they got a shot off.  both statistics are polluted by the actions of the other 9 skaters on the ice in exactly the same way.  

 

only difference is one counts scoring attempts and one counts scoring, period.  how the former can be valuable while the latter is not is beyond me.

[murraycraven]

Too many intangibles and contingencies in hockey, numerous micro-events in the course of every shift that are difficult to quantify.

For starters.

Also: nerds who've never played the game. At least Billy Beane played the game.

I love this post.

Most advanced stats are guys that are nerds who never got on the ice or held a stick.

[aziz]

I'm done. I tried my best. Enjoy the season guys.

 

 

wait, what?  you pulled up a soapbox, threw out a "I'm genuinely curious to hear the main arguments against" invitation, ...and then bail when someone points out that corsi is pretty much usless ryan lambert nonsense and wasn't swayed by a pair of graphs and an article someone else wrote?

 

i mean, you enjoy the season, too, but...that was weird.  we had a corsi expert ready to preach The Word, but then we didn't.  

[murraycraven]

Was this all a ploy to follow a twitter account? I ask this in all honesty....

[idahophilly]

well, corsi is a result of the past, too.  in the same way +/- is.  in the exact same way +/- is, really.  corsi measures the times a team *could have* scored.  +/- measures the times a team *did* score.  and they both use the same "things any player on the ice did during a particular player's shift" methodology.  which is to say, most of the things that increment someone's corsi number or their +/- will be because of someone else.  someone else scored, someone else lost their check and the other team scored, someone else squeezed off a desperation shot, someone else gave their check an extra two feet of space and they got a shot off.  both statistics are polluted by the actions of the other 9 skaters on the ice in exactly the same way.  

 

only difference is one counts scoring attempts and one counts scoring, period.  how the former can be valuable while the latter is not is beyond me.

 

Yeah, I see what you are saying. I thought that after I posted it. They are both based on past events. I'll say I look for the results more though, right or wrong. I'd rather have a guy who shoots twice and gets a goal versus a guy who makes 10 attempts and gets no goals...

[idahophilly]

Was this all a ploy to follow a twitter account? I ask this in all honesty....

 

I'm safe right now for not getting sucked in since I don't really use twitter except for some business stuff.

[idahophilly]

wait, what?  you pulled up a soapbox, threw out a "I'm genuinely curious to hear the main arguments against" invitation, ...and then bail when someone points out that corsi is pretty much usless ryan lambert nonsense and wasn't swayed by a pair of graphs and an article someone else wrote?

 

i mean, you enjoy the season, too, but...that was weird.  we had a corsi expert ready to preach The Word, but then we didn't.  

 

Good job Aziz, you scared him away! 

[aziz]

Good job Aziz, you scared him away! 

 

lol, didn't mean to.  thought this could be interesting, there's this whole wave of advanced stats people, and so much of it seems like useless fiddling with spreadsheets, wanted to see if there was something i was missing.

 

and, getting distracted by the corsi thing is too bad.  corsi is pointless (unless you are a goalie and are using it to scout your next start, who tends to shoot a lot, who tends to hold onto the puck, etc...which was what it was created to do), but there are some "advanced" stats that are interesting.  zone start versus zone finish, for one.  absolute possession numbers, if available, would be great.  line scored a goal/got scored on (+/-) says little about a given player.  line took a shot attempt/line gave up a shot attempt (corsi) also says very little.  line possessed the puck 80% of the time it was on the ice/line defended opposition possession 20% of the time it was on the ice says a lot, though.  those numbers aren't changed drastically by random things, a good possession line can recover from a hiccup here or there and maintain their possession time...a bad possession line will rarely accidentally maintain 45 seconds of possession pressure.  i'm not even sure if those numbers are tracked, though.

 

so, too bad a thread about advanced stats stalled out on the most useless of them all.