SoxProspects News
|
|
|
|
Legal
Forum Ground Rules
The views expressed by the members of this Forum do not necessarily reflect the views of SoxProspects, LLC.
© 2003-2024 SoxProspects, LLC
|
|
|
|
|
Forum Home | Search | My Profile | Messages | Members | Help |
Welcome Guest. Please Login or Register.
WAR and More (...what is it good for)
|
Post by jimed14 on Aug 6, 2018 14:17:36 GMT -5
Question for the stats guys, about this: Donald: Hey Keith, thanks for doing these chats. Last week, you said “xwOBA doesn’t work”, I’m not trying to argue it does and you are wrong, but I was just curious if you could elaborate on that a little more and explain why it doesn’t work.
Keith Law: I can’t find the link but someone – Russell Carleton, I think – showed that they don’t actually predict future wOBA (or whatever the stat is in the x-stat).
I always thought the 'x' meant what a stat should have been rather than what it actually was. If that's the case, what does predictability have to do with it ? I guess what I'm asking is, does the x (expected) look backwards or forwards. I always thought it meant backwards. It does exactly what you think it does, looks backwards. Keith seems to have the same issue I had with Erics use of xwOBA. It looks backwards so it's not an end all stat to be used to say who we should get for relievers or why we shouldn't acquire another reliever. It lets you know how much good or bad luck (somewhat) the player has had to get the results they have gotten. It lets you know how much room the player has to go in one direction or another without any major change to how they're playing. For example, Didi Gregorius had a wOBA around the same that Mookie had almost a month into the season, but his expected wOBA was about 150 points lower. They both kept hitting about the same and now you see the difference between them. No matter what anyone does with advanced statistics, they'll never be able to predict the future.
|
|
rjp313jr
Veteran
Posts: 14,039
Member is Online
|
Post by rjp313jr on Aug 6, 2018 14:57:23 GMT -5
It does exactly what you think it does, looks backwards. Keith seems to have the same issue I had with Erics use of xwOBA. It looks backwards so it's not an end all stat to be used to say who we should get for relievers or why we shouldn't acquire another reliever. It lets you know how much good or bad luck (somewhat) the player has had to get the results they have gotten. It lets you know how much room the player has to go in one direction or another without any major change to how they're playing. For example, Didi Gregorius had a wOBA around the same that Mookie had almost a month into the season, but his expected wOBA was about 150 points lower. They both kept hitting about the same and now you see the difference between them. No matter what anyone does with advanced statistics, they'll never be able to predict the future. It “supposedly” tells us what you described and what’s the point of that? To try and predict the future. However, if a study can debunk that it does that with any certainty then why are we buying that it tells us how things would have been in the first place?
|
|
|
Post by jimed14 on Aug 6, 2018 15:05:11 GMT -5
It lets you know how much good or bad luck (somewhat) the player has had to get the results they have gotten. It lets you know how much room the player has to go in one direction or another without any major change to how they're playing. For example, Didi Gregorius had a wOBA around the same that Mookie had almost a month into the season, but his expected wOBA was about 150 points lower. They both kept hitting about the same and now you see the difference between them. No matter what anyone does with advanced statistics, they'll never be able to predict the future. It “supposedly” tells us what you described and what’s the point of that? To try and predict the future. However, if a study can debunk that it does that with any certainty then why are we buying that it tells us how things would have been in the first place? What study debunked that? What xwOBA cannot do is account for future injuries, real improvements, and extended slumps. It can only predict what someone should do if they continue playing the same exact way for a long enough period of time. That's better than not being able to predict anything, but it's still not predicting the future. Every time there's some new stat, people start taking shots at it. That's kind of annoying. No one says that anything is perfect, but it's moving in the right direction. It's always a work in progress. At one point, all we had was BABIP and HR/FB ratio, which is almost meaningless in comparison to xwOBA.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,933
|
Post by ericmvan on Aug 6, 2018 16:15:40 GMT -5
Hembree is what he is, not a bad guy at the end of your bullpen. He really owes Cora a thank you for using him in the perfect situations. If you line him up against the bottom of crappy lineups he doesn't look bad. Thing is you don't want him pitching against teams with deep lineups. Which is what I've been saying for a month now. He's going to be rather useless in the playoffs if we have a long run. There's some truth in this, but the main point is backwards. He's been awful this year against great hitters. That does limit his usefulness in the post-season.
But he has absolutely dominated good hitters. He's been great against average hitters. These are actually the guys who fill out a deep lineup. He's been below average against everyone else. He's been just average against the worst hitters, and he's had real problems with the HR ball against merely subpar ones, largely high-K, low-BA, high-Iso types. He's given up 5 HR, all to batters with wRC+ below 90. (That no great hitters has gone yard yet, however, may just be luck.)
I defined "great hitters" by sorting on RC+ and stopping when the names ceased to be scary (wRC+ >= 127, which is Stanton). It of course includes some guys having great seasons without a reputation (yet) to match. I then defined average hitters as everyone with a wRC+ better than 90 but worse than 110. I defined the "poor" hitters to be the same number of guys as the good hitters (the guys 110 to 126), and that left a roughly equal number of terrible ones. Here's the data. I've just included how Hembree's numbers differ from what the rest of MLB has done against these hitters (weighting by how many times each hitter has faced Hembree). Opp wRC+ Guys PA BA OBP SA dBA dOBP dSA 127+ 16 21 .281 .369 .516 .249 .250 .190 110-126 29 45 .266 .338 .459 -.123 -.050 -.231 91-109 33 49 .255 .326 .408 -.041 -.020 -.194 71-90 29 38 .238 .300 .376 -.022 -.063 .219 0-70 34 45 .199 .261 .298 .015 -.017 .035 The great hitters: Machado is 2/3, 2B. Khris Davis and Choo are 1/1, 2B. Torres is 1/1, BB. Wilson Ramos and Mark Reynolds (the only guy with a SSS, but I left him in) are 1/1. Stanton is 1/2. Bregman, Chapman, and Daniel Robertson walked. Hicks is 0/2 and Soto, Cruz, Hoskins, and Altuve are 0/1. That's .529 / .619 / .706.
The group of 29 hitters with wRC+ of 110 to 126 include Smoak, Soler, Merrifield, Lowrie, Rosario, Odor, Abreu, Andujar, Springer, Gregorius, and Segura. Hembree has held them to .143 / .289 / .229. If you combine the 0-70 and 71-90 groups, they've hit .217 / .279 / .334, but they've gone .215 / .241 / .456 against Hembree.
My scouting take on this:
He doesn't have the stuff to get the best hitters out, because they won't chase his slider or the FB to weak zones. The good hitters are more confident and more aggressive than lesser hitters and he eats them alive. Ditto for the average hitters versus the subpar, to a lesser degree. He's actually in theory a bit better than average against the subpar, as you'd expect, but he's had some concentration lapses (mostly early in the season) against the guys who can hurt you but shouldn't (Sanchez, Carlos Gomez, Matt Joyce, Mike Zunino, Pedro Severino).
This is a really useful post-season guy, if you know enough to keep him away from the best hitters when the game is on the line.
|
|
|
Post by umassgrad2005 on Aug 6, 2018 16:36:33 GMT -5
It does exactly what you think it does, looks backwards. Keith seems to have the same issue I had with Erics use of xwOBA. It looks backwards so it's not an end all stat to be used to say who we should get for relievers or why we shouldn't acquire another reliever. It lets you know how much good or bad luck (somewhat) the player has had to get the results they have gotten. It lets you know how much room the player has to go in one direction or another without any major change to how they're playing. For example, Didi Gregorius had a wOBA around the same that Mookie had almost a month into the season, but his expected wOBA was about 150 points lower. They both kept hitting about the same and now you see the difference between them. No matter what anyone does with advanced statistics, they'll never be able to predict the future. It can be useful, depending on how you use it. My point was Eric was using it as predictive. You don't need another reliever because Hembree has this xwOBA. The killer is that players don't preform the same way for a whole season, well most don't. They go up and down for a bunch of reasons. Hembree is a perfect example of that. Bradley is another player who proves xwOBA is rather useless, it wasn't just luck either. It was things like shifts and who knows what else. Nevermind it seems just like BAbip players can under or over perform for a whole year or year after year. It also doesn't take the quality of the opponent into account or even the quality of the batters. So yea more information is good, but its not close to an end all everything type stat that some people were using it as.
|
|
rjp313jr
Veteran
Posts: 14,039
Member is Online
|
Post by rjp313jr on Aug 6, 2018 16:43:54 GMT -5
It “supposedly” tells us what you described and what’s the point of that? To try and predict the future. However, if a study can debunk that it does that with any certainty then why are we buying that it tells us how things would have been in the first place? What study debunked that? What xwOBA cannot do is account for future injuries, real improvements, and extended slumps. It can only predict what someone should do if they continue playing the same exact way for a long enough period of time. That's better than not being able to predict anything, but it's still not predicting the future. Every time there's some new stat, people start taking shots at it. That's kind of annoying. No one says that anything is perfect, but it's moving in the right direction. It's always a work in progress. At one point, all we had was BABIP and HR/FB ratio, which is almost meaningless in comparison to xwOBA. Keith Law referenced a study. And why do you say xOBA does that? Because someone says it does? Weight a bunch of adjustments come up with a number and say that’s what really should have happened? It may be the intention of the formula, but it doesn’t make it accurate. Just saying, when a guy like Keith Law, who loves advanced stats, says one is useless, I’m going to pay attention. Edit: to put it another way, if there is no correlation to how a player will do the following season it’s a useless statistic. Having the conversation “according to this number, player A should have done y instead of x last year but that means nothing as to what to expect from them next year” leads me to believe maybe player A shouldn’t have really done y instead of x.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,933
|
Post by ericmvan on Aug 6, 2018 16:44:46 GMT -5
Question for the stats guys, about this: Donald: Hey Keith, thanks for doing these chats. Last week, you said “xwOBA doesn’t work”, I’m not trying to argue it does and you are wrong, but I was just curious if you could elaborate on that a little more and explain why it doesn’t work.
Keith Law: I can’t find the link but someone – Russell Carleton, I think – showed that they don’t actually predict future wOBA (or whatever the stat is in the x-stat).
I always thought the 'x' meant what a stat should have been rather than what it actually was. If that's the case, what does predictability have to do with it ? I guess what I'm asking is, does the x (expected) look backwards or forwards. I always thought it meant backwards. It does exactly what you think it does, looks backwards. Keith seems to have the same issue I had with Erics use of xwOBA. It looks backwards so it's not an end all stat to be used to say who we should get for relievers or why we shouldn't acquire another reliever. I've been guilty of doing that a bit. You can't look up everyone's game log and deep data (unless they're paying you to do it). I crossed off Herrera as a trade target because his xwOBA wasn't good enough, along with a bunch of other guys that I appear to have been right about. It turns out that may have actually had a trade for Herrera that Nats' ownership vetoed ... if that trade had happened, I'm pretty sure that I would have liked it once I did look at the aforementioned further info.
Soria is a guy I did look at deeper data for, and I didn't like what I saw. I looked at the same data for Romo and Stammen before saying I liked them better ... but there were many guys that got no more than a look at xwOBA, and I 100% agree that it's just the starting point of any assessment. (In Herrera's case it was misleadingly weak.) You still have to project how a guy will pitch, but any time you do that, you want to start with how well a guy has actually pitched, so any stat that removes noise is good.
Eovaldi had the best xwOBA of any rental starter. That seems to be working out so far, in part because it was inflated, relative to predictive talent, by both his lesser early starts coming back from the TJ, and by a pair of incomprehensible slow hooks. I obviously knew about the former but discovered the latter only after the deal.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,933
|
Post by ericmvan on Aug 6, 2018 17:05:22 GMT -5
It lets you know how much good or bad luck (somewhat) the player has had to get the results they have gotten. It lets you know how much room the player has to go in one direction or another without any major change to how they're playing. For example, Didi Gregorius had a wOBA around the same that Mookie had almost a month into the season, but his expected wOBA was about 150 points lower. They both kept hitting about the same and now you see the difference between them. No matter what anyone does with advanced statistics, they'll never be able to predict the future. It can be useful, depending on how you use it. My point was Eric was using it as predictive. You don't need another reliever because Hembree has this xwOBA. The killer is that players don't preform the same way for a whole season, well most don't. They go up and down for a bunch of reasons. Hembree is a perfect example of that. Bradley is another player who proves xwOBA is rather useless, it wasn't just luck either. It was things like shifts and who knows what else. Nevermind it seems just like BAbip players can under or over perform for a whole year or year after year. It also doesn't take the quality of the opponent into account or even the quality of the batters. So yea more information is good, but its not close to an end all everything type stat that some people were using it as. What does Hembree (and your irrational distaste for him) have anything to do with acquiring another reliever? His role has been very clearly defined and it has nothing to do with the end of a post-season game. It's to bail out an inherited-runner situation in the 6th, and to go 2 innings when necessary. The question was always whether we needed to add a 7th / 8th inning guy. Here's what Dombrowski said, parroting what I was saying all along: you only want to get a guy who is better than Barnes and Thornburg.
The guys who had too high an asking price seem to have been Tyler Yates and any one of several guys with 3 or 4 years of control, since it was widely reported that the main purpose of a reliever trade would be to acquire a Kimbrel replacement. The guys where "they went in a different direction" were Britton (dealt him to another team) and Herrera (decided not to be sellers). But none of the rental guys other than Britton and Herrera appear to have been of interest.
|
|
|
Post by jimed14 on Aug 6, 2018 17:10:27 GMT -5
What study debunked that? What xwOBA cannot do is account for future injuries, real improvements, and extended slumps. It can only predict what someone should do if they continue playing the same exact way for a long enough period of time. That's better than not being able to predict anything, but it's still not predicting the future. Every time there's some new stat, people start taking shots at it. That's kind of annoying. No one says that anything is perfect, but it's moving in the right direction. It's always a work in progress. At one point, all we had was BABIP and HR/FB ratio, which is almost meaningless in comparison to xwOBA. Keith Law referenced a study. And why do you say xOBA does that? Because someone says it does? Weight a bunch of adjustments come up with a number and say that’s what really should have happened? It may be the intention of the formula, but it doesn’t make it accurate. Just saying, when a guy like Keith Law, who loves advanced stats, says one is useless, I’m going to pay attention. Edit: to put it another way, if there is no correlation to how a player will do the following season it’s a useless statistic. Having the conversation “according to this number, player A should have done y instead of x last year but that means nothing as to what to expect from them next year” leads me to believe maybe player A shouldn’t have really done y instead of x. That's not the only thing statistics are used for. According to xwOBA, you shouldn't bench this guy who is hitting the crap out of the ball and not getting results. According to xwOBA, you should not be trading for this guy who is getting results way better than they should be. According to xwOBA, you could possibly trade low for this guy. If you don't think xwOBA is accurate, then you should be waiting for it to be fixed. It's an incredibly useful advanced statistic that will undoubtedly be improved. I think it's lousy at dealing with park factors and shifting right now, but it should get improved over time, not just thrown away and ignored.
|
|
|
Post by philsbosoxfan on Aug 6, 2018 18:28:35 GMT -5
LOL, I seem to have touched off a debate and it also appears that the debate isn't just here as I interpret scottysmalls post, it appears to be among the pros as well.
Glad I had it right, Law's statement confused me. No statistic viewing past performance can predict future performance. They can, however be applied to probability bands which has major value in trying to quantify the future.
The prediction game is ummm marginal at best. Later in the LAW chat someone pointed out that in the PECOTA team wins predictions (which should be somewhat more accurate than predicting individuals) their BEST year was off by an average of 9 wins per team.
For me, since the intent of scouting is inherently prediction, when it comes to predicting player futures I will still prefer scouting to stats.
It also looks like SABR needs to clear up the confusion on the word 'expected'. That's a bad look for them.
And for the record, I never viewed eric's posts using x numbers as anything more than an examination of past performance, comparing what happened to various players.
|
|
|
Post by philsbosoxfan on Aug 6, 2018 18:43:20 GMT -5
|
|
|
Post by philarhody on Aug 6, 2018 19:18:24 GMT -5
@jlebowski:
In reality, the statistical analysis was pretty good in 2008. But our experience of baseball has changed mightily. We've brought baseball to bear using scientific analysis and some of the goodness of the game has squished from it. A 5th grader says his favorite player is Jackie Bradley Jr. because he roams center field with majesty and runs the bases splendidly. His classmate snorts and recites an OPS number. The game has been so quantitivatively packaged that there is hardly a debate to be had anymore.
With that said, Cain can be the perfect example of how analysis can support the good and true of the game. Sure he only hits 12 home runs a year. But he is a tremendous defensive centerfielder and a smart hitter who is awesome to watch.
|
|
|
Post by manfred on Aug 6, 2018 19:47:15 GMT -5
I think ultimately one of my two issues with many of these stats is that they are always “predictive” in a sense: something like WAR compares a real player with myriad factors mental, physical, contingent, to a mythical replacement and works on the assumption that we can measure by that constant. I don’t mind that for conversations that are limited to discussing some forms of statistical outcomes, but I am suspicious of the utopian idea that the replacement player has some kind of universal exchange value. One simple example: Xander’s error yesterday is possibly a result of the moment, a moment the shortstop for, say, the Orioles simply never has to deal with (cause, you know, they suck). So that error is at once more forgivable on one level and more costly on another than any error a shortstop on crap team makes. I think these statistics grossly underestimate the human factor in an effort to quantify something that is often as much about art and character as skill.
That problem of universal exchange value also bugs me with the weighting by season and position. It is fine, then, to measure just that: but it means that when, say, catchers are down but there are a bunch of strong, say, first basemen, it takes a much better player to get WAR at 1B than C. Similarly, a 5 WAR 1B in a season of strong 1B is better than one in a weaker season. Is this all a misunderstanding? But — unless that is wrong — the WAR argument without all kind of extra info and context is deceptive.
Again, maybe I am misconstruing this, and if I am, I am happy to come around. It would be less bruising on these boards to be a sabermetric firebrand.
|
|
|
Post by jimed14 on Aug 7, 2018 8:18:12 GMT -5
First of all, I have no idea why people would think that expected anything would be predictive since it's still analyzing the past. I always knew before I looked up the definition that xwOBA meant the wOBA that a player should have had with neutral luck and defense during the time period the xwOBA was measured. Second, xwOBA is more predictive about the future than any other stat is. I can't stand this crap every time a new stat is introduced. Oh I found a small flaw, throw it out and ignore it forever now!
|
|
|
Post by James Dunne on Aug 7, 2018 8:28:26 GMT -5
First of all, I have no idea why people would think that expected anything would be predictive since it's still analyzing the past. I always knew before I looked up the definition that xwOBA meant the wOBA that a player should have had with neutral luck and defense during the time period the xwOBA was measured. Second, xwOBA is more predictive about the future than any other stat is. I can't stand this crap every time a new stat is introduced. Oh I found a small flaw, throw it out and ignore it forever now! Isn't Law's point that xwOBA ends up being less predictive than wOBA, and since wOBA is also a better accounting of past performance that he believes it's a preferable stat? Or did I misunderstand him entirely? It's not like Keith Law is going to put his head in the sand and ignore a new metric just because it's new.
|
|
|
Post by jimed14 on Aug 7, 2018 8:38:12 GMT -5
First of all, I have no idea why people would think that expected anything would be predictive since it's still analyzing the past. I always knew before I looked up the definition that xwOBA meant the wOBA that a player should have had with neutral luck and defense during the time period the xwOBA was measured. Second, xwOBA is more predictive about the future than any other stat is. I can't stand this crap every time a new stat is introduced. Oh I found a small flaw, throw it out and ignore it forever now! Isn't Law's point that xwOBA ends up being less predictive than wOBA, and since wOBA is also a better accounting of past performance that he believes it's a preferable stat? Or did I misunderstand him entirely? It's not like Keith Law is going to put his head in the sand and ignore a new metric just because it's new. I haven't read whatever Law was talking about because no one linked it. I still don't see how xwOBA is less predictive or a worse accounting of past performance than wOBA. I want luck and defense taken out of it because a hitter has no control over that. It's especially better in short samples. Is a batter hitting the ball well or not? xwOBA will tell you without penalizing him for hitting 110 mph line drives right at defenders or crediting him for hitting bloop singles.
|
|
|
Post by James Dunne on Aug 7, 2018 8:51:23 GMT -5
Isn't Law's point that xwOBA ends up being less predictive than wOBA, and since wOBA is also a better accounting of past performance that he believes it's a preferable stat? Or did I misunderstand him entirely? It's not like Keith Law is going to put his head in the sand and ignore a new metric just because it's new. I haven't read whatever Law was talking about because no one linked it. I still don't see how xwOBA is less predictive or a worse accounting of past performance than wOBA. I want luck and defense taken out of it because a hitter has no control over that. It's especially better in short samples. Is a batter hitting the ball well or not? xwOBA will tell you without penalizing him for hitting 110 mph line drives right at defenders or crediting him for hitting bloop singles. The problem is that it's impossible to define what is luck and what is skill with that kind of precision. So then we're getting into the descriptive vs. predictive debate again. And I suppose my argument would be that if a statistic that is made to filter out luck ends up being less predictive than one that is not, then it follows logically that it's less descriptive as well, because it must be filtering out some thing or things that is/are not luck.
|
|
|
Post by umassgrad2005 on Aug 7, 2018 8:52:30 GMT -5
Hembree is what he is, not a bad guy at the end of your bullpen. He really owes Cora a thank you for using him in the perfect situations. If you line him up against the bottom of crappy lineups he doesn't look bad. Thing is you don't want him pitching against teams with deep lineups. Which is what I've been saying for a month now. He's going to be rather useless in the playoffs if we have a long run. There's some truth in this, but the main point is backwards. He's been awful this year against great hitters. That does limit his usefulness in the post-season.
But he has absolutely dominated good hitters. He's been great against average hitters. These are actually the guys who fill out a deep lineup. He's been below average against everyone else. He's been just average against the worst hitters, and he's had real problems with the HR ball against merely subpar ones, largely high-K, low-BA, high-Iso types. He's given up 5 HR, all to batters with wRC+ below 90. (That no great hitters has gone yard yet, however, may just be luck.)
I defined "great hitters" by sorting on RC+ and stopping when the names ceased to be scary (wRC+ >= 127, which is Stanton). It of course includes some guys having great seasons without a reputation (yet) to match. I then defined average hitters as everyone with a wRC+ better than 90 but worse than 110. I defined the "poor" hitters to be the same number of guys as the good hitters (the guys 110 to 126), and that left a roughly equal number of terrible ones. Here's the data. I've just included how Hembree's numbers differ from what the rest of MLB has done against these hitters (weighting by how many times each hitter has faced Hembree). Opp wRC+ Guys PA BA OBP SA dBA dOBP dSA 127+ 16 21 .281 .369 .516 .249 .250 .190 110-126 29 45 .266 .338 .459 -.123 -.050 -.231 91-109 33 49 .255 .326 .408 -.041 -.020 -.194 71-90 29 38 .238 .300 .376 -.022 -.063 .219 0-70 34 45 .199 .261 .298 .015 -.017 .035 The great hitters: Machado is 2/3, 2B. Khris Davis and Choo are 1/1, 2B. Torres is 1/1, BB. Wilson Ramos and Mark Reynolds (the only guy with a SSS, but I left him in) are 1/1. Stanton is 1/2. Bregman, Chapman, and Daniel Robertson walked. Hicks is 0/2 and Soto, Cruz, Hoskins, and Altuve are 0/1. That's .529 / .619 / .706.
The group of 29 hitters with wRC+ of 110 to 126 include Smoak, Soler, Merrifield, Lowrie, Rosario, Odor, Abreu, Andujar, Springer, Gregorius, and Segura. Hembree has held them to .143 / .289 / .229. If you combine the 0-70 and 71-90 groups, they've hit .217 / .279 / .334, but they've gone .215 / .241 / .456 against Hembree.
My scouting take on this:
He doesn't have the stuff to get the best hitters out, because they won't chase his slider or the FB to weak zones. The good hitters are more confident and more aggressive than lesser hitters and he eats them alive. Ditto for the average hitters versus the subpar, to a lesser degree. He's actually in theory a bit better than average against the subpar, as you'd expect, but he's had some concentration lapses (mostly early in the season) against the guys who can hurt you but shouldn't (Sanchez, Carlos Gomez, Matt Joyce, Mike Zunino, Pedro Severino).
This is a really useful post-season guy, if you know enough to keep him away from the best hitters when the game is on the line.
Isn't wRC+ just like OPS+ in that 100 is average? Why are you redefining an advanced stat numbers to make Hembree look better? I consider good hitters guys that are above average. Go look at the Yankees line-up its filled with guys above 100. When he has to face those guys one after another he gets lit up. Just look at his stats this year against the Yankees. Look at the Astros hitters, those are the likely teams we need to beat. Those teams aren't filled with crappy hitters.
|
|
|
Post by umassgrad2005 on Aug 7, 2018 9:01:13 GMT -5
It does exactly what you think it does, looks backwards. Keith seems to have the same issue I had with Erics use of xwOBA. It looks backwards so it's not an end all stat to be used to say who we should get for relievers or why we shouldn't acquire another reliever. I've been guilty of doing that a bit. You can't look up everyone's game log and deep data (unless they're paying you to do it). I crossed off Herrera as a trade target because his xwOBA wasn't good enough, along with a bunch of other guys that I appear to have been right about. It turns out that may have actually had a trade for Herrera that Nats' ownership vetoed ... if that trade had happened, I'm pretty sure that I would have liked it once I did look at the aforementioned further info.
Soria is a guy I did look at deeper data for, and I didn't like what I saw. I looked at the same data for Romo and Stammen before saying I liked them better ... but there were many guys that got no more than a look at xwOBA, and I 100% agree that it's just the starting point of any assessment. (In Herrera's case it was misleadingly weak.) You still have to project how a guy will pitch, but any time you do that, you want to start with how well a guy has actually pitched, so any stat that removes noise is good.
Eovaldi had the best xwOBA of any rental starter. That seems to be working out so far, in part because it was inflated, relative to predictive talent, by both his lesser early starts coming back from the TJ, and by a pair of incomprehensible slow hooks. I obviously knew about the former but discovered the latter only after the deal.
Eric wOBA is how a guy has pitched, xwOBA is how they think a guy should have pitched. So using expected wOBA is not looking at how a guy has pitched. That is what I kept telling you Soria for example had a way better wOBA and has for years over Hembree, yet you just kept looking at xwOBA. Then you started adjusting xwOBA, acting like wOBA which is the results didn't matter, only the xwOBA. Edit: My whole point was the past performance matters greatly when getting a pitcher at the deadline. You are getting 2-3 months of future production not past production. Its why Soria wasn't close to Hembree, who's numbers are scewed by one unreal month where he had an BAbip of .000. Its why they wanted Herrera, he's been a great pitcher for years. They weren't going to let 12 innings change anything about that. I think you need to step back my friend and try and look at things fairly. You keep trying to pump up the Red Sox players. No doubt in my mind you would have went from thinking Herrera wasn't good to all of a sudden thinking he was elite again.
|
|
|
Post by jimed14 on Aug 7, 2018 9:06:05 GMT -5
I haven't read whatever Law was talking about because no one linked it. I still don't see how xwOBA is less predictive or a worse accounting of past performance than wOBA. I want luck and defense taken out of it because a hitter has no control over that. It's especially better in short samples. Is a batter hitting the ball well or not? xwOBA will tell you without penalizing him for hitting 110 mph line drives right at defenders or crediting him for hitting bloop singles. The problem is that it's impossible to define what is luck and what is skill with that kind of precision. So then we're getting into the descriptive vs. predictive debate again. And I suppose my argument would be that if a statistic that is made to filter out luck ends up being less predictive than one that is not, then it follows logically that it's less descriptive as well, because it must be filtering out some thing or things that is/are not luck. Isn't the goal for a hitter to hit the ball as hard as possible at the optimum launch angle? That can be measured with xwOBA. And again, this stat can be improved again and again and again to do a better job measuring that. I want to know who is hitting the ball the hardest at the most optimum launch angle because in my opinion, he will always be the best hitter, no matter what their wOBA is. I also want to know who will be coming back down to earth because they're getting unsustainable results.
|
|
|
Post by James Dunne on Aug 7, 2018 9:11:20 GMT -5
The problem is that it's impossible to define what is luck and what is skill with that kind of precision. So then we're getting into the descriptive vs. predictive debate again. And I suppose my argument would be that if a statistic that is made to filter out luck ends up being less predictive than one that is not, then it follows logically that it's less descriptive as well, because it must be filtering out some thing or things that is/are not luck. Isn't the goal for a hitter to hit the ball as hard as possible at the optimum launch angle? That can be measured with xwOBA. And again, this stat can be improved again and again and again to do a better job measuring that. I want to know who is hitting the ball the hardest at the most optimum launch angle because in my opinion, that will always be the best hitter, no matter what their wOBA is. No, the goal is to get to first base. And if intentionally altering the bat angle from what produces an optimal exit velocity allows him to do that then it's not luck. At an extreme example, think of Wade Boggs actually having the combination of pitch recognition and bat control to sort through a pitch until he got the one that he could hit precisely between fielders. Or in an even more basic scenario, bunting on an overshift. Or even a player who alters his approach with two strikes, going from a power to contact approach that might negatively affect his SLG but have a positive effect on his OBP.
|
|
|
Post by umassgrad2005 on Aug 7, 2018 9:22:53 GMT -5
It can be useful, depending on how you use it. My point was Eric was using it as predictive. You don't need another reliever because Hembree has this xwOBA. The killer is that players don't preform the same way for a whole season, well most don't. They go up and down for a bunch of reasons. Hembree is a perfect example of that. Bradley is another player who proves xwOBA is rather useless, it wasn't just luck either. It was things like shifts and who knows what else. Nevermind it seems just like BAbip players can under or over perform for a whole year or year after year. It also doesn't take the quality of the opponent into account or even the quality of the batters. So yea more information is good, but its not close to an end all everything type stat that some people were using it as. What does Hembree (and your irrational distaste for him) have anything to do with acquiring another reliever? His role has been very clearly defined and it has nothing to do with the end of a post-season game. It's to bail out an inherited-runner situation in the 6th, and to go 2 innings when necessary. The question was always whether we needed to add a 7th / 8th inning guy. Here's what Dombrowski said, parroting what I was saying all along: you only want to get a guy who is better than Barnes and Thornburg.
The guys who had too high an asking price seem to have been Tyler Yates and any one of several guys with 3 or 4 years of control, since it was widely reported that the main purpose of a reliever trade would be to acquire a Kimbrel replacement. The guys where "they went in a different direction" were Britton (dealt him to another team) and Herrera (decided not to be sellers). But none of the rental guys other than Britton and Herrera appear to have been of interest.
I don't have any irrational distaste for Hembree. I like the guy as a back of the pen type guy. The only irrational thing going on with Hembree is you pumping him up, mainly because of a crazy June that makes his numbers look better than they are. I don't want Hembree anywhere near the postseason facing the lineups the Yankees and Astros have. Just remember you were the one using expected results then adjusting them showing Hembree was close to an 8th inning guy. The guy they didn't even have pitch in the postseason yet, he was passed over. Hembree won't be some weapon against good teams getting us out of jams, he'll most likely kill us. It will be just like the recent Yankee game all over again, because the teams in the playoffs don't have tons of crappy hitters.
|
|
|
Post by jimed14 on Aug 7, 2018 9:26:14 GMT -5
Isn't the goal for a hitter to hit the ball as hard as possible at the optimum launch angle? That can be measured with xwOBA. And again, this stat can be improved again and again and again to do a better job measuring that. I want to know who is hitting the ball the hardest at the most optimum launch angle because in my opinion, that will always be the best hitter, no matter what their wOBA is. No, the goal is to get to first base. And if intentionally altering the bat angle from what produces an optimal exit velocity allows him to do that then it's not luck. At an extreme example, think of Wade Boggs actually having the combination of pitch recognition and bat control to sort through a pitch until he got the one that he could hit precisely between fielders. Or in an even more basic scenario, bunting on an overshift. Or even a player who alters his approach with two strikes, going from a power to contact approach that might negatively affect his SLG but have a positive effect on his OBP. If it were the 1910s and half the league were slap hitters, I'd tend to agree. But most players today are trying to hit the ball as hard as possible on every swing. So yeah, xwOBA wouldn't work with guys who bunt for base hits. How many of those are there every year now?
|
|
|
Post by jmei on Aug 7, 2018 9:40:33 GMT -5
The problem is that it's impossible to define what is luck and what is skill with that kind of precision. So then we're getting into the descriptive vs. predictive debate again. And I suppose my argument would be that if a statistic that is made to filter out luck ends up being less predictive than one that is not, then it follows logically that it's less descriptive as well, because it must be filtering out some thing or things that is/are not luck. Isn't the goal for a hitter to hit the ball as hard as possible at the optimum launch angle? That can be measured with xwOBA. And again, this stat can be improved again and again and again to do a better job measuring that. I want to know who is hitting the ball the hardest at the most optimum launch angle because in my opinion, he will always be the best hitter, no matter what their wOBA is. I also want to know who will be coming back down to earth because they're getting unsustainable results. There's this perception that being able to isolate exit velocity and launch angle removes a significant portion of luck from the equation. That's not really how that works. Especially in smaller samples, players can have lucky or unlucky exit velocities/launch angles in the same way that players have lucky/unlucky walk rates or strikeout rates (which stabilize a lot more quickly than any batted ball statistic). It's more sophisticated than the old expected BABIP-type analysis that we used to do, but only by a little bit and not by leaps and bounds, and there is still a lot of mostly-random noise, especially in smaller samples.
|
|
|
Post by jimed14 on Aug 7, 2018 9:58:59 GMT -5
Isn't the goal for a hitter to hit the ball as hard as possible at the optimum launch angle? That can be measured with xwOBA. And again, this stat can be improved again and again and again to do a better job measuring that. I want to know who is hitting the ball the hardest at the most optimum launch angle because in my opinion, he will always be the best hitter, no matter what their wOBA is. I also want to know who will be coming back down to earth because they're getting unsustainable results. There's this perception that being able to isolate exit velocity and launch angle removes a significant portion of luck from the equation. That's not really how that works. Especially in smaller samples, players can have lucky or unlucky exit velocities/launch angles in the same way that players have lucky/unlucky walk rates or strikeout rates (which stabilize a lot more quickly than any batted ball statistic). It's more sophisticated than the old expected BABIP-type analysis that we used to do, but only by a little bit and not by leaps and bounds, and there is still a lot of mostly-random noise, especially in smaller samples. I just don't see how luck is a factor in exit velocity. If it is, I don't think it would be significant. It's not like Brock Holt is ever going to luck into a 120 mph 480 foot home run. And how is it unlucky for any batter to hit a weak ground ball? The best hitters will always be near the top of the xwOBA leaderboard. What xwOBA doesn't do is account for shifting and that's the most significant issue with it IMO.
|
|
|