95 replies to this topic

[DJ MC]

Doesn't it get old saying, "Everything I can't immediately see or don't understand is magic and witchcraft?"

[Pedro Cerrano]

Nope.

[RShack]

It's also possible that the O's have been an outlier for more random reasons. Maybe some of it has been tangible reasons, but a lot of it has been simple variance. 

 

Well, it might be possible, but I think it's highly unlikely.  

 

Just because stats can't see something in complex human team performance, that's no reason to conclude it doesn't exist.

 

BTW, on what basis do you conclude that "maybe some of it" is real, but "a lot of it" is simple variance?  That's a very value-laden statement.  You could just as well have sad "maybe some of it" is simple variance but "a lot of it" is real... but you didn't... and I expect we can agree that it's not just an accident that you phrased it the way you did.   Unless there's something I'm missing, this seems to be a faith-based conclusion, not an evidence-based conclusion.  Is there something I'm missing?

[mweb08]

Well, it might be possible, but I think it's highly unlikely.  

 

Just because stats can't see something in complex human team performance, that's no reason to conclude it doesn't exist.

 

BTW, on what basis do you conclude that "maybe some of it" is real, but "a lot of it" is simple variance?  That's a very value-laden statement.  You could just as well have sad "maybe some of it" is simple variance but "a lot of it" is real... but you didn't... and I expect we can agree that it's not just an accident that you phrased it the way you did.   Unless there's something I'm missing, this seems to be a faith-based conclusion, not an evidence-based conclusion.  Is there something I'm missing?

 

Use the maybe for that entire sentence. So maybe there has just been a lot of simple variance and maybe some of it has been tangible reasons. Better? Ricker is the one being definitive about this, not me.

[bnickle]

3 years.Off by almost 20 games last year. God knows how many in '12. Screw random variance. Nothing to do with random variance. At the least, there is something the system isn't measuring with this team in particular. More likely it's just a heavily flawed system.


I get looking at advanced metrics and stats once the season is underway or over. It's retarded to base any predictions on human players heading into a season. So many variables that can't be measured. Then to take it a step further and collectively add up all those flawed evaluations and predictions and make a sweeping prediction of the team win total. Ridiculous.

[bnickle]

You know Webers variance defense really has me worked up right now. So we were predicted for how many wins last year 75-76. Whatever it was. You are telling me that with variance if we played that season 100 times that numerous times we only win 76 games last year. Sometimes less. Not even close. That's without Wieters and Machado for a combined 8 months. Their stats were part of the prediction system.

[bnickle]

Really though doesn't it get old saying: "Computer stat says this, so this is what I think"?

Some of these writers pen whole articles about teams based on nothing but advanced metrics. So Weber has no leg to stand on in regards to some people when he says it's insulting to say they are a slave to them. I guarantee you some of these guys write articles and don't watch that team play a game more than a few times a year. They are completely a slave to them when they pen an article about a team they don't watch play the game.

[mweb08]

You know Webers variance defense really has me worked up right now. So we were predicted for how many wins last year 75-76. Whatever it was. You are telling me that with variance if we played that season 100 times that numerous times we only win 76 games last year. Sometimes less. Not even close. That's without Wieters and Machado for a combined 8 months. Their stats were part of the prediction system.

 

Well I have no idea why it would get you or anyone else worked up. No, I'm not telling you that at all Brandon and there is no way that what I've said can reasonably be interpreted that way.

 

PECOTA, any other predictive model, you, me, everyone else on here, Vegas, the national media, etc are all flawed in the ways we come up with predictions. Every person and model is going to be off on certain teams, and sometimes by a wide margin. Sometimes they are off because of weird stuff happening, sometimes they are off due to great or poor injury luck, sometimes they are off because of variance, sometimes they are off because people or models over or underestimated certain aspects, etc, and most likely they are off because of some combination of all of it. And even if they nail a number, a lot of those parts about being off would still apply, they would just balance out.

 

If I'm not mistaken, this board has underestimated the O's quite a bit over the past 3 seasons, not as much as PECOTA, but still nothing to be pounding our chest over. The predictions at the foxhole before that were always over optimistic as a group. So maybe our collective predictions have been useless too? I don't get the vitriol over these systems and advanced stats. If people really were slaves to them, then sure, but that's not at all what's happening here. Again, I have a lot of divergence from PECOTA in my predictions, but I find it and other systems like it to be useful tools for some baseline data and then you can make adjustments as you see fit.

[mweb08]

Some of these writers pen whole articles about teams based on nothing but advanced metrics. So Weber has no leg to stand on in regards to some people when he says it's insulting to say they are a slave to them. I guarantee you some of these guys write articles and don't watch that team play a game more than a few times a year. They are completely a slave to them when they pen an article about a team they don't watch play the game.

I said people, by and large. So are there some outliers that are like you guys say, yeah, probably, but it is insulting to act like that's a position that is common.

Also, Ricker made a similar insulting comment directly to me. So yes, I have a leg to stand on when BS like that is said.

[You Play to Win the Game]

Weber - I apologize for the insult. But what you said on the show was far more insulting than anything I said... that me being "critical" of PECOTA is ironic since I like to cite KenPom. Really? I can't have a problem with one measure, while liking another? All advanced metrics are created equal eh? I even criticized KenPom the other night when Stoner was talking about tempo. How I'm not going to put much stock into a tempo metric where a bunch of god-awful teams fill out the top 60 in that metric.

 

All I wanted to hear some people say is that, yes, when it comes to TEAM predictions, PECOTA struggles to peg the O's due to the way they win games. You aren't even willing to say that. You just say, eh, it's a prediction tool, no prediction tool is perfect, it's probably just due to some variance (for 3 years and counting now). And the thing is, we KNOW that they aren't capturing defense properly, that is a fact.

 

And I'm not a fan at all of comparing science model based predictions to that of human beings. Isn't the science based model supposed to remove human bias with the goal in mind of being MORE accurate? It has eliminated human bias, yet is is LESS accurate with the O's on a consistent basis. Why? Because what the O's do is hard to quantify with advanced metrics. I have no doubt that one day, that won't be the case, but it isn't there yet. So using PECOTA for TEAM projections as a baseline for the O's, to me, isn't useful. In normal cases, sure, it's fine. I never said otherwise, and I'm not throwing the baby out with the bath water.

 

So yeah, excuse me for taking issue with a system that keeps getting my team wrong, while, oh my gosh, still using some other stats and metrics that I like. How ironic. What a sin.

[mweb08]

Weber - I apologize for the insult. But what you said on the show was far more insulting than anything I said... that me being "critical" of PECOTA is ironic since I like to cite KenPom. Really? I can't have a problem with one measure, while liking another? All advanced metrics are created equal eh?

 

You can't be serious here. No, that is not far more insulting than anything you said. You use KenPom in the same manner that many people use PECOTA and various advanced stats. If you can't see that, then I don't what to tell you and I'm not apologizing for saying that. That wasn't insulting at all. You on the other hand were being a complete dick just for the sake of being one with zero credence behind what you said.

[mweb08]

As for the rest of the post. Take issue with it, that's perfectly fine and I never said it's close to perfect. I just said that I think it's useful and I've explained why. If it weren't for this thread, I might never even mention the PECOTA predicted win total for the O's this year and if I did, I'd be sure to mention that I think it's low and that they've underestimated the O's in recent years.

[You Play to Win the Game]

You can't be serious here. No, that is not far more insulting than anything you said. You use KenPom in the same manner that many people use PECOTA and various advanced stats. If you can't see that, then I don't what to tell you and I'm not apologizing for saying that. That wasn't insulting at all. You on the other hand were being a complete dick just for the sake of being one with zero credence behind what you said.

 

You are lying to yourslf dude. You smugly said, it's ironic he has an issue with PECOTA given that he cites KenPom 10 times a day. Seriously?

 

Whatever man, tell yourself what you want, but that was a dick move and not even remotely an accurate summation of my stance on PECOTA, KenPom or advanced stats in general.

[You Play to Win the Game]

As for the rest of the post. Take issue with it, that's perfectly fine and I never said it's close to perfect. I just said that I think it's useful and I've explained why. If it weren't for this thread, I might never even mention the PECOTA predicted win total for the O's this year and if I did, I'd be sure to mention that I think it's low and that they've underestimated the O's in recent years.

 

Which is what I was doing, so what exactly is the problem? I guess that I take it a step further and suggest that PECOTA is incapable of pegging the O's? Is that what has you twisted up about my stance the past couple days?

[bnickle]

KenPom is also not a predictive metric. I think he does do a preseason ranking but it's as baseless as PECOTA. Ricker is citing Kenpom while in the middle of the season. Big difference.

[You Play to Win the Game]

KenPom is also not a predictive metric. I think he does do a preseason ranking but it's as baseless as PECOTA. Ricker is citing Kenpom while in the middle of the season. Big difference.

 

Another good point. And I definitely don't cite KenPom to form my thoughts, I cite it to see if what I'm seeing matches up. Big difference there. I have some issues with it as well.

 

Look, maybe I just took what he said the wrong way and got all worked up over nothing. It certainly wouldn't be the first time, and we all know it won't be the last either. For whatever reason, I seem to be wired to assume that something said in jest is meant as more of a dig than it really is. If you truly weren't being an ass there, then I'm sorry for blowing this whole thing out of proportion. It's kinda what I do though.

[bnickle]

Well I have no idea why it would get you or anyone else worked up. No, I'm not telling you that at all Brandon and there is no way that what I've said can reasonably be interpreted that way.

PECOTA, any other predictive model, you, me, everyone else on here, Vegas, the national media, etc are all flawed in the ways we come up with predictions. Every person and model is going to be off on certain teams, and sometimes by a wide margin. Sometimes they are off because of weird stuff happening, sometimes they are off due to great or poor injury luck, sometimes they are off because of variance, sometimes they are off because people or models over or underestimated certain aspects, etc, and most likely they are off because of some combination of all of it. And even if they nail a number, a lot of those parts about being off would still apply, they would just balance out.

If I'm not mistaken, this board has underestimated the O's quite a bit over the past 3 seasons, not as much as PECOTA, but still nothing to be pounding our chest over. The predictions at the foxhole before that were always over optimistic as a group. So maybe our collective predictions have been useless too? I don't get the vitriol over these systems and advanced stats. If people really were slaves to them, then sure, but that's not at all what's happening here. Again, I have a lot of divergence from PECOTA in my predictions, but I find it and other systems like it to be useful tools for some baseline data and then you can make adjustments as you see fit.

A lot of people put more credence into a PECOTA prediction vs a human prediction and you know that. The thought being there is no bias coming from the computer, but in fact, the Os are proving there is indeed just that.


Anyway, my biggest gripe is with excusing this in almost anyway with variance. You can't chalk something as big as a 20 game swing and 3 years worth of bitching the system up to variance. If nearly everything goes wrong for the Os last year they still win 76. They could play that season from the beginning 100 times and they will win more than 76 games 95 times. That's not variance, that's a flaw that shows the system is off.

[mweb08]

You are lying to yourslf dude. You smugly said, it's ironic he has an issue with PECOTA given that he cites KenPom 10 times a day. Seriously?

 

Whatever man, tell yourself what you want, but that was a dick move and not even remotely an accurate summation of my stance on PECOTA, KenPom or advanced stats in general.

 

Nope. And what I'm saying is totally accurate if you can undertand the context I'm using it in.

 

Basically, someone said something to you about you being too relient on KenPom, and you responded that you do watch the games and have divergent opinions from KenPom. But when it comes to this discussion, you attack people for being slaves to the computer and attack me specifically for never thinking advanced stats/models have flaws in them. Well my defense is the same as yours, I watch games and have divergent opinions, plus I do find flaws in them. And in both cases, the person defending themselves is just to do so IMO. That's why it's ironic IMO.

[You Play to Win the Game]

Nope. And what I'm saying is totally accurate if you can undertand the context I'm using it in.

 

Basically, someone said something to you about you being too relient on KenPom, and you responded that you do watch the games and have divergent opinions from KenPom. But when it comes to this discussion, you attack people for being slaves to the computer and attack me specifically for never thinking advanced stats/models have flaws in them. Well my defense is the same as yours, I watch games and have divergent opinions, plus I do find flaws in them. And in both cases, the person defending themselves is just to do so IMO. That's why it's ironic IMO.

 

What I'm annoyed about with ths PECOTA garbage is that when to predicting the O's wins, there isn't any debate about bein divergent about it. It's a fact that it's off base and is unable to appropriately predict the O's. It's not variance, it's not just a bad luck year with predicting the O's... it's just simply not capable, with it's component parts, to capture how the O's win games. That is what I'm getting pissed about when it comes to some of the more stat favored people not being able to acknowledge that point. It's like, acknowledging a real flaw that basically diminishes it's worth to the O's TEAM predictions is admitting a flaw in using advanced metrics, so you don't want to admit it.

[mweb08]

I really don't know what I'm supposed to admit? I've admitted it's flawed. I've admitted it's been very wrong about the O's the last 3 years. I've admitted that my predictions in the past and for the future are not dependent on PECOTA at all and are often far apart from the system's projection. As far as how exactly it is flawed in relation to the O's, I speculated on that last night as did Rob, and you and others speculated on it on here. That's all it is, though, speculation. We don't know exactly why PECOTA has underestimated the O's and we don't know how much variance has contributed to it. I think that's a pretty fair stance. I guess you want more. Sorry.