Does anyone miss hockey analysis before analytics?

[Pablo El Perro]

So... this morning the Edmonton Oilers fired their coach after the team started 3-9-1 (probably they decided to fire him after the team started 2-9-1, they beat Seattle last night).

In the meantime, as of today, here is how the Oilers measure up with regard to expected goals for percentage at 5v5:

View attachment 767407

Yes, that 57.82% expected goals for percentage is good for best in the league.

I'd say actual results are much more important than the hypotheticals implied by fancy stats. How about you

Again, statistics can't tell you anything but a should about a singular element of a sample. A should isn't a would or will.

 

[Mattilaus]

So I guess my take is analytics are helpful for explaining this or that after the fact, but they aren't the game itself.

For example,the Blackhawks PDO was better than the Panthers today, the Hawks still lost in regulation.

Using PDO to predict or describe a single game is useless. The sample size is too small. If I see a guy who eats right and exercises every day and he drops dead from a brain aneurysm at 36, do I take from this that exercise and eating well don't matter? Of course not, because it's only one person and we all know over the grand scheme of things you are much more likely to live a long and healthy life by exercising and eating well.

There will always be outliers, it's bad faith to say one outlier means all statistics are garbage or useless.

 

[Voight]

Analytics in a nustshell (quote is from the JFK movie)

The FBI says they can prove it through physics in a nuclear laboratory. Of course they can prove it. Theoretical physics can also prove that an elephant can hang off a cliff with its tail tied to a daisy! But use your eyes, your common sense.

 

[Ianturnedbull]

Analysis based on analytics means the analysis is actually based on some truth.

Before you had idiots like Mike Milbury who had analyst jobs simply because they had a history in the league.

what does this even mean?

if someone sees a good team play well: goaltending, passing, good defensive play, timely scoring, etc. that is not the truth? a person who ignores the goaltending, passing, good defensive play, in favour of analysis based on Corsi, Fenwick, PDO, Zone Starts, RAPM, etc.is following a truth?

 

[KaseMeOutside]

The influx of analytics have made virtually every sport worse from an entertainment POV.

In regards to hockey, the analytics have definitely caused a decrease in the physicality part of the game and the sport is worse for it

 

[TheRumble]

I'm still waiting for hockey to get out of the stone age to be honest with you. The current state of analytics in hockey is far behind other sports.

I wish their were more tactical analysis in hockey. We hear terms like 1-1-3, 1-3-1, 1-2-2 with barely any understanding of what those systems mean.

 

[TheGreatOne11]

I don't know why anybody uses "analytics", everybody knows hockey players born when bulbasaur is in retrograde are the best players of their generation.

 

[Ianturnedbull]

I'm still waiting for hockey to get out of the stone age to be honest with you. The current state of analytics in hockey is far behind other sports.

I wish their were more tactical analysis in hockey. We hear terms like 1-1-3, 1-3-1, 1-2-2 with barely any understanding of what those systems mean.

what do they mean?

 

[Pablo El Perro]

Using PDO to predict or describe a single game is useless. The sample size is too small. If I see a guy who eats right and exercises every day and he drops dead from a brain aneurysm at 36, do I take from this that exercise and eating well don't matter? Of course not, because it's only one person and we all know over the grand scheme of things you are much more likely to live a long and healthy life by exercising and eating well.

There will always be outliers, it's bad faith to say one outlier means all statistics are garbage or useless.

That's what I said. I didn't say they are garbage. Just statistics and anti-statistics crowds don't usually acknowledge it. I used to teach stats, which makes the hate both ways annoying to say the least.

what do they mean?

The 1s, 2s, and 3s are just positional coverage.

 

[Adam Fox Time]

what does this even mean?

if someone sees a good team play well: goaltending, passing, good defensive play, timely scoring, etc. that is not the truth? a person who ignores the goaltending, passing, good defensive play, in favour of analysis based on Corsi, Fenwick, PDO, Zone Starts, RAPM, etc.is following a truth?

It’s the difference between the Eye Test and objective statistics.

 

[Mattilaus]

That's what I said. I didn't say they are garbage. Just statistics and anti-statistics crowds don't usually acknowledge it. I used to teach stats, which makes the hate both ways annoying to say the least.

Ah, my mistake.

 

[Mattilaus]

what does this even mean?

if someone sees a good team play well: goaltending, passing, good defensive play, timely scoring, etc. that is not the truth? a person who ignores the goaltending, passing, good defensive play, in favour of analysis based on Corsi, Fenwick, PDO, Zone Starts, RAPM, etc.is following a truth?

You know what kind of evidence leads to the most wrongful convictions? Eye witness testimony. People have biases of all kinds and are often bad at evaluating play based off their eyes. The strongest cases have both eye witnesses AND physical evidence or in this case, stats, that both back up the same conclusion.

 

[Ezekial]

Hockey is the hardest sport to handicap for sportsbooks, it has the least amount of predictability. Shots for and against lead to goals either way, in simplest terms they are the easiest way to determine the likelihood you will win - just like goals for and against at ES are an indicator if you will win or not. It's not the only thing that matters though and not all shots from the same spot have the same level of danger. But like was said in a great post earlier, +/- isn't exactly the best indicator that whether or not you move the needle in the + direction all the time.

 

[Strangle]

Yep. All this XY%FU garbage means nothing to me. l watch hockey with my eyes, not a stat sheet.

The only thing these stats would be good for is if they were predictive, but as it is they are only ever interesting to look at in hindsight to try to argue which historical teams or players may or may not have been better.

We have much better stats to use to explain actual performance. Mainly wins/losses and Goals/Assists.

If you’re trying to evaluate players who have garbage production, I guess these might be able to split hairs on that, but I think most of this is driven by coaching anyway. So just coach the player in what you want them to do on the ice , instead of running analytics to figure out if they’re already being coached to perform that way or not

 

[Ianturnedbull]

You know what kind of evidence leads to the most wrongful convictions? Eye witness testimony. People have biases of all kinds and are often bad at evaluating play based off their eyes. The strongest cases have both eye witnesses AND physical evidence or in this case, stats, that both back up the same conclusion.

No. This is a really terrible comparison.

This is hockey and not a court of law.

I do agree that wacky/emotional fans have biases, but people with experience have a pretty grounded view of teams, and talented players. They know enough when a team is good or bad. They also know enough about hockey to know how unpredictable the playoffs can be.

 

[TLEH]

So... this morning the Edmonton Oilers fired their coach after the team started 3-9-1 (probably they decided to fire him after the team started 2-9-1, they won their game last night).

In the meantime, as of today, here is how the Oilers measure up with regard to expected goals for percentage at 5v5:

View attachment 767407

Yes, that 57.82% expected goals for percentage is good for best in the league.

I'd say actual results are much more important than the hypotheticals implied by fancy stats. How about you

I think you just made a great point. You are using a 12 game sample size to say that xGF% is bad. The exact people this thread is talking about!

You will find exactly zero teams that lead the league in xGF% and finish last in the league after 82 games.

 

[Ianturnedbull]

It’s the difference between the Eye Test and objective statistics.

When did subjectivity of any sort (not just hockey) become a bad word?

 

[Pablo El Perro]

No. This is a really terrible comparison.

This is hockey and not a court of law.

I do agree that wacky/emotional fans have biases, but people with experience have a pretty grounded view of teams, and talented players. They know enough when a team is good or bad. They also know enough about hockey to know how unpredictable the playoffs can be.

Yeah, it's a bad analogy. Ideally, stats measure what eyewitnesses see as far as shots, goals, etc. Stats themselves don't do any seeing or counting. They require a witness.

When did subjectivity of any sort (not just hockey) become a bad word?

When stats people assumed nobody watching was accounting for the very elements put into statistical analysis. Trying to make stats purely objective misunderstands basic epistemology.

 

[Mattilaus]

No. This is a really terrible comparison.

This is hockey and not a court of law.

I do agree that wacky/emotional fans have biases, but people with experience have a pretty grounded view of teams, and talented players. They know enough when a team is good or bad. They also know enough about hockey to know how unpredictable the playoffs can be.

Yeah, it's a bad analogy. Ideally, stats measure what eyewitnesses see as far as shots, goals, etc. Stats themselves don't do any seeing or counting. They require a witness.

You can dislike the analogy and that's fine, but pretending the "experts" who go off their "feel" are always right and not subject to the same bias and mistakes that anyone else is is foolhardy. We see multiple instances every year of GMs signing bad players for too much money because they just like them and most of the time it ends predictably bad. I doubt however this will move the discussion forward because no matter what I say it can simply be shrugged off because I am not an "expert" and the stats don't matter.

 

[Ianturnedbull]

You can dislike the analogy and that's fine, but pretending the "experts" who go off their "feel" are always right and not subject to the same bias and mistakes that anyone else is is foolhardy. We see multiple instances every year of GMs signing bad players for too much money because they just like them and most of the time it ends predictably bad. I doubt however this will move the discussion forward because no matter what I say it can simply be shrugged off because I am not an "expert" and the stats don't matter.

No one is arguing that someone is "always right".

 

[Pablo El Perro]

You can dislike the analogy and that's fine, but pretending the "experts" who go off their "feel" are always right and not subject to the same bias and mistakes that anyone else is is foolhardy. We see multiple instances every year of GMs signing bad players for too much money because they just like them and most of the time it ends predictably bad. I doubt however this will move the discussion forward because no matter what I say it can simply be shrugged off because I am not an expert and the stats don't matter.

Subjectivity isn't the same as feel. All stats require a subject who collects them. Of course, some witnesses are suspect. And nobody is shrugging anything. My point was that stats aren't objective, meaning without subjective points of view. Quit deliberately misunderstanding my point.

 

[Gaylord Q Tinkledink]

I miss Glenn Healy saying a team is trash and isn't good, but they're losing 1-0 going into the 3rd and then Healy talking about how that team has been dominant all game when they win 4-1

 

[PainForShane]

I think you just made a great point. You are using a 12 game sample size to say that xGF% is bad. The exact people this thread is talking about!

You will find exactly zero teams that lead the league in xGF% and finish last in the league after 82 games.

Yeah, what I believe about this is a bit more nuanced, harder to explain in a few sentences on a message board but I'll try. Basically I believe all of this data is a tool, if you use it in a way that makes sense, it will help to explain some things some of the time. So, why not use this tool, if you understand what you're doing it can only show different parts of the game that maybe you wouldn't have considered otherwise.

However, analytics will not be able to explain all things all of the time (ie EDM's league best xGF% combined with a close to league worst record), and in those divergent cases, imo real-life wins / losses and actual goals for / goals against clearly matter a lot more than Corsi and Fenwick percentages. And of course your point about sample sizes is also true (especially with PDO), but that's not really what I was looking to show.

Anyway, not sure that what I said in paragraphs 1 and 2 made sense to most ppl reading this, but there it is in writing

 

[TLEH]

Yeah, what I believe about this is a bit more nuanced, harder to explain in a few sentences on a message board but I'll try. Basically I believe all of this data is a tool, if you use it in a way that makes sense, it will help to explain some things some of the time. So, why not use this tool, if you understand what you're doing it can only show different parts of the game that maybe you wouldn't have considered otherwise.

However, analytics will not be able to explain all things all of the time (ie EDM's league best xGF% combined with a close to league worst record), and in those divergent cases, imo real-life wins / losses and actual goals for / goals against clearly matter a lot more than Corsi and Fenwick percentages. And of course your point about sample sizes is also true (especially with PDO), but that's not really what I was looking to show.

Anyway, not sure that what I said in paragraphs 1 and 2 made sense to most ppl reading this, but there it is in writing

Hockey is way too luck based to be able to make any conclusions on a small sample size. Those "deserve to win" meters are a joke.

This isn't baseball. You can't quantify things as easily because there isn't enough controlled stops and starts in a game.

But over the course of seasons there are statistics that are a good predictor of success overall.

 

[Mattilaus]

Subjectivity isn't the same as feel. All stats require a subject who collects them. Of course, some witnesses are suspect. And nobody is shrugging anything. My point was that stats aren't objective, meaning without subjective points of view. Quit deliberately misunderstanding my point.

I am not deliberately misunderstanding your point of view my response was largely toward the other gentleman. I included you because you also responded to my post.

There are different levels of subjectivity and data collection is different than interpretation. A person using their own subjective feel to say how good a team is performing is different than someone counting shots. For shots either it hit the goalie, hit the net, or it didn't. Can mistakes be made? Sure, but lets not pretend like that's the same as someone watching the game and then just declaring a player played well based off nothing other than their own emotions or opinions. Thats the argument I was making against the other poster who indicated that experts or people with experience are grounded and know when a team is good or bad better than analytics can identify.

I included you because you chimed in with him because you didn't like the analogy. Using what you have said everything is subjective. DNA? Well a human had to collect it, transport it, test it, and then interpret the data, so it is subjective too. Ballistics? Still reliant on an expert saying they match. There is almost nothing I can think of in the world that doesn't require some sort of human interaction where data can be manipulated.