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.
8/28-8/30 Red Sox @ Mets Series Thread
|
Post by geostorm on Aug 30, 2015 18:33:47 GMT -5
Didn't see today's game, but Mets Twitter seems just as angry with West as my Red Sox contingency. Joe West is such a jerk that everyone hated him even when he was a competent umpire. Now that he isn't that either, he's the single worst thing about baseball. Hopefully, his jumping into a crew he typically isn't part of, per Sox radio team, towards expediting his getting to 5000 games, means he's (selfishly) looking to hit that mark this season, and then (fngers crossed) retire...
|
|
|
Post by benogliviesbrother on Aug 30, 2015 19:58:53 GMT -5
Honest question: do the Red Sox bring back any relievers? I'd be fine with them releasing the entire collection of bums? Honest question: do you hold yourself to as high a critical standard as you do others?
|
|
wcp3
Veteran
Posts: 3,809
|
Post by wcp3 on Aug 30, 2015 21:26:19 GMT -5
Honest question: do the Red Sox bring back any relievers? I'd be fine with them releasing the entire collection of bums? Honest question: do you hold yourself to as high a critical standard as you do others? Well I'm perfect, so yes.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,881
|
Post by ericmvan on Aug 30, 2015 23:07:25 GMT -5
For as good as the results have been for Kelly over the last 5 starts, I don't think the peripherals match. His LOB% for his last 5 starts is 91.2% which is unsustainable and his K-BB% is 9.1 which isn't great. His ERA was 1.69, but his xFIP has been only slightly lower than it has been all season at 4.04. It seems that he's getting as lucky now as he was unlucky earlier. He has a BABIP skill. He's had it his whole career. He's using it now, whereas earlier in the season, he wasn't. Now, I made that argument when we got him, after looking at his situational splits. Now I've taken his whole career and looked at the relationship between his BABIP and his BB% and SO%, game by game. You know how noisy game-by-game BABIP rates are? They shouldn't correlate with anything with p = .013 over 70 starts. But, damn, they do, and in a way that both makes baseball sense and fits with observations of Kelly. I'm going to look at some pitch/fx data tonight to see if I can glean some more info, and then start a thread, where I'll explain the findings. BTW, it looks like the Sox brain trust tried to get him away from his BABIP skill, have him stop nibbling and try to strike more guys out. Someone dropped the ball analytically.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,881
|
Post by ericmvan on Aug 30, 2015 23:11:55 GMT -5
Robbie Ross Jr.: The best post-ASG reliever on the 2015 Red Sox. Seriously, while Breslow and Ogando have to go, I wouldn't mind seeing Ross back with the team next year. I actually think the BP is in better shape than most think. Between Uehara, Tazawa, Ross, and Layne, that's four useful-to-better relievers, and Wright can be your long man (he'll be out of options and is unlikely to crack the rotation). They still need two high-leverage guys, but that's something you can add in an offseason. That's exactly how I had it in the 2016 Roster thread, except with Wright in the rotation and Kelly as the long man, in part because you'd trust Kelly more to also pitch in high-leverage situations, and because Wright has the knuckleball effect on the next day's lineup. A bullpen with Tazawa as the 3rd or 4th best option and therefore not getting burnt to a crisp would be very good.
|
|
|
Post by mgoetze on Aug 31, 2015 4:43:52 GMT -5
A bullpen with Tazawa as the 3rd or 4th best option and therefore not getting burnt to a crisp would be very good. Tazawa doesn't automatically have to get burnt to a crisp just because he's the 2nd best option. You could also just hire a manager who would find a better time for his first multi-inning appearance of the year than a rain-shortened blowout.
|
|
|
Post by ethanbein on Aug 31, 2015 9:48:46 GMT -5
He has a BABIP skill. He's had it his whole career. He's using it now, whereas earlier in the season, he wasn't. Now, I made that argument when we got him, after looking at his situational splits. Now I've taken his whole career and looked at the relationship between his BABIP and his BB% and SO%, game by game. You know how noisy game-by-game BABIP rates are? They shouldn't correlate with anything with p = .013 over 70 starts. But, damn, they do, and in a way that both makes baseball sense and fits with observations of Kelly. I'm going to look at some pitch/fx data tonight to see if I can glean some more info, and then start a thread, where I'll explain the findings. BTW, it looks like the Sox brain trust tried to get him away from his BABIP skill, have him stop nibbling and try to strike more guys out. Someone dropped the ball analytically. Eric, you know well enough that a p-value isn't any sort of estimate of the size of an effect. I'm maybe willing to believe Joe Kelly might have a BABIP skill. But - what do you estimate the size of that skill to be? And, if it requires him to have worse K and BB rates to tap into it, that makes it even less of a useful skill. (We can table this discussion for now if you're going to start a new thread).
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,881
|
Post by ericmvan on Sept 1, 2015 1:09:56 GMT -5
He has a BABIP skill. He's had it his whole career. He's using it now, whereas earlier in the season, he wasn't. Now, I made that argument when we got him, after looking at his situational splits. Now I've taken his whole career and looked at the relationship between his BABIP and his BB% and SO%, game by game. You know how noisy game-by-game BABIP rates are? They shouldn't correlate with anything with p = .013 over 70 starts. But, damn, they do, and in a way that both makes baseball sense and fits with observations of Kelly. I'm going to look at some pitch/fx data tonight to see if I can glean some more info, and then start a thread, where I'll explain the findings. BTW, it looks like the Sox brain trust tried to get him away from his BABIP skill, have him stop nibbling and try to strike more guys out. Someone dropped the ball analytically. Eric, you know well enough that a p-value isn't any sort of estimate of the size of an effect. I'm maybe willing to believe Joe Kelly might have a BABIP skill. But - what do you estimate the size of that skill to be? And, if it requires him to have worse K and BB rates to tap into it, that makes it even less of a useful skill. (We can table this discussion for now if you're going to start a new thread). The thread will have to wait a bit because it turns out that his two-seamer and four-seamer blend together in a way that makes it impossible for even the superior Pitch Info classification methodology to tell them apart. So I will either attempt to do so myself semi-manually, or just lump them together. In this case, to get a significant result despite the huge amount of BABIP noise in any individual start, you would need a large effect size. Now, his career BABIP as a starter is .293. But it's .264 in the 42 games where he walked 8% or more of batters (on 701 balls in play), and .333 in the 28 where he walked less than 8% (on 526). (Yes, his BABIP is an inverse function of his BB%.) The odds of getting the BABIP split at random are 1 in 113 (p = .009). You can also do a t-test on the unweighted means of the two groups of starts (which have identical variances), and p = .005. And I finally plotted BB% vs. BABIP and immediately saw that there was one huge outlier where he walked a career-high 25.0% and got killed (2 HR, .385 BABIP, versus Houston last 8/17). It's one of two games in his career that are huge xFIP and FIP outilers, games where he apparently had nothing and any wildness was going to just get him killed. (The other was his 7 R in 1.2 IP, 0 SO, 1 BB game this year in Minnesota.) Remove those two outliers, and you get p = .0017 for the inverse correlation, start by start.
|
|
|
Post by ethanbein on Sept 1, 2015 10:59:10 GMT -5
Eric, you know well enough that a p-value isn't any sort of estimate of the size of an effect. I'm maybe willing to believe Joe Kelly might have a BABIP skill. But - what do you estimate the size of that skill to be? And, if it requires him to have worse K and BB rates to tap into it, that makes it even less of a useful skill. (We can table this discussion for now if you're going to start a new thread). The thread will have to wait a bit because it turns out that his two-seamer and four-seamer blend together in a way that makes it impossible for even the superior Pitch Info classification methodology to tell them apart. So I will either attempt to do so myself semi-manually, or just lump them together. In this case, to get a significant result despite the huge amount of BABIP noise in any individual start, you would need a large effect size. Now, his career BABIP as a starter is .293. But it's .264 in the 42 games where he walked 8% or more of batters (on 701 balls in play), and .333 in the 28 where he walked less than 8% (on 526). (Yes, his BABIP is an inverse function of his BB%.) The odds of getting the BABIP split at random are 1 in 113 (p = .009). You can also do a t-test on the unweighted means of the two groups of starts (which have identical variances), and p = .005. And I finally plotted BB% vs. BABIP and immediately saw that there was one huge outlier where he walked a career-high 25.0% and got killed (2 HR, .385 BABIP, versus Houston last 8/17). It's one of two games in his career that are huge xFIP and FIP outilers, games where he apparently had nothing and any wildness was going to just get him killed. (The other was his 7 R in 1.2 IP, 0 SO, 1 BB game this year in Minnesota.) Remove those two outliers, and you get p = .0017 for the inverse correlation, start by start. Alright, I grabbed the data myself. The average relationship (whether or not you include K%) is about 6 points of BABIP improvement for every extra 1 percentage point of BB%. But, of course, I think you'd agree that you have to regress that to the mean, since its highly, highly likely that at least some of this is caused by chance. I'd guess you have to regress 95%+ of the way to the mean, because I don't think this is a normal split to have, but let's say for now that it's mostly real. 6 points of BABIP of a 1% increase in BB% doesn't seem like a great tradeoff. You'd have to run the numbers with some run values, but that's probably only a small net positive. So, once you regress to the mean a little bit, I'm pretty skeptical that high BB, low BABIP Joe Kelly, if he exists, is better than just average old Joe Kelly.
|
|
|
Post by jmei on Sept 1, 2015 11:16:53 GMT -5
I'll also note that walk rate and BABIP might be negatively correlated for all pitchers. If you're wild and throwing lots of balls out of the zone, the contact you give up should be worse-than-average contact. The difficulty is that wild pitchers are also more often behind in the count, and BABIP when behind in the count is higher.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,881
|
Post by ericmvan on Sept 2, 2015 15:52:13 GMT -5
The thread will have to wait a bit because it turns out that his two-seamer and four-seamer blend together in a way that makes it impossible for even the superior Pitch Info classification methodology to tell them apart. So I will either attempt to do so myself semi-manually, or just lump them together. In this case, to get a significant result despite the huge amount of BABIP noise in any individual start, you would need a large effect size. Now, his career BABIP as a starter is .293. But it's .264 in the 42 games where he walked 8% or more of batters (on 701 balls in play), and .333 in the 28 where he walked less than 8% (on 526). (Yes, his BABIP is an inverse function of his BB%.) The odds of getting the BABIP split at random are 1 in 113 (p = .009). You can also do a t-test on the unweighted means of the two groups of starts (which have identical variances), and p = .005. And I finally plotted BB% vs. BABIP and immediately saw that there was one huge outlier where he walked a career-high 25.0% and got killed (2 HR, .385 BABIP, versus Houston last 8/17). It's one of two games in his career that are huge xFIP and FIP outilers, games where he apparently had nothing and any wildness was going to just get him killed. (The other was his 7 R in 1.2 IP, 0 SO, 1 BB game this year in Minnesota.) Remove those two outliers, and you get p = .0017 for the inverse correlation, start by start. Alright, I grabbed the data myself. The average relationship (whether or not you include K%) is about 6 points of BABIP improvement for every extra 1 percentage point of BB%. But, of course, I think you'd agree that you have to regress that to the mean, since its highly, highly likely that at least some of this is caused by chance. I'd guess you have to regress 95%+ of the way to the mean, because I don't think this is a normal split to have, but let's say for now that it's mostly real. 6 points of BABIP of a 1% increase in BB% doesn't seem like a great tradeoff. You'd have to run the numbers with some run values, but that's probably only a small net positive. So, once you regress to the mean a little bit, I'm pretty skeptical that high BB, low BABIP Joe Kelly, if he exists, is better than just average old Joe Kelly. I'll also note that walk rate and BABIP might be negatively correlated for all pitchers. If you're wild and throwing lots of balls out of the zone, the contact you give up should be worse-than-average contact. The difficulty is that wild pitchers are also more often behind in the count, and BABIP when behind in the count is higher. The only way you get this negative correlation is if you have a BABIP skill and go back and forth between challenging hitters in the zone and nibbling. Pitchers who try to work the edges will have a higher walk rate when they are just mediocre at doing so successfully, but a lower walk rate when their control is at their best. Their BABIP may be low, but if anything, when they are wilder (as jmei points out) their BABIP will be higher because they'll miss more with pitches that get into the zone. So a guy who nibbles all the time isn't going to have this negative correlation and may have a slight positive one. However, if that same pitcher goes through a stretch where he's just challenging hitters in the zone without much success, he's going to have his lowest walk rate and his highest BABIP. And he'll have the inverse correlation. There are probably more pitchers who could have a BABIP skill if they needed to try to have one. Most guys with Kelly's stuff can work within the zone, keep their walks low, and keep their BABIP at league average. Kelly keep trying that for stretches, and then realizing it's not working, and then he goes back to nibbling, where his BABIP is a lot lower and his BB rate is higher, regardless of start-to-start variations in command. *** Now, what I've found out last night is that Kelly, throughout his career, has always messed significantly with his fastball %. I mean, he'll throw four or five games with every one at least FB% > x, and then the next four will be every game < X. Very clear-cut. His current great stretch is not the first time he's decided to throw significantly fewer fastballs and more offspeed stuff: it's the eighth. And these changes (including the six times he decided to switch to throwing more fastballs) have had a very mixed and frankly puzzling or bizarre pattern of success or lack of it. *** I'm just starting to look into that (including a breakdown of which off-speed pitch he favors, which he also seems to tinker with.) I have not yet, BTW, tried to correlate these changes in repertoire with the hypothesized challenge vs. nibble changes in strategy. I'm still thinking about how to quantify that and even what data I can use. Reasons to doubt Kelly this month is for real: everything between *** above. Reasons to think it may be for real: 1) This is the lowest FB% he's ever tried. He basically had three levels he'd switch between; this is below the previous lowest. And 2) He's been more consistent over this stretch of starts than any previous stretch. Right now, though, I would have to guess that's random. I'm going to start a thread for this with (for the time being) links to this discussion. If the mods can move all those posts into that thread, terrific!
|
|
|