Lane Hutson Burgeoning Star Watch

[ponder]

• Quality of teammates
• Quality of opposition
• Are you getting put out for mostly o-zone work (o-zone faceoffs, or hopping over the boards on the fly when your team has possession) or d-zone work (the opposite)?

All of this information can also be found and used (although QoC and zone starts are vastly overrated in terms of impact outside of the outliers).

Eg, you can see on MTL, the Guhle pairing plays ridiculous (hard) minutes, the Xhekaj pairing plays ridiculous (easy) minutes.

• Simply knowing "slap shot from the right dot" isn't actually very good at predicting the danger of a shot. None of this sort of info is available:
  • Was the goalie moving, or set?
  • Was the goalie screened?
  • Did the player have the time/space to get off a quality shot, hard and with great placement, or were they heavily pressured and could only get off a weak shot with bad placement?

Better models use this information to varying extents, but yes, it is a developing stat.

Also, players who consistently generate chances in ways that models do not account for will outperform those models over a large sample size (which can be easily recognized if you again look). It's why most WAR models use actual goals for offense, and expected goals for defense.

If you are expecting it to be a perfect stat, you're crazy.

Ignoring it's value is also crazy.

Agreed that there are other stats using quality of teammates and opposition, and zone starts - e.g. most of the stats with "adjusted" somewhere in the name. But xGF%, which was being discussed, does not. And also, these "quality of teammates and opposition" bits are ALSO based on the same NHL event data, which is just super limited, so it's a somewhat circular definition. If these stats have serious errors/caveats about your own quality, they have those same caveats/errors about the quality of your teammates and opposition.

• Simply knowing "slap shot from the right dot" isn't actually very good at predicting the danger of a shot. None of this sort of info is available:
  • Was the goalie moving, or set?
  • Was the goalie screened?
  • Did the player have the time/space to get off a quality shot, hard and with great placement, or were they heavily pressured and could only get off a weak shot with bad placement?

Better models use this information to varying extents, but yes, it is a developing stat.

AFAIK this data is simply not available. For all advanced stats available to the general public (that I've seen), the source data is the official NHL stats/event log, which I gave a sample event of in my previous post, and all of this information is simply not there. Some stats try to infer it via things like "if the shot came 2 seconds since the last event, that's more likely to be a one-timer than a shot 5 seconds since the last event," but this is stretching things massively and a very low quality analysis IMO.

I don't think xGF% is completely useless, but I do think the error/noise within it is mostly larger than the signal, when it comes to trying to determine a player's overall impact.

For example, according to MoneyPuck's xGF% model, last year Mark Jankowski had the highest xGF% of any non-Edmonton player (with 300+ 5v5 minutes) - I say non-Edmonton, because Edmonton players absolutely dominated xGF% last year (again indicating it's more of a team than individual stat). Or if you want to use a higher minutes cutoff (say 1000 5v5 minutes), the highest xGF% of any non-Edmonton player was Jordan Staal. Or on the flip side, guys in the bottom 50 for xGF% (minimum 1000 5v5 mins) include Kyle Connor, Mark Scheifele, Alex Ovechkin and Connor Bedard. You could take this literally, and say "Mark Jankowski is a massively better player than Mark Scheifele", but IMO a better interpretation is that the stat just isn't very good at describing a player's value on its own, and Mark Jankowski is nowhere close to Mark Scheifele as a player.

 

[Baksfamous112]

You have this view of the "team" as some magical entity holding players back rather than a collection of individual players.

Yes, when players play better, such as hutson having an excellent game, the team will do better.

You could apply that logic you just did for hutson to every single other player on the habs, blaming their past failures on the team around them and justifying them as a good player, into circular logic that tries to prop up players on bad teams and bring down players on good teams.

You could go down every single player from the 1st line to the 4th line and from the 1st pairing to the 3rd pairing blaming "the team", and saying the player on the shit team is better than the player on the good team.

So what happens when Player A play a great (individual) game and generate as much as a single entity possibly can but that player is also playing with 4 other teammates that just doesn’t have it or are having a bad collective stretch?

The difference between a win and a loss makes basically every single player xGF% go up 20%. You can't say that suddenly every single players got so much better.

Yes each individual performances helps a team, but it goes the other way around. A great team will make everyone's individual performance much better. You cannot deny that.

but he will.

Basically, it's a stat to take with a massive, massive grain of salt. It's very common to see players that are clearly not very good (coaches don't play them much, little value in trades/signings, don't look good on the eye test) who have great xGF%, because they're getting really soft/sheltered minutes on a strong team. Or, also very common to see players that are clearly very good (coaches play them heavily, lots of trade/signing value, look great on the eye test) have terrible xGF%, because they're getting really tough minutes on a bad team. And that's over the course of entire seasons - over the course of a single game, it says even less about the player.

FWIW, I've got an MSc in a stats heavy field, and used to be a Data Scientist (am now a Software Engineer), so I've got pretty strong experience in stats and data analysis, and I personally put minimal value in xGF%. It's one statistic about a player, but to make sense of it you need to fully understand the (massive) weaknesses it has, and fully understand the context around the player's team and role within their team. Acting like it's a single number that meaningfully describes a player's value is IMO crazy - even over the course of a full season, and especially so for a single game.

Thank you for the explanation. That’s what I’ve been trying to articulate for the past 2 months but some people just doesn’t have the capacity to understand this.

 

[dgibb10]

Agreed that there are other stats using quality of teammates and opposition, and zone starts - e.g. most of the stats with "adjusted" somewhere in the name. But xGF%, which was being discussed, does not. But also, these "quality of teammates and opposition" bits are ALSO based on the same NHL event data, which is just super limited, so it's a somewhat circular definition.


AFAIK this data is simply not available. For all advanced stats available to the general public (that I've seen), the source data is the official NHL stats/event log, which I gave a sample event of in my previous post, and all of this information is simply not there. Some stats try to infer it via things like "if the shot came 2 seconds since the last event, that's more likely to be a one-timer than a shot 5 seconds since the last event," but this is stretching things massively and a very low quality analysis.

I don't think xGF% is completely useless, but I do think the error/noise within it is mostly larger than the signal, when it comes to trying to determine a player's overall impact. For example, last year Mark Jankowski had the highest xGF% of any non-Edmonton player (with 300+ 5v5 minutes). Or if you want to use a higher minutes cutoff (say 1000 5v5 minutes), it's Jordan Staal. Or on the flip side, guys in the bottom 50 for xGF% (minimum 1000 5v5 mins) include Kyle Connor, Mark Scheifele, Alex Ovechkin and Connor Bedard. You could take this literally, and say "Mark Jankowski is a massively better player than Mark Scheifele", but IMO a better interpretation is that the stat just isn't very good at describing a player's value on its own, and Mark Jankowski is nowhere close to Mark Scheifele as a player.

Jordan Staal is an absolute stud.

I'm confused what your issue is there? Because Staal plays hard minutes, with primarly dzone starts. And he dominates them. He has for years. That 3rd line in carolina has been a massive part of their team success.

I will agree that Staal isn't a great finisher, which is a seperate metric, but he is one of the best defensive players in hockey


Schiefele, Ovi, Bedard, Connor.

All of those players are BAD defensively. Is that the issue? that xGoals takes into account defensive play and how much you give up?

And they are snipers.

Yes, expected goals do not account for shooting talent, but it is very easy to account for that by just looking at a players career. But again, these elite shooters are the outliers.

Using outliers as the means to disregard a stat is dumb.

your problem with xGoals is starting to become clear to me. And it's that it doesn't line up with players point totals, which is your baseline for evaluation.

No, expected goals is not a metric to evaluate finishing ability. It's whole point is to look at play driving and chance generation/prevention.

Onto Jankowski. His actual goal metrics 71% were even BETTER than his expected goal metrics. So what was the tracking missing, what usage/team play is skewing his numbers? With Jankowski on the ice, Nashville dominated last year. His box score stats were quality too, considering he scored at a virtually identical or better rate than all the other players you mentioned.

Btw all the players you mentioned as being underrated by xGoals, had ridiculously offensively skewed usage that you say is a big reason why players are OVERRATED by xgoals, and all play with 1st line linemates. Your logic doesn't line up.

This seems to a sample size issue rather than a stat issue. Jankowski was excellent last year, which happens sometimes.

 

[dgibb10]

So what happens when Player A play a great (individual) game and generate as much as a single entity possibly can but that player is also playing with 4 other teammates that just doesn’t have it or are having a bad collective stretch?

but he will.

Thank you for the explanation. That’s what I’ve been trying to articulate for the past 2 months but some people just doesn’t have the capacity to understand this.

Everything is everyone else's fault but Hutson's.

It's very clear that is your view.

You start with the assumption that hutson is playing excellent, and any negative is someone else's fault, and every positive is because of him.

 

[the paisanos guy]

Everything is everyone else's fault but Hutson's.

It's very clear that is your view.

You start with the assumption that hutson is playing excellent, and any negative is someone else's fault, and every positive is because of him.

Maybe you're starting with the reverse assumption? to the point of stat-hunting for proof instead of just watching with your eyes the impact he has on the ice.

I mean you sure seem invested in him

 

[dgibb10]

Maybe you're starting with the reverse assumption? to the point of stat-hunting for proof instead of just watching with your eyes the impact he has on the ice.

I use the same metrics quite consistently across the board.

I acknowledge these metrics have issues with:

Overrating players with ridiculously sheltered usage (Xhekaj)
Underrating players with ridiculously hard usage (Guhle)
Overrating guys who are really bad at shooting the puck
Underrating guys who are really good at shooting the puck.
Underrating guys who generate the types of chances that models struggle to account for.

Hutson doesn't fall into either of the usage categories, his usage is normal. He's certainly not a sniper, if anything his shot is poor.

Hutson is on the ice for 2.31 xGoals/60 5v5, and 2.26 actual goals/60 5v5. This does not suggest the models are underrating his chance quality at all.

Meanwhile MTL fans try to walk a tightrope of blaming david savard for every single thing that goes wrong, while simultaneously trying to convince people he is worth a 1st.

 

[MMC]

Zellweger clears this 4th forward