Stats....

[cccsberg]

I thought I would start a thread specifically about stats, their uses and abuses.  Many on here believe stats tell the tale, while others push them too far or out of context.  It's the new age of stats-driven hockey, does the eye test still hold water?  Weigh in with your favourite stats, their implications and other observations.

[cccsberg]

Here's a quote I pulled off of the Flames Nation's post-discussion of the Blue Jackets game.  All stats are 5v5 for the season.

 

"Did you know, that a line can be an offensive juggernaut, and defensively responsible? Check out NYI’s line of Anders Lee, Brock Nelson, and Josh Bailey. 5.20 GF/60, 0.00 GA/60. 10 goals for, none against in 115 minutes of play. Compare that to Mony's line at 3.96 GF/60 and 2.77 GA/60, for 20 GF and 14 GA in 303 minutes."

 

Here's a great example of some interesting but perhaps not-so-great stats, with serious implications if you really put absolute faith in bare stats.  One of the implications is that the NYI trio is far superior to the Flames trio, and if that is so, would anyone be willing to do a 3 for 3 straight-up trade between the teams?  What could go wrong?

[flames-fan-in-jets-land]

I think in that scenario the main question is  why is Calgarys top line playing more than double the 5v5 minutes?  Are PP and PK affecting time that much? Is that even the Isles top line? Does their coach like to juggle lines?  Hard stats can be useful in painting a vague picture but there are always so many variables for them to be the only thing to base ones judgement or decision on.

[cross16]

1 hour ago, cccsberg said:

Here's a quote I pulled off of the Flames Nation's post-discussion of the Blue Jackets game.  All stats are 5v5 for the season.

 

"Did you know, that a line can be an offensive juggernaut, and defensively responsible? Check out NYI’s line of Anders Lee, Brock Nelson, and Josh Bailey. 5.20 GF/60, 0.00 GA/60. 10 goals for, none against in 115 minutes of play. Compare that to Mony's line at 3.96 GF/60 and 2.77 GA/60, for 20 GF and 14 GA in 303 minutes."

 

Here's a great example of some interesting but perhaps not-so-great stats, with serious implications if you really put absolute faith in bare stats.  One of the implications is that the NYI trio is far superior to the Flames trio, and if that is so, would anyone be willing to do a 3 for 3 straight-up trade between the teams?  What could go wrong?

 

Big flaw with that analysis is small sample size bias. 115 minutes of 5 on 5 play is pretty much nothing, it's probably 8-10 games. So if you are willing to draw conclusions based on that sample size set I don't think the stats are the issues it's the conclusions being drawn and IMO that's where stats get dangerous. 

 

I like stats and as i'm sure most here know an analytics guy as well. Specifically I put most of my focus on chances (both created and given up) per 60 mins at 5 on 5 as my main stats. I look at shooting percentage a lot too, and while I look at both Corsi/.Fenwick on a frequent basis i don't put as much value in that stat anymore. I still, and always will, use stats in conjunction with the eye test. Reason I like analytics so much is they offer a lot more depth of information and allow you to challenge what your eyes see. Quite often I find myself changing an opinion, or re looking at a certain player/situation, because the stats didn't support my initial eye test and I need to look deeper. 

[travel_dude]

1 hour ago, cross16 said:

 

Big flaw with that analysis is small sample size bias. 115 minutes of 5 on 5 play is pretty much nothing, it's probably 8-10 games. So if you are willing to draw conclusions based on that sample size set I don't think the stats are the issues it's the conclusions being drawn and IMO that's where stats get dangerous. 

 

I like stats and as i'm sure most here know an analytics guy as well. Specifically I put most of my focus on chances (both created and given up) per 60 mins at 5 on 5 as my main stats. I look at shooting percentage a lot too, and while I look at both Corsi/.Fenwick on a frequent basis i don't put as much value in that stat anymore. I still, and always will, use stats in conjunction with the eye test. Reason I like analytics so much is they offer a lot more depth of information and allow you to challenge what your eyes see. Quite often I find myself changing an opinion, or re looking at a certain player/situation, because the stats didn't support my initial eye test and I need to look deeper. 

 

Good points.

I think the stats can change dramatically during a game (chasing, score effects, etc.), leading to it looking like we lost the "Corsi Battle" but won the game.  For instance, Neal was dominant in the CHI game, while the whole team was buried in the CBJ game.

In the CHI, it was not even close for most of the game, but we let up.  

 

Anyway, I focus on the normal stats that easily support or reject the eye test.  SA%, GAA, +/-, shots, faceoff%...

I like the inclusion of scoring chances, which can be subjective.  HDSC is one that is easily supported by the eye test.  Or so it seems.

[flames-fan-in-jets-land]

1 hour ago, travel_dude said:

 

Good points.

I think the stats can change dramatically during a game (chasing, score effects, etc.), leading to it looking like we lost the "Corsi Battle" but won the game.  For instance, Neal was dominant in the CHI game, while the whole team was buried in the CBJ game.

In the CHI, it was not even close for most of the game, but we let up.  

 

Anyway, I focus on the normal stats that easily support or reject the eye test.  SA%, GAA, +/-, shots, faceoff%...

I like the inclusion of scoring chances, which can be subjective.  HDSC is one that is easily supported by the eye test.  Or so it seems.

I think thats the point stats people are getting to. Stats are stats, backed by hard numbers. When you start adding subjective areas such as high or low scoring chances then your hard stats start becoming probabilities and trends which may be more reflective of a game or player given the dynamics of the game.

[travel_dude]

17 minutes ago, flames-fan-in-jets-land said:

I think thats the point stats people are getting to. Stats are stats, backed by hard numbers. When you start adding subjective areas such as high or low scoring chances then your hard stats start becoming probabilities and trends which may be more reflective of a game or player given the dynamics of the game.

 

I don't think it take a rocket scientist to recognize a scoring chance.  

High danger that's really a dangerous chance is subjective more than whether it's in a high danger area.

[cross16]

2 minutes ago, travel_dude said:

 

I don't think it take a rocket scientist to recognize a scoring chance.  

High danger that's really a dangerous chance is subjective more than whether it's in a high danger area.

 

This is one of my beefs with those who question (for lack of a better term) analytics and scoring chances. Why is a high danger scoring chance more subjective than just a general shot? And even if it is do we ignore that just simply looking at things like shots present a false comparison because how can you judge a single statistics that has so many different variations?

So what's worse the small potential for subjective bias or a false comparison? IMO, simply looking at things like shots, saves/save % etc are far more risky and present more opportunities for error. 

[travel_dude]

1 hour ago, cross16 said:

 

This is one of my beefs with those who question (for lack of a better term) analytics and scoring chances. Why is a high danger scoring chance more subjective than just a general shot? And even if it is do we ignore that just simply looking at things like shots present a false comparison because how can you judge a single statistics that has so many different variations?

So what's worse the small potential for subjective bias or a false comparison? IMO, simply looking at things like shots, saves/save % etc are far more risky and present more opportunities for error. 

 

Did it sound like I was against using those stats?

What I was suggesting is that a shot from a high danger area isn't always a great scoring chance.

 

Shots only tell part of the story.  Last year we generated a ton, but they weren't great chances.

+/- tells very little, but is a good barometer of even strength goals for/against.

 

I look at the game summary, who's on for what.  It gives you an idea of the performance of players in situations.

 

[cross16]

17 minutes ago, travel_dude said:

 

Did it sound like I was against using those stats?

What I was suggesting is that a shot from a high danger area isn't always a great scoring chance.

 

 

 

No, but you provided the commentary I hear a lot. I get what you are saying, but at the same time what i'm saying is why is a shot a shot? Is a dump in from the blue line a shot? Is it a save?

 

It works both ways. That's what i'm saying. Not necessarily directed at you

[flames-fan-in-jets-land]

It sound like we're all arguing the same point.  My issue is the use of the term "stats" when discussing something open to analytics.

[cccsberg]

8 hours ago, flames-fan-in-jets-land said:

I think in that scenario the main question is  why is Calgarys top line playing more than double the 5v5 minutes?  Are PP and PK affecting time that much? Is that even the Isles top line? Does their coach like to juggle lines?  Hard stats can be useful in painting a vague picture but there are always so many variables for them to be the only thing to base ones judgement or decision on.

I don’t watch the NYI games but I suspect that Barzal is their line 1.  Yes, the TOI disparity between the two trios is dramatic.

[cccsberg]

7 hours ago, cross16 said:

 

Big flaw with that analysis is small sample size bias. 115 minutes of 5 on 5 play is pretty much nothing, it's probably 8-10 games. So if you are willing to draw conclusions based on that sample size set I don't think the stats are the issues it's the conclusions being drawn and IMO that's where stats get dangerous. 

 

I like stats and as i'm sure most here know an analytics guy as well. Specifically I put most of my focus on chances (both created and given up) per 60 mins at 5 on 5 as my main stats. I look at shooting percentage a lot too, and while I look at both Corsi/.Fenwick on a frequent basis i don't put as much value in that stat anymore. I still, and always will, use stats in conjunction with the eye test. Reason I like analytics so much is they offer a lot more depth of information and allow you to challenge what your eyes see. Quite often I find myself changing an opinion, or re looking at a certain player/situation, because the stats didn't support my initial eye test and I need to look deeper. 

One of the big problems I have with stats like this is the behind the scene assumptions that are implied but almost never true.  As you mentioned, TOI is a good example.  Quality of competition as well as quality of opponent (team) I submit are two others.  Game situation is another one.  Stats are fine as far as they go, but guys who put up stats alone to prove/make a point are usually out to lunch when you dig deeper.

[cccsberg]

4 hours ago, cross16 said:

 

This is one of my beefs with those who question (for lack of a better term) analytics and scoring chances. Why is a high danger scoring chance more subjective than just a general shot? And even if it is do we ignore that just simply looking at things like shots present a false comparison because how can you judge a single statistics that has so many different variations?

So what's worse the small potential for subjective bias or a false comparison? IMO, simply looking at things like shots, saves/save % etc are far more risky and present more opportunities for error. 

I don’t have a problem with analytics, I just have a problem with the misapplication of analytics and inconsistencies in data collection.  Watching a game the other night on a face-off the Centre drew it back and to the side, yet the opposition player beat the Flames player to the puck and skated away with possession.  So, did the Centre “win” the draw or not?  And if not, how is the Centre at ‘fault’? (I.e. lower FO%)

[travel_dude]

3 hours ago, cross16 said:

 

No, but you provided the commentary I hear a lot. I get what you are saying, but at the same time what i'm saying is why is a shot a shot? Is a dump in from the blue line a shot? Is it a save?

 

It works both ways. That's what i'm saying. Not necessarily directed at you

 

I also think that shots are counted differently in every rink.

I watch the game and it show 10 shots, and look up after a chance, and suddenly it's gone up by 3.

[conundrumed]

The blown coverage by one player affects the stats for the other 5.

And there are far too many instances of goal or high danger chance due to 1 players gaffe or the goalie letting in bad goals.

So realistically, 1 player’s gaffe is 16.7% affecting the other 83.3% of players on the ice.

Is that accounted for?

[conundrumed]

And I'm not railing against stats, but stop calling them "advanced". What I do rail against is using the stats alone to prove a point, which we're inundated with now.

That's what gives the use the hairy eyeball imho.

As cross spells it out why he enjoys, that's fine. There are more not cross' than cross' as purveyors of stats though.

I'm sure I'm not alone in saying it can be manipulated to beyond annoyance and you really have to dig in and decide what is a legitimate trend, what is circumstances and what is simply a drive by.

My biggest bias is, something useful is created as a means of assisting, potentially helping. Great.

The bandwagon turns it into an industry and it gets buried in too many halftards (can I say that?) with their hands in the cookie jar.

Like every industry everywhere with some new angle, now there's 2,000 stats guys making an economy out of it.

The problem will always be making it human-friendly.

These guys are beat up, getting BTCed out by their wives, playing sick, etc etc.

As far as I know, stats only apply to numbers and are mechanical.

They fall far short of being able to label specific players.

But...it's an industry, they will.

As much as I read the hard work some Athletic writers put in on stats, I'm left not agreeing a lot.

I'd be hardcore anti-"advanced" stats if not for cross and kehatch tbh.

There has to be a balance, not inundated like it's, ummm, grospel?

 

[cross16]

And to be fair even though I like analytics they still aren't great. I think they are better than "traditional" stats, but I don't think they are the answer at all.

 

Hockey is well behind other sports in terms of stat tracking and understanding.

[cccsberg]

21 hours ago, cross16 said:

And to be fair even though I like analytics they still aren't great. I think they are better than "traditional" stats, but I don't think they are the answer at all.

 

Hockey is well behind other sports in terms of stat tracking and understanding.

When the league comes up with a means to accurately and consistently track player movements and gather them into useful stats we'll have something useful to discuss.  Until then, its kind of a crapshoot.  The NHL has been experimenting with uniform-embedded RFID chips that they have highlighted a handful of games the past couple of years.  Perhaps that will pave the way?

[Cowtownguy]

On 2018-12-10 at 10:21 AM, cross16 said:

And to be fair even though I like analytics they still aren't great. I think they are better than "traditional" stats, but I don't think they are the answer at all.

 

Hockey is well behind other sports in terms of stat tracking and understanding.

I find reference to analytics rather funny because that is just another name for statistics. Somehow, analytics are smart statistics as opposed to dumb statistics. In reality, much of the best statistics (by mathematicians) were generated in the late 1800s and the 1930s. The statistics that we use today are really just extensions of those developments. In any case, the NHL may well have improved the data to analyze, but they are not about to release what they may have. All of that data will go to the gambling companies. They don't want the regular fan to be able to beat the house on the basis of knowledge.