All Purpose Analytics and Extended Stats Discussion

[Hivemind]

He’s leaving baseball to take a role with Fenway Sports Group as of October 2020. I’d venture this is his last season. Article I read said he most likely shifts his focus to European soccer

Either way, he certainly wasn’t forced out of baseball for being “too focused on stats.”

 

[kicksavedave]

Either way, he certainly wasn’t forced out of baseball for being “too focused on stats.”

This, plus the fact that he decided to stick with Oakland and their stingy owner vs taking more high profile jobs is what cost him wins, especially as the mega teams started to dominate when they began using stats as well. He started a revolution, pure and simple and he was copied through out baseball and other sports, so his advantage ended, but he could have moved on to a rich team and won a ton more.

 

[kicksavedave]

Bumping this thread so a few forums regulars can give it a quick read.

One comment about analytics and the eye test. It was said in the Rags thread that Lavi has access to all the same stats and numbers as we do (probably even more that aren't published on the web). We've heard BMac state that he uses analytics some but he isn't ruled by them, as in he doesn't use them exclusively to make his decisions. We haven't heard (or at least I haven't) from Lavi about how much does he use analytics in his coaching decisions. We do know that some coaches (and GMs) use them more than others do, likely from a range of almost none to too much so.

But when we say "Lavi has access to all the same data" and his decisions don't coincide with your position" its a little disingenuous to suggest Lavi sees the stats differently. We simply don't know how much he looks at the fancies, how much he bases it on his decisions or not. He could be an "eye test" only coach and its hard for any of us to really know, but who could blame him? He's taken three teams to Cup finals, he's been a successful NHL HC for almost 20 years, his "eye test" is much better than our eye tests, despite 20 years of us posting on message boards. So lets not presume to know exactly how he makes his decisions on ice time and deployments. When his decisions agree with ours, we assume he uses the same information we use in the same ratios, but we simply don't know that for a fact.

Lavi may be very much fancy stat averse, as some coaches (or GMs) tend to be. The assumptions that he's all knowing, or even infallible, merely to support our arguments, doesn't float.

TL;DR Maybe if he played Kuzy more late in games we wouldn't be losing so many big leads like we have been? What he's doing instead, ain't working.

 

[Calicaps]

One comment about analytics and the eye test. It was said in the Rags thread that Lavi has access to all the same stats and numbers as we do (probably even more that aren't published on the web). We've heard BMac state that he uses analytics some but he isn't ruled by them, as in he doesn't use them exclusively to make his decisions. We haven't heard (or at least I haven't) from Lavi about how much does he use analytics in his coaching decisions. We do know that some coaches (and GMs) use them more than others do, likely from a range of almost none to too much so.

But when we say "Lavi has access to all the same data" and his decisions don't coincide with your position" its a little disingenuous to suggest Lavi sees the stats differently. We simply don't know how much he looks at the fancies, how much he bases it on his decisions or not. He could be an "eye test" only coach and its hard for any of us to really know, but who could blame him? He's taken three teams to Cup finals, he's been a successful NHL HC for almost 20 years, his "eye test" is much better than our eye tests, despite 20 years of us posting on message boards. So lets not presume to know exactly how he makes his decisions on ice time and deployments. When his decisions agree with ours, we assume he uses the same information we use in the same ratios, but we simply don't know that for a fact.

Lavi may be very much fancy stat averse, as some coaches (or GMs) tend to be. The assumptions that he's all knowing, or even infallible, merely to support our arguments, doesn't float.

TL;DR Maybe if he played Kuzy more late in games we wouldn't be losing so many big leads like we have been? What he's doing instead, ain't working.

I said that. And I don't assume he's using my metrics/eye test. My point was, there's likely very little info that we have that he doesn't and it's more likely... indeed almost certain... that he has a lot of info we don't, so to assume that he isn't factoring in various metrics is naïve. There's the notion among the stat fans that the fancies present an obvious choice, and I would counter that in the full light of the knowledge that someone inside the dressing room and with decades of experience would have, what we think is obvious, simply isn't. And maybe what he's doing isn't working... or maybe it just isn't working yet. I don't know, but I'm not arrogant enough to claim any certainty.

 

[g00n]

One comment about analytics and the eye test. It was said in the Rags thread that Lavi has access to all the same stats and numbers as we do (probably even more that aren't published on the web). We've heard BMac state that he uses analytics some but he isn't ruled by them, as in he doesn't use them exclusively to make his decisions. We haven't heard (or at least I haven't) from Lavi about how much does he use analytics in his coaching decisions. We do know that some coaches (and GMs) use them more than others do, likely from a range of almost none to too much so.

But when we say "Lavi has access to all the same data" and his decisions don't coincide with your position" its a little disingenuous to suggest Lavi sees the stats differently. We simply don't know how much he looks at the fancies, how much he bases it on his decisions or not. He could be an "eye test" only coach and its hard for any of us to really know, but who could blame him? He's taken three teams to Cup finals, he's been a successful NHL HC for almost 20 years, his "eye test" is much better than our eye tests, despite 20 years of us posting on message boards. So lets not presume to know exactly how he makes his decisions on ice time and deployments. When his decisions agree with ours, we assume he uses the same information we use in the same ratios, but we simply don't know that for a fact.

Lavi may be very much fancy stat averse, as some coaches (or GMs) tend to be. The assumptions that he's all knowing, or even infallible, merely to support our arguments, doesn't float.

TL;DR Maybe if he played Kuzy more late in games we wouldn't be losing so many big leads like we have been? What he's doing instead, ain't working.

Why don't we get some facts, then?

Peter Laviolette’s Use Of Advanced Analytics: Skeptic, Believer, Or In-Betweener?

Continue reading past that part for more.

So yeah. Lavi is no luddite or babe in the woods here. It's arrogant and presumptuous to think NHL coaches and franchises today don't have either a very robust analytics information bank, or at least one person on staff providing some sort of access.

The rest of the article is a mixed bag of opinions from Preds fans/bloggers about how much analytics were evident in Lavi's moves. I would expect most fancy stats enthusiasts would have similar opinions for any team that isn't 100% Twabbified.

 

[g00n]

Before anyone gets upset that we're talking about this, I'm also going to repost the local rule and OP of this thread, since there was some confusion and my reply was deleted yesterday in what looks like a cleanup of QDP...

This is not and has never been a "safe space" exclusively for discussing analytics entirely among true-believers. Nor are analytics banned from other threads. This is simply a deep-dive thread that's also the landing spot for repetitious or obnoxious-to-the-point-of-threadjacking debates about analytics.

The rule, with emphasis added as related to this point:


MOD NOTE: We recognize that not everyone is into deep statistical analysis. This is a thread for "extended stats" discussions. Stats can still be used in the general roster discussion to make a single point, but if someone wants to debate a player's statistical value ad nauseum*, or engage in back and forth about the value of X or Y stat, it must go here. If you have extended analysis that applies in another thread, rather than jam up that thread with large charts and so forth you may want to link to a post here, or use a collapse tag in your original post. Contact a mod if you do not know how to do this.

Mods please delete if not appropriate.

*and feel free to correct my flub in spelling "ad nauseam"

 

[twabby]

I said that. And I don't assume he's using my metrics/eye test. My point was, there's likely very little info that we have that he doesn't and it's more likely... indeed almost certain... that he has a lot of info we don't, so to assume that he isn't factoring in various metrics is naïve. There's the notion among the stat fans that the fancies present an obvious choice, and I would counter that in the full light of the knowledge that someone inside the dressing room and with decades of experience would have, what we think is obvious, simply isn't. And maybe what he's doing isn't working... or maybe it just isn't working yet. I don't know, but I'm not arrogant enough to claim any certainty.

So if we follow this train of logic, how can an outsider ever levy a criticism against a coach? They will always have access to more resources than us.

I'm not saying with 100% certainty he is wrong in his management of late game situations with the lead, I'm saying there is some evidence to suggest that he isn't doing a great job in this regard and there is some evidence that a specific alternative strategy may yield better results.

 

[g00n]

So if we follow this train of logic, how can an outsider ever levy a criticism against a coach? They will always have access to more resources than us.

I'm not saying with 100% certainty he is wrong in his management of late game situations with the lead, I'm saying there is some evidence to suggest that he isn't doing a great job in this regard and there is some evidence that a specific alternative strategy may yield better results.

There's a big difference between admittedly semi-informed fans forwarding opinions about coaching decisions and obviously observable gameplay and quotes...

vs.

...outsiders insisting they know for certain (or near certain) that debatable statistical analysis alone decides the issues regardless of any unknown mitigating factors that only a coach or player would know, and that the team insiders are somehow completely unaware of the now-commonly-available analytics, or else are refusing to act on some clear-cut statistical conclusion they undoubtedly would have at their fingertips, probably before we do.


I mean how is that not insanely obvious?

 

[kicksavedave]

There's a big difference between admittedly semi-informed fans forwarding opinions about coaching decisions and obviously observable gameplay and quotes...

vs.

...outsiders insisting they know for certain (or near certain) that debatable statistical analysis alone decides the issues regardless of any unknown mitigating factors that only a coach or player would know, and that the team insiders are somehow completely unaware of the now-commonly-available analytics, or else are refusing to act on some clear-cut statistical conclusion they undoubtedly would have at their fingertips, probably before we do.


I mean how is that not insanely obvious?

Who has said anything like that?

I see Twabby suggesting, even in the post immediately above this, that there is evidence that supports other decisions. That's not saying he "knows for certain" that the coach is making mistakes.

 

[Calicaps]

So if we follow this train of logic, how can an outsider ever levy a criticism against a coach? They will always have access to more resources than us.

I'm not saying with 100% certainty he is wrong in his management of late game situations with the lead, I'm saying there is some evidence to suggest that he isn't doing a great job in this regard and there is some evidence that a specific alternative strategy may yield better results.

You have said repeatedly that he should be doing X or Y based on xGwhatever... . One can question coaching decisions for sure and suggest why alternatives may be better but the idea that one data source always governs ignores the fact that it's only one of many.

There's a big difference between admittedly semi-informed fans forwarding opinions about coaching decisions and obviously observable gameplay and quotes...

vs.

...outsiders insisting they know for certain (or near certain) that debatable statistical analysis alone decides the issues regardless of any unknown mitigating factors that only a coach or player would know, and that the team insiders are somehow completely unaware of the now-commonly-available analytics, or else are refusing to act on some clear-cut statistical conclusion they undoubtedly would have at their fingertips, probably before we do.


I mean how is that not insanely obvious?

this.

Who has said anything like that?

I see Twabby suggesting, even in the post immediately above this, that there is evidence that supports other decisions. That's not saying he "knows for certain" that the coach is making mistakes.

I find the "this stat means this is right/wrong" posts pretty certain. And that is bolstered by the "told you so" posts that presume his preferred metric is the reason a certain decision was made or a certain outcome reached. Which is a HUGE presumption.

 

[twabby]

You have said repeatedly that he should be doing X or Y based on xGwhatever... . One can question coaching decisions for sure and suggest why alternatives may be better but the idea that one data source always governs ignores the fact that it's only one of many.

I find the "this stat means this right/wrong" posts pretty certain. And they are bolstered by the "told you so" posts that presume his preferred metric is the reason a certain decision was made or a certain outcome reached. Which is a HUGE presumption.

I specifically went through some of the predictive power of expected goals, and I specifically mentioned that it is not perfectly predictive or even close to it. The bolded sentence flies in the face of things that I have specifically posted, and it makes me think that you aren't actually even trying to read or understand analytics or their application. Which is fine, you don't have to if you don't want to.

But please don't mischaracterize things like the bolded if that's the case.

For reference:

Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2
Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2

I asserted that it [xGF] has a correlation of about r^2 = 0.23 to future GF, nothing more. A noticeable, but not incredibly strong predictor.

 

[g00n]

Who has said anything like that?

I see Twabby suggesting, even in the post immediately above this, that there is evidence that supports other decisions. That's not saying he "knows for certain" that the coach is making mistakes.

What Cali said.

And it has definitely been said that stats trump the eye test and in many cases can be relied upon themselves for a reasonable level of certainty, at least beyond eye test or other stats. These statements are often couched when convenient and forgotten when the opportunity to gloat presents itself, however premature or exaggerated.

Point being, Twabby is abusing logic in trying to twist Cali's argument to mean we can never criticize coaches unless we accept his analytics-based statements, and that's just not true.

 

[Calicaps]

I specifically went through some of the predictive power of expected goals, and I specifically mentioned that it is not perfectly predictive or even close to it. The bolded sentence flies in the face of things that I have specifically posted, and it makes me think that you aren't actually even trying to read or understand analytics or their application. Which is fine, you don't have to if you don't want to.

But please don't mischaracterize things like the bolded if that's the case.

For reference:

Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2
Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2

If it's "not close" to predictive then WTF do you talk about it so much for? Seriously? And why do you then rush to say "who could've predicted that" every time?

You know, what? Nevermind. I don't actually care.

 

[twabby]

If it's "not close" to predictive then WTF do you talk about it so much for? Seriously? And why do you then rush to say "who could've predicted that" every time?

You know, what? Nevermind. I don't actually care.

It's not close to perfectly predictive. It's right there. The word "perfectly".

I don't think you're engaging in good faith right now.

 

[g00n]

I specifically went through some of the predictive power of expected goals, and I specifically mentioned that it is not perfectly predictive or even close to it. The bolded sentence flies in the face of things that I have specifically posted, and it makes me think that you aren't actually even trying to read or understand analytics or their application. Which is fine, you don't have to if you don't want to.

But please don't mischaracterize things like the bolded if that's the case.

For reference:

Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2
Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2

I notice you didn't post my reply.

Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2

In other posts you were claiming a stat with 23% predictability was "more likely than not" going to be right. Which is not how predictions or probability work.

Nobody has time to search thousands and thousands of posts then argue semantics about what's a declaration of fact vs some squishy statistical hedge.

 

[twabby]

I notice you didn't post my reply.

Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2

In other posts you were claiming a stat with 23% predictability was "more likely than not" going to be right. Which is not how predictions or probability work.

Nobody has time to search thousands and thousands of posts then argue semantics about what's a declaration of fact vs some squishy statistical hedge.

I don't think you understand how correlation works. An r^2 of 0.23 doesn't mean it's worse than a coin flip. A coin-flip has an r^2 of 0 because there is no correlation between one flip on the next.

 

[Calicaps]

It's not close to perfectly predictive. It's right there. The word "perfectly".

I don't think you're engaging in good faith right now.

Oh Christ! It's not bad faith. I'm just not engaging in full because I really don't care. I'm going back to the hockey threads.

 

[kicksavedave]

I find the "this stat means this is right/wrong" posts pretty certain. And that is bolstered by the "told you so" posts that presume his preferred metric is the reason a certain decision was made or a certain outcome reached. Which is a HUGE presumption.

I just disagree. I find the "stats support this opinion" or "stats helped me form this opinion" 1000X more compelling than the occasional "Player X is an ECHL'r" junk we have to endure (not from you but you know what I am referring to). I just don't see Twabby saying "I'm right and Lavi is wrong" any more or less than dozens of other posters here who form their opinions in all sorts of different ways. You all are singling him (her?) out for using stats to support his opinions, but misrepresenting how he characterizes said opinions.

I think our collective "eye tests" are much more susceptible to bias, both recency and personal, than looking at statistics, but neither one tells a complete picture without the other. This forum, like all forums, is littered with people who think they know more than the coaches. Most of them use no evidence whatsoever to support their takes, Twabby provides evidence for his takes all the time. I personally hope he continues posting these analytics because it truly helps form a more complete picture, even if we all draw our own different conclusions in our own different ways.

Also, Lavi has been everything we hoped for and he seems to be getting virtually everything right so far, but it would be pretty boring if we all just sat around thinking there's no room for improvement. In the end we're really spoiled, bitching at each other about how to fix our 18-2 3rd period lead issues on a team tied for points lead in the NHL. But then again, Reirden had solid RS records too, but the flaws we saw creeping in crushed us in the playoffs. So, we're Caps fans after all. Even in 1st place, we're going to complain about something. Its what we do.

 

[Calicaps]

I just disagree. I find the "stats support this opinion" or "stats helped me form this opinion" 1000X more compelling than the occasional "Player X is an ECHL'r" junk we have to endure (not from you but you know what I am referring to). I just don't see Twabby saying "I'm right and Lavi is wrong" any more or less than dozens of other posters here who form their opinions in all sorts of different ways. You all are singling him (her?) out for using stats to support his opinions, but misrepresenting how he characterizes said opinions.

I think our collective "eye tests" are much more susceptible to bias, both recency and personal, than looking at statistics, but neither one tells a complete picture without the other. This forum, like all forums, is littered with people who think they know more than the coaches. Most of them use no evidence whatsoever to support their takes, Twabby provides evidence for his takes all the time. I personally hope he continues posting these analytics because it truly helps form a more complete picture, even if we all draw our own different conclusions in our own different ways.

Also, Lavi has been everything we hoped for and he seems to be getting virtually everything right so far, but it would be pretty boring if we all just sat around thinking there's no room for improvement. In the end we're really spoiled, bitching at each other about how to fix our 18-2 3rd period lead issues on a team tied for points lead in the NHL. But then again, Reirden had solid RS records too, but the flaws we saw creeping in crushed us in the playoffs. So, we're Caps fans after all. Even in 1st place, we're going to complain about something. Its what we do.

OK

 

[twabby]

It's especially disappointing that analytics posts have already been relegated to an outcast thread because people can't care to read (or ignore!) analytics posts in the main thread, and then a moderator of all people Kramers her way into this thread in a low effort way, admittedly saying that they "don't really care" about the topic at hand, and shits up the thread with blatant mischaracterizations.

 

[Calicaps]

It's especially disappointing that analytics posts have already been relegated to an outcast thread because people can't care to read (or ignore!) analytics posts in the main thread, and then a moderator of all people Kramers her way into this thread in a low effort way, admittedly saying that they "don't really care" about the topic at hand, and shits up the thread with blatant mischaracterizations.

I'm sorry you're disappointed. For the record, you've not been "relegated" anywhere. So if you object to mischaracterizations, stop making them.

PS: And I commented because Dave invoked my post from another thread and I wanted to respond. Not sure how that's "kramering."

 

[g00n]

I don't think you understand how correlation works. An r^2 of 0.23 doesn't mean it's worse than a coin flip. A coin-flip has an r^2 of 0 because there is no correlation between one flip on the next.

I don't remember the details of the argument from a month ago but I recall they were related to RELATIVE predictive power, not absolute predictive power. It wasn't 23% predictability but it was still a low enough number vs 100% correlation.

But that's not even the real issue.

Posts like these project an air of relative certainty or NEAR-certainty that the chosen stat is accurate to reality, beyond what people are observing or believe:

Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2

Speculation: - Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021 "Season" Pt. 2

Let's not focus entirely on the semantics of stat accuracy or certainty. The thing that rubs people the wrong way the most is being told that things with no quantifiable metric have no value, or that only players with X value fancy stats should be playing.

That should not be debatable as a fact. We've had many discussions in the past about puck-moving d-men and icing a squad of nothing but. Or how players with low offensive potential or low analytics values are seen as being overused by the coach.

This is a STAPLE of almost every argument by analytics fans and this board is no exception.

 

[ALLCAPSALLTHETIME]

Who has said anything like that?

I see Twabby suggesting, even in the post immediately above this, that there is evidence that supports other decisions. That's not saying he "knows for certain" that the coach is making mistakes.

Agree 100%.

 

[g00n]

It's especially disappointing that analytics posts have already been relegated to an outcast thread because people can't care to read (or ignore!) analytics posts in the main thread, and then a moderator of all people Kramers her way into this thread in a low effort way, admittedly saying that they "don't really care" about the topic at hand, and shits up the thread with blatant mischaracterizations.

Personally IDGAF if you post charts and stats out the wazoo in the other threads. Knock yourself out. Some I'm even interested in. The issues you have are with the rules, and other posters who don't like looking at that stuff all the time. No need to go over it again.

What burns my beans is when you post shit like this (quoted here):

Speculation: - Acq./Rost. Bldg./Cap/Lines etc. Part LXVII

Granted it's an older example but I think the principle is the same: advanced offensive stats trump "unmeasurables", or anything that smacks of defense-over-offense.

Didn't we learn anything about overfocus on these things from that period, and the Cup run?

 

[CapitalsCupReality]

Personally IDGAF if you post charts and stats out the wazoo in the other threads. Knock yourself out. Some I'm even interested in. The issues you have are with the rules, and other posters who don't like looking at that stuff all the time. No need to go over it again.

What burns my beans is when you post shit like this (quoted here):

Speculation: - Acq./Rost. Bldg./Cap/Lines etc. Part LXVII

Granted it's an older example but I think the principle is the same: advanced offensive stats trump "unmeasurables", or anything that smacks of defense-over-offense.

Didn't we learn anything about overfocus on these things from that period, and the Cup run?

If he had his wish Orpik is in the pressbox and we (probably) don’t have a Cup yet.

this is where I see stats and real knowledge of the game part paths.....over and over and over.

I love this thread.....I wish it had more action. It’s fun to see the clay pigeons going up and the shattered parts land.

what I really wish is that we had another hardcore math guy or gal who was also say.....also a 35 year fan and student of this game.

I think we wouldn’t see the numbers constantly crammed down our throats in such disingenuous fashion as they would be accompanied by a more compete understanding of the game and how other factors matter....factors that often can’t be assigned a number.

on any given night we know so little about these guys and their “readiness” to compete or be loaded with additional responsibilities.

we don’t know what player has nagging injuries, what player sucked at practice the last 2 days, or which players were on fire in practice.

Those are all things (and there are sooooo many more) that I trust Lavi and his staff have a handle on, and when it comes to roster choices or deployment or playing time, he’s probably doing pretty f’ing alright overall.