2018-19 stats and underlying metrics thread

[garret9]

Thought it was time to move on to the next year.

For those new to the board, I run a stats thread like this each year in order to not spam those who do not want to see certain things.

Here you will find things like:
* random numbers I find interesting that are not being written in an article
* me brainstorming ideas for articles
* links to statistical articles I (or others) have found around the in
* questions you have of things you want me to research, and my answers
* any stats work YOU have done

WINNIPEG JETS DEPTHCHART

LW | C | RW Andrew Ladd | Bryan Little | Blake Wheeler Evander Kane | Matthieu Perreault | Michael Frolik TJ Galiardi | Mark Scheifele | Dustin Byfuglien
| Jim Slater | Chris Thorburn
| Eric O`Dell | Anthony Peluso
| | Matt Halischuk

LD | RD Tobias Enstrom | Jacob Trouba Grant Clitsome | Zach Bogosian Mark Stuart | Paul Postma Adam Pardy |
Keaton Ellerby |

G Ondrej Pavelec
Michael Hutchinson

* Click the links to see an analytical focused write up on each player
* These are not lines so do not get stuck up on the order at any position... they are not even in order of my opinion of good to bad

INTRODUCTION TO SHOT METRICS
* Underlying metrics are stupid.... stupid easy
* Limits of Corsi: What can and can't it do
* Importance and misconceptions of hockey analytics
* Some of the most important pieces in hockey analytics

STATISTICAL BASED WEBSITES:

Arctic Ice Hockey - Winnipeg Jets blog that commonly dips into shot metrics (by yours truly) but not limited to
Extra Skater - the most in-depth
Behind The Net - the original website for gaining shot metrics on players and teams
Nice Time on Ice - another raw data site (but different) like Behind the Net
Hockey Analysis - third raw data site I (and others) commonly use, has advantages and disadvantages compared to BTN
Some Kind of Ninja - player usage charts and super shot tracker
Hockey Abstract - place to get GVT, player usage charts and quality starts (amongst other things), by Rob Volman of HP and ESPN

Click to read more...

 

[YWGinYYZ]

Moved a bunch of stats heavy posts that were taking some threads OT to this thread.

 

[Jack722]

Honestly PLE wasn’t too good either. Their efficiency was at best average. LWC was a clearly more efficient line with their goalscoring and their 5 on 5 goal difference, even though there was the inexperienced center in Wheeler there. Yet the line seems to get quite a lot of criticism at these boards.

Funny really that in reality LWC was more efficient and a better line than PLE, but still for some reason PLE gets praised and LWC usually criticised. I guess it’s about the useless Corsi-loving or liking busy looking lines in general. Otherwise I don’t really get the reason why PLE is so much praised by some people here.

I think you're over the top with your corsi criticism. It's not "useless"... it may sometimes have noise, but more often than not it just indicates that hey, you're tending to control the play more. Players who have great shots and great shot selection will probably tend to not be that well-represented in Corsi, but that doesn't mean that corsi is meaningless. It's just something to keep in mind.

For example, Perreault is typically a great corsi player and that's because he's a very smart and tenacious player who both knows how to break up plays and make a pass. He isn't a superb shooter, so his scoring is not great but his lines often tilt the ice. Even when they are in their own end the opponent is having trouble stringing passes together and developing shot attempts. Anyway that's my 'eye-test' of Perreault's game, and I think things like corsi capture this facet of the game very well.

PLE probably gets praised because it both looked good and outshot the opposition.



Also the more I think about it the less I like ELL, for all their sakes. One of Little's best attributes over the years has been that he's pretty opportunistic, with a good shot... he's got a scorer's instinct, he's great at trailing the play and potting goals. That was his bread and butter with Ladd and Wheeler; Wheels driving play and Ladd doing dirty work and getting to the net. Little playmaking and cleaning up for goals. It seems like complete waste to put him with Laine, who is good at exactly the same things.

 

[Romang67]

Honestly PLE wasn’t too good either. Their efficiency was at best average. LWC was a clearly more efficient line with their goalscoring and their 5 on 5 goal difference, even though there was the inexperienced center in Wheeler there. Yet the line seems to get quite a lot of criticism at these boards.

Funny really that in reality LWC was more efficient and a better line than PLE, but still for some reason PLE gets praised and LWC usually criticised. I guess it’s about the useless Corsi-loving or liking busy looking lines in general. Otherwise I don’t really get the reason why PLE is so much praised by some people here.

That you don't understand the use of something doesn't make it useless.

 

[Tommigun]

Honestly PLE wasn’t too good either. Their efficiency was at best average. LWC was a clearly more efficient line with their goalscoring and their 5 on 5 goal difference, even though there was the inexperienced center in Wheeler there. Yet the line seems to get quite a lot of criticism at these boards.

Funny really that in reality LWC was more efficient and a better line than PLE, but still for some reason PLE gets praised and LWC usually criticised. I guess it’s about the useless Corsi-loving or liking busy looking lines in general. Otherwise I don’t really get the reason why PLE is so much praised by some people here.

Some people are just so fancy that they like fancy stats more than the actual result. Corsi etc are just MODELS someone has come up with to predict an outcome, of course the actual outcome should be more important (and it is for most people). Weather is hard to predict so there’s weather forecasts built on different models, but I’ve yet to meet a single person who would hold on to a forecast if it said it should be sunny when it’s raining outside, and claim that the model was right while the weather got it wrong. It never ceases to amaze me.

 

[Tommigun]

I think you're over the top with your corsi criticism. It's not "useless"... it may sometimes have noise, but more often than not it just indicates that hey, you're tending to control the play more. Players who have great shots and great shot selection will probably tend to not be that well-represented in Corsi, but that doesn't mean that corsi is meaningless. It's just something to keep in mind.

For example, Perreault is typically a great corsi player and that's because he's a very smart and tenacious player who both knows how to break up plays and make a pass. He isn't a superb shooter, so his scoring is not great but his lines often tilt the ice. Even when they are in their own end the opponent is having trouble stringing passes together and developing shot attempts. Anyway that's my 'eye-test' of Perreault's game, and I think things like corsi capture this facet of the game very well.

PLE probably gets praised because it both looked good and outshot the opposition.



Also the more I think about it the less I like ELL, for all their sakes. One of Little's best attributes over the years has been that he's pretty opportunistic, with a good shot... he's got a scorer's instinct, he's great at trailing the play and potting goals. That was his bread and butter with Ladd and Wheeler; Wheels driving play and Ladd doing dirty work and getting to the net. Little playmaking and cleaning up for goals. It seems like complete waste to put him with Laine, who is good at exactly the same things.

I think it’s very telling that Perreault is a Corsi friendly player, and he went like 30 games without a goal. Corsi is a completely arbitrary model someone has come up with. Of course actual results should take precedence, Corsi is just a tool that’s better than a guess to predict said results. I don’t think Corsi himself meant the Corsi numbers should eclipse the actual results. Some people just take them too far.

 

[Ippenator]

I think you're over the top with your corsi criticism. It's not "useless"... it may sometimes have noise, but more often than not it just indicates that hey, you're tending to control the play more. Players who have great shots and great shot selection will probably tend to not be that well-represented in Corsi, but that doesn't mean that corsi is meaningless. It's just something to keep in mind.

For example, Perreault is typically a great corsi player and that's because he's a very smart and tenacious player who both knows how to break up plays and make a pass. He isn't a superb shooter, so his scoring is not great but his lines often tilt the ice. Even when they are in their own end the opponent is having trouble stringing passes together and developing shot attempts. Anyway that's my 'eye-test' of Perreault's game, and I think things like corsi capture this facet of the game very well.

PLE probably gets praised because it both looked good and outshot the opposition.



Also the more I think about it the less I like ELL, for all their sakes. One of Little's best attributes over the years has been that he's pretty opportunistic, with a good shot... he's got a scorer's instinct, he's great at trailing the play and potting goals. That was his bread and butter with Ladd and Wheeler; Wheels driving play and Ladd doing dirty work and getting to the net. Little playmaking and cleaning up for goals. It seems like complete waste to put him with Laine, who is good at exactly the same things.

And I agree with your estimation about Perreault as a player. He is exactly how you described. But his kind is not a very useful player unless he is exactly the great kind of a supporting puck hound with above average skill that is an extremely good player to support the kind of a duo that Scheifele and Laine are when they play together.

But with less efficient players like Little and even Ehlers the line becomes more like some busy bees without doing too well the things that still win the games, which are scoring clearly more goals than their opponents score. Thus the Corsi that he brings is pretty much useless by itself, unless he plays with the kind of players that are able to put the puck in the net. This is exactly why Corsi by itself is just next to useless. Some people just look at the Corsi numbers so blindly without understanding how much different kind of skills and efficiency between different players just simply make those shot statistics next to useless when comparing exactly individuals. With teams I can see a bit more of use with those kind of statistics, but by comparing individuals and their importance to their team, Corsi and Corsi-based stats are simply bad and misleading stats.

 

[Halberdier]

I don’t think Corsi himself meant the Corsi numbers should eclipse the actual results. Some people just take them too far.

Jim Corsi was a goalie who wanted to measure better the amount of work goalies need to do during games. That was because NA statistics for shots did get it wrong and they still do: only shots against goal are counted as shots for example in SH%.

So he invented this valuable tool for measuring the amount of work for goalies, and other people used that Corsi-hammer for other nails too.

It was going OK as long as they used Corsi to predict things in future. But some guys got it backwards and since then many people genuinely think that Corsi and Corsi-based stats tells the reality, while stats based on actual results are just too random. When they do not match their Corsi-based results, they say that and that player is either "lucky" or "unlucky/snake-bitten". And when guys like Perreault are "unlucky/snake-bitten" for season after season, that's just OK for them, while it is actually impossible in the real world.

 

[Duke749]

I think it’s very telling that Perreault is a Corsi friendly player, and he went like 30 games without a goal. Corsi is a completely arbitrary model someone has come up with. Of course actual results should take precedence, Corsi is just a tool that’s better than a guess to predict said results. I don’t think Corsi himself meant the Corsi numbers should eclipse the actual results. Some people just take them too far.

It’s telling of nothing more then it says yet you’re trying to say it says more then that but it doesn’t. You’re actually overthinking something that I’m not sure you understand.

 

[Mortimer Snerd]

I think it’s very telling that Perreault is a Corsi friendly player, and he went like 30 games without a goal. Corsi is a completely arbitrary model someone has come up with. Of course actual results should take precedence, Corsi is just a tool that’s better than a guess to predict said results. I don’t think Corsi himself meant the Corsi numbers should eclipse the actual results. Some people just take them too far.

Corsi is a model but it is not arbitrary. It has been proven to be an effective predictor but it is like the difference between climate and weather. It is not great for predicting micro results. It is very good at predicting macro results.

During that 30 game span Perreault had a sh% of 0. You don't expect that to be the norm going forward. That is 'luck', or variance if you prefer. It is an outlying data point as far as the model is concerned.

Of course Corsi should not eclipse actual results. Has anyone ever said they should? Anyone? Ever?

 

[Halberdier]

Corsi is a model but it is not arbitrary. It has been proven to be an effective predictor but it is like the difference between climate and weather. It is not great for predicting micro results. It is very good at predicting macro results.

During that 30 game span Perreault had a sh% of 0. You don't expect that to be the norm going forward. That is 'luck', or variance if you prefer. It is an outlying data point as far as the model is concerned.

Of course Corsi should not eclipse actual results. Has anyone ever said they should? Anyone? Ever?

Corsi will never match actual results. Neither macro nor micro, unless that guy is Joe Average. Average shooter, average IQ, average line mates and average game plan. Then, of course, in periods long enough Corsi pretty much tells you also those results.

But. And this is really a huge BUT. For example Laine is not average. He is not average shooter. He doesn't have average IQ. He is picky with his shots etc. etc. Therefore even if his Corsi over long time is pretty average, somehow, magically his 5-on-5 goal differential is best in the team (shared with Scheifele, who is another non-average-Corsi guy).

Therefore all those claims how this and that guy is better than the other guy since he has better Corsi should be taken with a shovel full of salt. And you don't really claim that you haven't seen those flying HfJets & HfBoards all the time?

Does that mean that Corsi is useless? No, it doesn't. Corsi events lack completely the quality factor of eg. goals, but the quantity is always magnitude or more higher than that of goals, so that Corsi is more predictable and reliable as an indicator for trends than goals. Like not everyone scores 0.50 GPG like Laine, but most (goal scorers at least) will have several Corsi events per game. So if someone all in the sudden has a significant drop of his Corsi production for several games, you might need to worry as there might be something wrong going on with his game, and you can probably adjust things tens of games earlier than you could just by looking goals scored, as they are much more rare events.

PS, I agree with you that Corsi is not arbitrary model. The "only" problem with it, and that is quite huge, is that it lacks the quality factor completely. As the goal of the game is to score more goals than the opponent, not to have more Corsi events than the opponent or even outshoot the opponent.

 

[Tom ServoMST3K]

I never understood the hate towards Corsi from the old school crowd. I remember on so many occasions in my youth Don Cherry yelling something like: "ALWAYS SHOOT YA NEVER KNOW WHAT COULD HAPPEN!!!"

Basically corsi measures that kind of thing and the old school types hate it?

 

[garret9]

Some people are just so fancy that they like fancy stats more than the actual result. Corsi etc are just MODELS someone has come up with to predict an outcome, of course the actual outcome should be more important (and it is for most people). Weather is hard to predict so there’s weather forecasts built on different models, but I’ve yet to meet a single person who would hold on to a forecast if it said it should be sunny when it’s raining outside, and claim that the model was right while the weather got it wrong. It never ceases to amaze me.

There's a lot wrong here. If you want I can help you...

Corsi isn’t a model. It’s just a shot different. It’s a specific measure of how much a team is outshooting it’s opponents. It's a specific explanation of who is outshooting who. No one came up with the model to predict anything.

Corsi matters because the average player who outscores but doesn’t out Corsi will tend to not outscore in the future. Also the average player who does not outscore but does out Corsi will tend I outscore in the future. This is simply a fact and it was a fact that was found out simply when researching such things.

In other words: when a player (or team) performs differently based on goals vs shots, shots is right much more often.

Weather isn't at all an apt comparison anyways. Hockey has two groups battling it out with specific and opposed objectives: outscore the opposition. There is a pattern to hockey because the game is *not* random, and where there is patterns there are statistical trends which shows you the better team that is more likely to win in the long run or the better player that is more likely to outscore in the long run.

One just has to remember that Corsi is a portion of the game. Offense is about generating chances.

• You want as many as possible (shot volume) and want the ones you get to be as good as possible (shot quality) and you want to score on as many of those as possible (finishing talent).
• You want to prevent as many as possible (shot volume) and want the ones you allow to be as poor as possible (shot quality) and you want to stop as many of those as possible (goaltending).

Corsi is just 2/6 of those factors (both the shot quality for and against). It's not a model of shot quantity but an actual counting measure of it.

 

[Tommigun]

Corsi isn’t a model. It’s just a shot different. It’s a specific measure of how much a team is outshooting it’s opponents.

Corsi matters because the average player who outscores but doesn’t out Corsi will tend to not outscore in the future. Also the average player who does not outscore but does out Corsi will tend I outscore in the future.

In other words: when a player (or team) performs differently based on goals vs shots, shots is right much more often.

That is all.

It’s a model. It’s a model that tries to predict the future, just as you said. The fact that it calculates shot differential is an implementation detail. It’s a very simple model but a model nevertheless.

Statistics are not a model. That’s the realized end result that the Corsi model is trying to predict.

 

[Tommigun]

When I said it was arbitrary I meant that Jim Corsi just happened to decide that for example shot quality doesn’t matter. You could calculate any action and call it a model, Corsi just happens to calculate shot volume, and shot volume only. It’s not the end all be all model (or “metric” if you focus on the value itself without trying to extrapolate goals and success out of it). It’s not bad either, but it’s not perfect in any way as it favors certain play styles and skill sets.

 

[garret9]

Shot quantity isn't the whole picture, but it's a BIG part of the picture. If you look only at it, you miss some of the picture... but if you ignore it you miss an even greater part of the picture:

 

[garret9]

When I said it was arbitrary I meant that Jim Corsi just happened to decide that for example shot quality doesn’t matter. You could calculate any action and call it a model, Corsi just happens to calculate shot volume, and shot volume only. It’s not the end all be all model (or “metric” if you focus on the value itself without trying to extrapolate goals and success out of it). It’s not bad either, but it’s not perfect in any way as it favors certain play styles and skill sets.

Jim Corsi didn't decide that shot quality didn't matter. He was using all shot types to measure a goaltender's workload (how often they reacted to shots).

Tim Barnes was the one that studied shot volume, noticed it outperformed goal differentials in predicting future success, and named it Corsi.

Ironically Barnes didn't know it was Jim Corsi who decided to use all shot types, but just heard a member of the Sabres management in an interview saying they do that, and he named it after Corsi because of his moustache.

 

[Tommigun]

Jim Corsi didn't decide that shot quality didn't matter. He was using all shot types to measure a goaltender's workload (how often they reacted to shots).

Tim Barnes was the one that studied shot volume, noticed it outperformed goal differentials in predicting future success, and named it Corsi.

Ironically Barnes didn't know it was Jim Corsi who decided to use all shot types, but just heard a member of the Sabres management in an interview saying they do that, and he named it after Corsi because of his moustache.

Thanks, great info. But Corsi most definitely did decide that shot quality would not be a factor in his formula as there is no weighing so all shots are treated as equal, no matter if it’s a stray dump in from a grinder or a shot from the slot by Ovechkin. Every model is made up by someone and that person decides what variables it contains. Just saying that it’s a pretty simple model for predicting success as it heavily favors for example volume shooters, and puts 100% of the focus on quantity vs quality. Wouldn’t zone time be a better predictor, as it wouldn’t judge exclusively on how trigger happy someone is?

 

[garret9]

Thanks, great info. But Corsi most definitely did decide that shot quality would not be a factor in his formula as there is no weighing so all shots are treated as equal, no matter if it’s a stray dump in from a grinder or a shot from the slot by Ovechkin. Every model is made up by someone and that person decides what variables it contains. Just saying that it’s a pretty simple model as it heavily favors for example volume shooters, and puts 100% of the focus on quantity vs quality.

Well ya. Of course shot quality would not factor when trying to estimate how often a goalie has to react to a shot.

As to shots not being created equal, that doesn't necessarily make it a bad thing.

You want to be able to break down the game into its individual components sometimes. A team or player that is outscoring but failing in the shot quantity portion of the game has an area they are failing in. As we've noted, those players and teams on average fall more often than not.

As to the style thing, don't forget what is explanative is not necessarily predictive.
Example: Adding shot location information to Corsi for an xGoal model increases in sample correlation to goals but decreases out of sample correlation to goals. In other words, Corsi without shot location is more predictive than shot location with prediction. My theory has been that this is because players control shot quantity by a far greater degree than they control shot quality, and the distribution in talent is larger for the former than the latter.

There's also a danger to the style thing. People fall for this all the time. Every single time a team or player is "beating Corsi" we see explanations to why being due to their style, but far more often than not, we see these players and teams collapse.

 

[Tommigun]

Well ya. Of course shot quality would not factor when trying to estimate how often a goalie has to react to a shot.

As to shots not being created equal, that doesn't necessarily make it a bad thing.

You want to be able to break down the game into its individual components sometimes. A team or player that is outscoring but failing in the shot quantity portion of the game has an area they are failing in. As we've noted, those players and teams on average fall more often than not.

As to the style thing, don't forget what is explanative is not necessarily predictive.
Example: Adding shot location information to Corsi for an xGoal model increases in sample correlation to goals but decreases out of sample correlation to goals. In other words, Corsi without shot location is more predictive than shot location with prediction. My theory has been that this is because players control shot quantity by a far greater degree than they control shot quality, and the distribution in talent is larger for the former than the latter.

There's also a danger to the style thing. People fall for this all the time. Every single time a team or player is "beating Corsi" we see explanations to why being due to their style, but far more often than not, we see these players and teams collapse.

I edited my post a little for clarity but you’re so fast that you had already quoted it

But still, Corsi only accounts for quantity and not at all for quality. Goals are pretty much quantity x quality x luck I guess. Luck has to be omitted, but why quality also? So it in no way paints a complete picture. Wouldn’t zone time be a better indicator for success as it wouldn’t just base the results entirely on how trigger happy someone is?

Would you calculate salary for a year by 12 x 15000, or 12 x 1, nulling out the actual salary? Because that’s what Corsi does, it will just tell you how often you’d receive your salary but not necessarily much more.

 

[Jack722]

And I agree with your estimation about Perreault as a player. He is exactly how you described. But his kind is not a very useful player unless he is exactly the great kind of a supporting puck hound with above average skill that is an extremely good player to support the kind of a duo that Scheifele and Laine are when they play together.

But with less efficient players like Little and even Ehlers the line becomes more like some busy bees without doing too well the things that still win the games, which are scoring clearly more goals than their opponents score. Thus the Corsi that he brings is pretty much useless by itself, unless he plays with the kind of players that are able to put the puck in the net. This is exactly why Corsi by itself is just next to useless. Some people just look at the Corsi numbers so blindly without understanding how much different kind of skills and efficiency between different players just simply make those shot statistics next to useless when comparing exactly individuals. With teams I can see a bit more of use with those kind of statistics, but by comparing individuals and their importance to their team, Corsi and Corsi-based stats are simply bad and misleading stats.

Not sure I agree that PLE is bad per se but I definitely agree with your implication that in order to use both Laine and Perreault's talents to the fullest, P-S-L seems well worth trying. I think we tried it before but I can't remember the results. In theory it should be dynamite.

 

[garret9]

I edited my post a little for clarity but you’re so fast that you had already quoted it

But still, Corsi only accounts for quantity and not at all for quality. Goals are pretty much quantity x quality x luck I guess. Luck has to be omitted, but why quality also? So it in no way paints a complete picture. Wouldn’t zone time be a better indicator for success as it wouldn’t just base the results entirely on how trigger happy someone is?

Would you calculate salary for a year by 12 x 15000, or 12 x 1, nulling out the actual salary? Because that’s what Corsi does, it will just tell you how often you’d receive your salary but not necessarily much more.

As I noted above, it's actually:
GF (quantity, quality, finishing, luck)
GA (quantity, quality, goaltending, luck)

One thing you have to keep in mind is that there is "luck" in the opportunity. Luck is not just bounces. Part of what luck is can be that a player or team did better in shot quality than what we expect them to keep doing moving forward. This is why xGoals, which is just Corsi + shot location*, actually performs worse than just Corsi in predicting the future.

That's a weird example, as that's very apples and oranges.

1. There's a much, much, much stronger relationship in those that do well in shot quantity and those that do well in shot quality than there is years and AAV of contracts.
2. Part of what makes shot quantity so powerful in predicting which players, lines, and teams are best is because players control shot quantity by such a greater degree than the other factors and there is such a greater spread in talent distribution than there is in shot quality and finishing talent.**

* excluding DTMAH's no longer public xGoal that used regression adjustments to finishing talent and other factors
** doesn't mean there is no spread, which is why the other factors still matter

 

[garret9]

All this arguing about Corsi for Perreault-Little-Ehlers, and they didn't just do well in Corsi...

Corsi (All shots): 64% - 287 vs 164
Fenwick (Unblocked shots): 64% - 119 vs 64
Shots on net: 62% - 146 vs 91
Actual goals: 67% - 10 vs 5
Scoring chances: 63% - 126 vs 75
High Danger Corsi: 68% - 53 vs 25
High Danger goals: 67% - 6 vs 3

Unweighted average of these things: 65%

So looks like Corsi undervalued Perreault.

*High Danger and Scoring Chances by NST's definition*

 

[lomiller1]

Weather isn't at all an apt comparison anyways. Hockey has two groups battling it out with specific and opposed objectives: outscore the opposition. There is a pattern to hockey because the game is *not* random, and where there is patterns there are statistical trends which shows you the better team that is more likely to win in the long run or the better player that is more likely to outscore in the long run.

Climate vs weather. Past a week or so weather is nearly impossible to predict, but the average follows strong underlying patterns. Even while getting almost none of the daily predictions right the average tracks these underlying patterns closely which allows you to make very good predictions over longer periods (typically 10-30 years)

Hockey is similar. Smaller samples are dominated by luck but as you look at larger sample sized the underlying patterns emerge. Trying to predict an individual hockey game is like trying to predict the weather a week out. There is a good chance you will be wrong even with a model that provides good predictions, that doesn't mean there are not larger patterns to be discovered.

A further similarity is that for both you get a group of people utterly convinced the lager patterns show by the model are wrong because reasons.

 

[garret9]

Climate vs weather. Past a week or so weather is nearly impossible to predict, but the average follows strong underlying patterns. Even while getting almost none of the daily predictions right the average tracks these underlying patterns closely which allows you to make very good predictions over longer periods (typically 10-30 years)

Hockey is similar. Smaller samples are dominated by luck but as you look at larger sample sized the underlying patterns emerge. Trying to predict an individual hockey game is like trying to predict the weather a week out. There is a good chance you will be wrong even with a model that provides good predictions, that doesn't mean there are not larger patterns to be discovered.

A further similarity is that for both you get a group of people utterly convinced the lager patterns show by the model are wrong because reasons.

I don’t know much about climate or weather, obviously, hahah!

I do know weather doesn’t have a known objective though, which is one difference.