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.
The Reliever Reinforcement Effect Hypothesis
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,941
|
Post by ericmvan on Mar 18, 2016 17:59:39 GMT -5
Why are teams like the Royals, Sox and Yankees going out of their way to build stupidly deep bullpens?
The Yankees already had a guy good enough to close pitching the 8th, but instead of trading him as redundant, they traded for a closer in order to bump their two incumbents to the 7th and 8th. The Sox looked like they needed a top set-up guy, but instead paid a disturbingly high price for one of the best closers in the game, bumping their very good-to-great closer to the 8th, and then they went out and traded for a guy good enough to close to pitch the 7th.
Weird. It can't be that they're blindly imitating the Royals. Is there a sound analytic reason why this strategy makes sense?
If there is such a reason, it explains why the price for Kimbrel was so high: there were lots of teams that had made the same analytic discovery, and lots of teams who wanted to acquire Kimbrel regardless of how good their bullpen already was.
Here's a reason that might be true. Finding evidence for it would require parsing the pitch/fx database at the pitch-by-pitch level, something every FO in MLB is set up to do but I am not. So here's the scouting version. I'll give you the narrative as it occurred to me.
After 2013 John Farrell was gracious enough to appear at the winter meeting of the local SABR chapter and do a long Q&A. One of the questions asked him (perhaps by me, but I think it may have been by someone else after we had talked about it beforehand) concerned Uehara's sudden dominance as a closer, and the obvious fact that it happened because he was throwing his deadly splitter much more often. Was that something the Sox had figured out he should have been doing all along?
Farrell said no. He said that pitching the 9th allowed Koji to throw the splitter more often, because hitters in the 9th in opposing-team save situations will be more aggressive chasing pitches outside the zone.
Well, I bought this, and still do. You might recall that the Sox under Valentine had an incredibly bad K/W ratio when hitting in the 9th when the other team was in a save situation, numbers so bad they beggared belief (they went months at one point between walks).
When we traded for Kimbrel, my first thought was, there goes Koji's ability to throw the splitter more often. But a few months later I realized that that might not be true at all.
Why are hitters more aggressive in the 9th when trailing by a run or two, hence making certain opposing pitchers more effective? (Assuming that is true, as Farrell claimed.) Because they know that if they do not score this inning, they lose the game.
Well, if you're facing Koji Uehara in the 8th and Craig Kimbrel is waiting to pitch the 9th, the same psychology holds. You're thinking, Jesus, we'd better score now. And without quite realizing it, you start hacking a bit, in desperation.
Furthermore, if Koji is as good as we hope, they'll be thinking that in the 7th, making Smith better. And if Smith is as good as we hope, they'll be thinking that in the 6th, making Tazawa better.
I looked for evidence that Tazawa had been better when setting up Uehara than otherwise and there was none. But that didn't surprise me, because there is so much innate variation in reliever performance. Bad outings can happen any time, and their random distribution will totally skew the results of any pair of samples. You have to look at the deepest possible data and break it down by inning and score.
I strongly suspect that multiple teams have done so and found the effect to be real.
|
|
|
Post by telson13 on Mar 18, 2016 22:41:37 GMT -5
Why are teams like the Royals, Sox and Yankees going out of their way to build stupidly deep bullpens? The Yankees already had a guy good enough to close pitching the 8th, but instead of trading him as redundant, they traded for a closer in order to bump their two incumbents to the 7th and 8th. The Sox looked like they needed a top set-up guy, but instead paid a disturbingly high price for one of the best closers in the game, bumping their very good-to-great closer to the 8th, and then they went out and traded for a guy good enough to close to pitch the 7th. Weird. It can't be that they're blindly imitating the Royals. Is there a sound analytic reason why this strategy makes sense? If there is such a reason, it explains why the price for Kimbrel was so high: there were lots of teams that had made the same analytic discovery, and lots of teams who wanted to acquire Kimbrel regardless of how good their bullpen already was.
Here's a reason that might be true. Finding evidence for it would require parsing the pitch/fx database at the pitch-by-pitch level, something every FO in MLB is set up to do but I am not. So here's the scouting version. I'll give you the narrative as it occurred to me. After 2013 John Farrell was gracious enough to appear at the winter meeting of the local SABR chapter and do a long Q&A. One of the questions asked him (perhaps by me, but I think it may have been by someone else after we had talked about it beforehand) concerned Uehara's sudden dominance as a closer, and the obvious fact that it happened because he was throwing his deadly splitter much more often. Was that something the Sox had figured out he should have been doing all along? Farrell said no. He said that pitching the 9th allowed Koji to throw the splitter more often, because hitters in the 9th in opposing-team save situations will be more aggressive chasing pitches outside the zone. Well, I bought this, and still do. You might recall that the Sox under Valentine had an incredibly bad K/W ratio when hitting in the 9th when the other team was in a save situation, numbers so bad they beggared belief (they went months at one point between walks). When we traded for Kimbrel, my first thought was, there goes Koji's ability to throw the splitter more often. But a few months later I realized that that might not be true at all. Why are hitters more aggressive in the 9th when trailing by a run or two, hence making certain opposing pitchers more effective? (Assuming that is true, as Farrell claimed.) Because they know that if they do not score this inning, they lose the game.
Well, if you're facing Koji Uehara in the 8th and Craig Kimbrel is waiting to pitch the 9th, the same psychology holds. You're thinking, Jesus, we'd better score now. And without quite realizing it, you start hacking a bit, in desperation. Furthermore, if Koji is as good as we hope, they'll be thinking that in the 7th, making Smith better. And if Smith is as good as we hope, they'll be thinking that in the 6th, making Tazawa better. I looked for evidence that Tazawa had been better when setting up Uehara than otherwise and there was none. But that didn't surprise me, because there is so much innate variation in reliever performance. Bad outings can happen any time, and their random distribution will totally skew the results of any pair of samples. You have to look at the deepest possible data and break it down by inning and score. I strongly suspect that multiple teams have done so and found the effect to be real. That's a fascinating hypothesis. It's also very believable. I'm not a big fan of the (mis-)use of large-scale surface data (by that I mean wholesale analysis of unselected sets, like in our previous thread about the value of global studies on second-half performance as a predictor of future performance, or your example here about Tazawa). And by that I mean to say, just as SSS is "positively" misleading, so too can large, unselected sets be "negatively" misleading. So I appreciate your point here. It certainly passes the often erroneous but wholly convincing eyeball test. And that's not a criticism at all. Astute observation is the cornerstone of any scientific analysis (as is selecting/designing your experiment, eg data set). Hitters genuinely do seem to get overanxious as the end nears, especially in close games. And the über-bullpen phenomenon does seem to cluster some unreal performances (last year's Yankees, Royals, etc.) beyond what you might expect. It may be even more important in today's free-swinging game, where run manufacturing is kind of a lost art. Very cool. FWIW, I hope that this means Matt Barnes gets a shot. His FB, at least in velocity/movement, is the real deal. That's five potentially outstanding relievers, an absolute lefty-murdering LOOGY, and a knuckleballing swing man. Acquiring Kimbrel and Smith made this bullpen incredible. Really, in terms of peak performance of the last three, it isn't all that far off NY. And depth-wise, maybe better.
|
|
|
Post by jimed14 on Mar 19, 2016 11:53:28 GMT -5
Is this such an subconscious effect that hitters cannot effectively alter their approach late in games?
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,941
|
Post by ericmvan on Mar 19, 2016 14:52:34 GMT -5
Is this such an subconscious effect that hitters cannot effectively alter their approach late in games? I think that when it becomes widely known, many of them will try to do exactly that. It may thus be a bit of bubble that will burst when players adapt. I would classify it as equivalent to the shift; hitters know they need to go the other way more often, it's easier said than done, but it probably is indeed happening. There is, BTW, a lot of evidence of long-standing maladaptive strategies. Hitters as a group are ridiculously good in SF situations. Last year in MLB: .248 / .306 / .398 bases empty .316 / .351 / .490 runner on 3rd with less than 2 out (9415 PA) I'm sure there are some teams that have told their pitchers do not pitch as to avoid giving up a fly ball in that situation, but the fact that they do, and become way too predictable, is clearly driving this.
|
|
|
Post by raftsox on Mar 19, 2016 19:22:59 GMT -5
I remember seeing somewhere recently that teams with dominant bullpens averaged a little over 2 wins more than their baseruns wins suggest they should. It could be noise, obviously, but it was pretty consistent. That might explain why teams are loading up on dominant relievers. Or it might have to do with teams respect the "third time through the order" starter penalty more.
|
|
|
Post by lronhoyabembe on Mar 20, 2016 10:49:42 GMT -5
There is, BTW, a lot of evidence of long-standing maladaptive strategies. Hitters as a group are ridiculously good in SF situations. Last year in MLB This stat sounds like circular logic to me. It's mostly saying that when pitchers are putting runners on base, hitters tend to get on base more often. Also, sacrifice flies don't count as at bats, so fly balls that would score runners only penalize batting average when there is not a runner in scoring position.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,941
|
Post by ericmvan on Mar 20, 2016 14:39:10 GMT -5
There is, BTW, a lot of evidence of long-standing maladaptive strategies. Hitters as a group are ridiculously good in SF situations. Last year in MLB This stat sounds like circular logic to me. It's mostly saying that when pitchers are putting runners on base, hitters tend to get on base more often. Also, sacrifice flies don't count as at bats, so fly balls that would score runners only penalize batting average when there is not a runner in scoring position. As you suggest, let's count SF as outs to get a better sense of the actual change in BA and SA. Let's also exclude IBB from OBP. And let's compare two situations where the pitcher would be struggling equally: man just on 2nd versus just on 3B, with 0 or 1 outs. The difference there is mostly leadoff double followed by an out that either moves the runner over or doesn't. There are also of course a bunch of miscellaneous event sequences. Since the difference between a 2B and 3B is more the player's speed and the ballpark than the hardness of contact, there's no appreciable difference between leadoff doubles and triples in terms of our expectation for the pitcher. Data is from last year, but it holds up year after year. Man on PA* BA* OB* SA* SO% BB% HRC BABIP XB% 2nd 8088 .257 .348 .405 .204 .093 .038 .306 .234 3rd 2309 .281 .388 .430 .194 .104 .031 .340 .250 -5% 12% -18% 11% 7% (PA* excludes SH and IBB.) When pitchers try to prevent the SF, they strike out batters 5% less often, walk them 12% more often, and lose .034 points of BABIP. An extra 7% of hits in play go for extra bases, too. Because they're limiting fly balls, this is countered by a nice drop in HR/Contact, but the net result is they lose .040 points of effective OBP and .025 points of effective SA. Preventing the SF obviously makes sense when it represents the tying or go-ahead run at the end of a game, but otherwise, pitchers would be better off just trying to get the batter out in general, like always.
|
|
|
Post by lronhoyabembe on Mar 28, 2016 20:41:51 GMT -5
Interesting, thank you for running the numbers. The assumptions you made seem fair enough, and it does look like there could be a statistically significant difference here.
|
|
|
Post by umassgrad2005 on Mar 28, 2016 23:10:48 GMT -5
Why are teams like the Royals, Sox and Yankees going out of their way to build stupidly deep bullpens? The Yankees already had a guy good enough to close pitching the 8th, but instead of trading him as redundant, they traded for a closer in order to bump their two incumbents to the 7th and 8th. The Sox looked like they needed a top set-up guy, but instead paid a disturbingly high price for one of the best closers in the game, bumping their very good-to-great closer to the 8th, and then they went out and traded for a guy good enough to close to pitch the 7th. Weird. It can't be that they're blindly imitating the Royals. Is there a sound analytic reason why this strategy makes sense? If there is such a reason, it explains why the price for Kimbrel was so high: there were lots of teams that had made the same analytic discovery, and lots of teams who wanted to acquire Kimbrel regardless of how good their bullpen already was.
Here's a reason that might be true. Finding evidence for it would require parsing the pitch/fx database at the pitch-by-pitch level, something every FO in MLB is set up to do but I am not. So here's the scouting version. I'll give you the narrative as it occurred to me. After 2013 John Farrell was gracious enough to appear at the winter meeting of the local SABR chapter and do a long Q&A. One of the questions asked him (perhaps by me, but I think it may have been by someone else after we had talked about it beforehand) concerned Uehara's sudden dominance as a closer, and the obvious fact that it happened because he was throwing his deadly splitter much more often. Was that something the Sox had figured out he should have been doing all along? Farrell said no. He said that pitching the 9th allowed Koji to throw the splitter more often, because hitters in the 9th in opposing-team save situations will be more aggressive chasing pitches outside the zone. Well, I bought this, and still do. You might recall that the Sox under Valentine had an incredibly bad K/W ratio when hitting in the 9th when the other team was in a save situation, numbers so bad they beggared belief (they went months at one point between walks). When we traded for Kimbrel, my first thought was, there goes Koji's ability to throw the splitter more often. But a few months later I realized that that might not be true at all. Why are hitters more aggressive in the 9th when trailing by a run or two, hence making certain opposing pitchers more effective? (Assuming that is true, as Farrell claimed.) Because they know that if they do not score this inning, they lose the game.
Well, if you're facing Koji Uehara in the 8th and Craig Kimbrel is waiting to pitch the 9th, the same psychology holds. You're thinking, Jesus, we'd better score now. And without quite realizing it, you start hacking a bit, in desperation. Furthermore, if Koji is as good as we hope, they'll be thinking that in the 7th, making Smith better. And if Smith is as good as we hope, they'll be thinking that in the 6th, making Tazawa better. I looked for evidence that Tazawa had been better when setting up Uehara than otherwise and there was none. But that didn't surprise me, because there is so much innate variation in reliever performance. Bad outings can happen any time, and their random distribution will totally skew the results of any pair of samples. You have to look at the deepest possible data and break it down by inning and score. I strongly suspect that multiple teams have done so and found the effect to be real. I just don't see it. You act like making Koji a closer suddenly made him a much better pitcher. Looking at career stats I just don't see it. Koji was very dominant before becoming a closer and the last 3 years are all within his career norms. As to why Yankees got chapman? Simple it was a classic buy low guy. Sure they didn't need him, but they couldn't turn down getting him for 25 cents on the dollar. Kimbrels's high value was simply supply and demand. There were no elite closers on free agent market. We got Kimbrel and Smith to improve a pen that was very bad last year.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,941
|
Post by ericmvan on Mar 28, 2016 23:47:59 GMT -5
Why are teams like the Royals, Sox and Yankees going out of their way to build stupidly deep bullpens? I strongly suspect that multiple teams have done so and found the effect to be real. I just don't see it. You act like making Koji a closer suddenly made him a much better pitcher. Looking at career stats I just don't see it. Koji was very dominant before becoming a closer and the last 3 years are all within his career norms. As to why Yankees got chapman? Simple it was a classic buy low guy. Sure they didn't need him, but they couldn't turn down getting him for 25 cents on the dollar. Kimbrels's high value was simply supply and demand. There were no elite closers on free agent market. We got Kimbrel and Smith to improve a pen that was very bad last year. Koji had a 2.31 ERA as a setup guy (2.36 from 2010-2012, and 2.10 in 30 IP with the Sox). He had a 10.7 K/W ratio, but gave up a HR to a healthy 5.8% of hitters who made contact. Hitters hit about .178 (counting SF as AB). In his first 76 innings as a closer for the Sox, he had an 0.47 ERA with a 16.8 K/W and a HR allowed to 1.9% of hitters making contact. Hitters hit .121. And he did that starting at age 38! Note that in 2014, he had a 4.41 ERA after June 16, when he clearly was not right, and it blew up his season line. But for his first year as a closer he was insanely better than as a setup guy. Since this was something that every Red Sox fan was noticing and going bonkers over at the time that it happened, I find your response, well, puzzling. (If you suffered a major blow to the head after that, I apologize.)
|
|
|
Post by umassgrad2005 on Mar 29, 2016 1:58:53 GMT -5
I just don't see it. You act like making Koji a closer suddenly made him a much better pitcher. Looking at career stats I just don't see it. Koji was very dominant before becoming a closer and the last 3 years are all within his career norms. As to why Yankees got chapman? Simple it was a classic buy low guy. Sure they didn't need him, but they couldn't turn down getting him for 25 cents on the dollar. Kimbrels's high value was simply supply and demand. There were no elite closers on free agent market. We got Kimbrel and Smith to improve a pen that was very bad last year. Koji had a 2.31 ERA as a setup guy (2.36 from 2010-2012, and 2.10 in 30 IP with the Sox). He had a 10.7 K/W ratio, but gave up a HR to a healthy 5.8% of hitters who made contact. Hitters hit about .178 (counting SF as AB). In his first 76 innings as a closer for the Sox, he had an 0.47 ERA with a 16.8 K/W and a HR allowed to 1.9% of hitters making contact. Hitters hit .121. And he did that starting at age 38! Note that in 2014, he had a 4.41 ERA after June 16, when he clearly was not right, and it blew up his season line. But for his first year as a closer he was insanely better than as a setup guy. Since this was something that every Red Sox fan was noticing and going bonkers over at the time that it happened, I find your response, well, puzzling. (If you suffered a major blow to the head after that, I apologize.) You can't just look at first 76 innings and overlook the last two years. Looking at the stats you could easily just say Koji had a career year, it happens with players all the time. He was also healthy for a full year which he has struggled with for his career. If becoming a closer makes a set up guy insanley better please provide a bunch of examples rather then only one out of three seasons for Koji. If you have a problem with that logic you are the one that might of had a major blow to the head. What's your theory on Brady Anderson and his fluke 50 HR season?
|
|
|
Post by jodyreidnichols on Mar 29, 2016 6:55:49 GMT -5
Why are teams like the Royals, Sox and Yankees going out of their way to build stupidly deep bullpens? The Yankees already had a guy good enough to close pitching the 8th, but instead of trading him as redundant, they traded for a closer in order to bump their two incumbents to the 7th and 8th. The Sox looked like they needed a top set-up guy, but instead paid a disturbingly high price for one of the best closers in the game, bumping their very good-to-great closer to the 8th, and then they went out and traded for a guy good enough to close to pitch the 7th. Weird. It can't be that they're blindly imitating the Royals. Is there a sound analytic reason why this strategy makes sense? If there is such a reason, it explains why the price for Kimbrel was so high: there were lots of teams that had made the same analytic discovery, and lots of teams who wanted to acquire Kimbrel regardless of how good their bullpen already was.
Here's a reason that might be true. Finding evidence for it would require parsing the pitch/fx database at the pitch-by-pitch level, something every FO in MLB is set up to do but I am not. So here's the scouting version. I'll give you the narrative as it occurred to me. After 2013 John Farrell was gracious enough to appear at the winter meeting of the local SABR chapter and do a long Q&A. One of the questions asked him (perhaps by me, but I think it may have been by someone else after we had talked about it beforehand) concerned Uehara's sudden dominance as a closer, and the obvious fact that it happened because he was throwing his deadly splitter much more often. Was that something the Sox had figured out he should have been doing all along? Farrell said no. He said that pitching the 9th allowed Koji to throw the splitter more often, because hitters in the 9th in opposing-team save situations will be more aggressive chasing pitches outside the zone. Well, I bought this, and still do. You might recall that the Sox under Valentine had an incredibly bad K/W ratio when hitting in the 9th when the other team was in a save situation, numbers so bad they beggared belief (they went months at one point between walks). When we traded for Kimbrel, my first thought was, there goes Koji's ability to throw the splitter more often. But a few months later I realized that that might not be true at all. Why are hitters more aggressive in the 9th when trailing by a run or two, hence making certain opposing pitchers more effective? (Assuming that is true, as Farrell claimed.) Because they know that if they do not score this inning, they lose the game.
Well, if you're facing Koji Uehara in the 8th and Craig Kimbrel is waiting to pitch the 9th, the same psychology holds. You're thinking, Jesus, we'd better score now. And without quite realizing it, you start hacking a bit, in desperation. Furthermore, if Koji is as good as we hope, they'll be thinking that in the 7th, making Smith better. And if Smith is as good as we hope, they'll be thinking that in the 6th, making Tazawa better. I looked for evidence that Tazawa had been better when setting up Uehara than otherwise and there was none. But that didn't surprise me, because there is so much innate variation in reliever performance. Bad outings can happen any time, and their random distribution will totally skew the results of any pair of samples. You have to look at the deepest possible data and break it down by inning and score. I strongly suspect that multiple teams have done so and found the effect to be real. Presented as it is, this seems easy to buy into, aka it has a certain logic to it. However while it likely is true for the average hitter, I suspect it does not apply to the above average patient hitter which I typically think of as hitters whose OBP is .070 higher than their respective batting average. I noticed over the years that any hitter considered a superstar almost without exception has a difference of .070 between the two. Good hitters typically have a difference of .050, although that is less rigid. What makes batters such as Votto, Boggs ect as good as they are is the commitment to their strike zone. It amazes me how many fans think he should expand his zone to get more RBi's. It leads me to believe that they simply do not understand the game. But I digress, I believe the premise you laid out is typically true.
|
|
|
Post by deepjohn on Mar 29, 2016 8:04:00 GMT -5
From looking around a little, it seems that the occam's razor explanation may be that, under Dombrowski and the new front office, the Red Sox are using some variation of K% minus BB%.
Kimbrel > Carson Smith > Darren O'Day > Koji Uehara
Chris Sale > James Shields >> Miley/Porcello/Kelly
This could fit with your theory that K% - BB% improves when batters start hacking under pressure?
|
|
|
Post by mgoetze on Mar 29, 2016 8:10:57 GMT -5
From looking around a little, it seems that the occam's razor explanation may be that, under Dombrowski and the new front office, the Red Sox are using some variation of K% minus BB%. How does this fit your theory that the Red Sox front office uses super-advanced statistics and theories that mere mortals like us cannot even begin to comprehend? K%-BB%, while excellent (arguably better than FIP), is not exactly rocket science.
|
|
|
Post by deepjohn on Mar 29, 2016 8:31:54 GMT -5
I think the premise of Eric's post is that we may not be understanding why the front office put such a high value on Kimbrel. And by extension, they may also be valuing Shields more than some here do.
It need not be rocket science (and rocket science is not really very complicated once you understand it), as much as something that some here may be overlooking. Have you posted on this forum about K%-B%? If you did, I missed it.
|
|
|
Post by mgoetze on Mar 29, 2016 8:36:26 GMT -5
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,941
|
Post by ericmvan on Mar 29, 2016 11:17:42 GMT -5
Koji had a 2.31 ERA as a setup guy (2.36 from 2010-2012, and 2.10 in 30 IP with the Sox). He had a 10.7 K/W ratio, but gave up a HR to a healthy 5.8% of hitters who made contact. Hitters hit about .178 (counting SF as AB). In his first 76 innings as a closer for the Sox, he had an 0.47 ERA with a 16.8 K/W and a HR allowed to 1.9% of hitters making contact. Hitters hit .121. And he did that starting at age 38! Note that in 2014, he had a 4.41 ERA after June 16, when he clearly was not right, and it blew up his season line. But for his first year as a closer he was insanely better than as a setup guy. Since this was something that every Red Sox fan was noticing and going bonkers over at the time that it happened, I find your response, well, puzzling. (If you suffered a major blow to the head after that, I apologize.) You can't just look at first 76 innings and overlook the last two years. Looking at the stats you could easily just say Koji had a career year, it happens with players all the time. He was also healthy for a full year which he has struggled with for his career. If becoming a closer makes a set up guy insanley better please provide a bunch of examples rather then only one out of three seasons for Koji. If you have a problem with that logic you are the one that might of had a major blow to the head. What's your theory on Brady Anderson and his fluke 50 HR season? 1) You can look at any sample size and say, that's unlikely to be random, if the effect is extreme enough. This effect was enormous. 2) Koji actually did not have a huge difference in the effectiveness of his individual pitches. He just threw the splitter, his better pitch, much more often as a closer. It was Sox management's opinion that he could do do and not have it become less effective, when closing. IOW, the closing situation alters the game-theory devised optimum pitch mix, and makes him a better pitcher. 3) It's not any pitcher who benefits; it's guys who have an out pitch that hitters chase out of the zone. The first guy I thought of when I typed that was Brad Lidge. 3.60, .202 / .308 / .323, .283 K rate, age-26 rookie season as setup guy 2.45, .182 / .264 / .308, .398 K, through June 20 the next year as setup 1.49, .168 / .246 / .276, .456 K as closer the rest of the year We don't have pitch/fx data for that era, but it wouldn't surprise me if Lidge in '04 earned the closer job by throwing his slider more, and once he got it, ramped up the frequency even further. It's in the running for best slider of all time, after all. (As an aside, one of my favorite and proudest moment as a baseball fan was at the SABR conference in Philly, where Lidge was on a panel. During the Q&A, I asked him about the '05 All-Star game, where he struck out the side in the 7th on 11 pitches, 8 of the 9 strikes swinging, but it all happened in a box on the screen, and Lidge was not identified by the announcers until the end of the inning, because they were interviewing Kenny Rogers, who had just pitched for the AL, about his ongoing pine-tar accusations. I thought that was the all-time low point for baseball TV journalism. I asked Lidge how he found out that this historically great performance (most impressive in the AS game since Pedro in '99) had barely been broadcast. His answer: he was in the clubhouse afterwards, where of course the game was on TV, and Roger Clemens, who had pitched the 5th, came up to him and asked, "hey, did you throw?" ) I think there would be plenty of evidence for this if you looked into the data. Someone else can do that. I'm going to try to finish my DRS projection system over the next week or so and then call it quits on analytic projects until the end of 2017. Re Anderson, I don't think it's a stretch to guess that he juiced for a year and didn't like the side effects. There were certainly rumors to that effect. And I know that there have been other fluke seasons you can point to, and my answer would be that, if the season was a big enough outlier, the chances are that it was caused by something real. We now know that Norm Cash in '61 was corking his bat, for instance. You could measure the improbability of every season, and you'll find more improbable ones than chance would dictate, and maybe a lot more. You couldn't separate the real ones from the flukes, but you'd have a good idea of the proportion. The most common cause for non-fluke outlier seasons is probably a real effect that the player declined to duplicate because it was too much work or had undesirable consequences. The term for what you're doing here, BTW, is being Statistically Correct. All variation is not random unless proven otherwise to a degree that would satisfy a criminal justice system. The real versus random question is incredibly important, and there is no benefit to erring on the side of caution and being overly skeptical about the real. You have to get it right. And that means the preponderance of the evidence, not an argument that you could publish in a scientific journal. The penalty for declaring a false negative can be immense -- imagine if the Blue Jays had decreed that Jose Bautista's 2010 was a fluke and sold "high" on him. Obviously, they had scouting reasons to know it was real. But there's no way that any statistician could have proven that it was real and not random (although a good one could have demonstrated how improbable it was), and I bet there was at least one analyst online urging the Jays to dump him.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,941
|
Post by ericmvan on Mar 29, 2016 11:20:54 GMT -5
Why are teams like the Royals, Sox and Yankees going out of their way to build stupidly deep bullpens? I strongly suspect that multiple teams have done so and found the effect to be real. Presented as it is, this seems easy to buy into, aka it has a certain logic to it. However while it likely is true for the average hitter, I suspect it does not apply to the above average patient hitter which I typically think of as hitters whose OBP is .070 higher than their respective batting average. I noticed over the years that any hitter considered a superstar almost without exception has a difference of .070 between the two. Good hitters typically have a difference of .050, although that is less rigid. What makes batters such as Votto, Boggs ect as good as they are is the commitment to their strike zone. It amazes me how many fans think he should expand his zone to get more RBi's. It leads me to believe that they simply do not understand the game. But I digress, I believe the premise you laid out is typically true. Very interesting. Post-season HRs to backup catchers notwithstanding, Koji has always been a guy who had what I call a "steep opponent [quality] slope," that is, he's been much tougher on weak hitters than elite ones. That fits with your observation perfectly.
|
|
|
Post by mgoetze on Mar 29, 2016 11:23:00 GMT -5
2) Koji actually did not have a huge difference in the effectiveness of his individual pitches. He just threw the splitter, his better pitch, much more often as a closer. It was Sox management's opinion that he could do do and not have it become less effective, when closing. IOW, the closing situation alters the game-theory devised optimum pitch mix, and makes him a better pitcher. Yes, but this is an experiment without a control group.
|
|
|
Post by jmei on Mar 29, 2016 11:27:43 GMT -5
This stat sounds like circular logic to me. It's mostly saying that when pitchers are putting runners on base, hitters tend to get on base more often. Also, sacrifice flies don't count as at bats, so fly balls that would score runners only penalize batting average when there is not a runner in scoring position. As you suggest, let's count SF as outs to get a better sense of the actual change in BA and SA. Let's also exclude IBB from OBP. And let's compare two situations where the pitcher would be struggling equally: man just on 2nd versus just on 3B, with 0 or 1 outs. The difference there is mostly leadoff double followed by an out that either moves the runner over or doesn't. There are also of course a bunch of miscellaneous event sequences. Since the difference between a 2B and 3B is more the player's speed and the ballpark than the hardness of contact, there's no appreciable difference between leadoff doubles and triples in terms of our expectation for the pitcher. Data is from last year, but it holds up year after year. Man on PA* BA* OB* SA* SO% BB% HRC BABIP XB% 2nd 8088 .257 .348 .405 .204 .093 .038 .306 .234 3rd 2309 .281 .388 .430 .194 .104 .031 .340 .250 -5% 12% -18% 11% 7% (PA* excludes SH and IBB.) When pitchers try to prevent the SF, they strike out batters 5% less often, walk them 12% more often, and lose .034 points of BABIP. An extra 7% of hits in play go for extra bases, too. Because they're limiting fly balls, this is countered by a nice drop in HR/Contact, but the net result is they lose .040 points of effective OBP and .025 points of effective SA. Preventing the SF obviously makes sense when it represents the tying or go-ahead run at the end of a game, but otherwise, pitchers would be better off just trying to get the batter out in general, like always. I think a significant chunk of the higher BABIP and at least some of the higher extra-base hit percentage with a runner on third is because infield defenses will play in with a runner on third and no/one out, whereas they wouldn't with a runner on second.
|
|
|
Post by mgoetze on Mar 29, 2016 11:35:37 GMT -5
infield defenses will play in with a runner on third and no/one out, A tactic that, incidentally, I suspect managers use too liberally.
|
|
|
Post by jmei on Mar 29, 2016 11:40:02 GMT -5
More generally, I think the inability of analysts to find evidence that players play better or worse in clutch situations generally cuts against this theory. For instance, I'm not aware of any analysis which shows that starting pitchers can "pitch to the score" (in other words, that their performance varies depending on the leverage of the situation) or that closers pitch better or worse in save situations than they do in non-save situations, and there is plenty of analysis which cuts the other way. Similar analysis has been performed on hitters, with similar results.
Considering the above, unless you can find real "smoking gun"-type data, I'm inclined to think that this is an interesting theory but no more than that.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,941
|
Post by ericmvan on Apr 3, 2016 23:42:50 GMT -5
More generally, I think the inability of analysts to find evidence that players play better or worse in clutch situations generally cuts against this theory. For instance, I'm not aware of any analysis which shows that starting pitchers can "pitch to the score" (in other words, that their performance varies depending on the leverage of the situation) or that closers pitch better or worse in save situations than they do in non-save situations, and there is plenty of analysis which cuts the other way. Similar analysis has been performed on hitters, with similar results. Considering the above, unless you can find real "smoking gun"-type data, I'm inclined to think that this is an interesting theory but no more than that. This is fundamentally asymmetrical. No one has really studied un-clutch performance, except as an adjunct to clutch, in very broad studies. Indeed, it makes little psychological sense that the average player could be better in high-leverage situations than in ordinary ones -- you immediately run into the "why isn't he that good all of the time" question. The extra adrenaline tends to make the nervous system noisier. It's far more credible that there are certain types of high-leverage situations where a subset of players perform sub-optimally, and largely because of a change in approach rather than any "clutch" or "choke" factor.. Elsewhere, I posted evidence that Pedroia is too aggressive in high-leverage situations when anyone other than Ortiz is the next hitter. The splits in hitting rates were small but collectively significant, but the split in WPA relative to OPS was massive, suggesting that all of the difference happened when the game was on the line. And that fits with what I remember about his performance perfectly. That's the sort of effect I'm hypothesizing here -- one too narrow to be discovered without a lot of work. My guess is that it's very clear if you look at every pitch with pitch/fx, and quite possibly you need hit/fx data as well. I don't have either databse, but MLB teams do, and they seem to be acting as if this hypothesis is true.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,941
|
Post by ericmvan on Apr 4, 2016 0:24:27 GMT -5
2) Koji actually did not have a huge difference in the effectiveness of his individual pitches. He just threw the splitter, his better pitch, much more often as a closer. It was Sox management's opinion that he could do do and not have it become less effective, when closing. IOW, the closing situation alters the game-theory devised optimum pitch mix, and makes him a better pitcher. Yes, but this is an experiment without a control group. Only half-true. Yes, Koji did not try to close while throwing his splitter at the old set-up rate. I don't think that would have gotten past the ethics committee. However, it's almost certainly the case that his splitter frequency while setting up was optimized by Koji and his catchers via trial and error, in accordance with game theory. If your best pitch is an off-speed pitch, you try to throw it as often as possible, but not so often that it becomes too predictable. The way you do that is to throw it more often and back off when you start to see diminishing returns. So there was presumably a small data set that showed that throwing the splitter at something approaching or approximating the closer rate while setting up was not only not more effective, it was less effective. Note that Farrell and his staff believed both that Koji's splitter frequency as a set-up guys was optimal, and that he could ramp it up as a closer, with predictable positive results. Which is what happened. Were they somehow (along with all his previous teams) just stupid about his optimal splitter rate in general? That seems unlikely. While it's credible that only the Sox had the idea that closing would change the optimal rate in a positive way.
|
|
|
Post by mgoetze on Apr 4, 2016 3:08:33 GMT -5
Yes, but this is an experiment without a control group. Only half-true. Yes, Koji did not try to close while throwing his splitter at the old set-up rate. I don't think that would have gotten past the ethics committee. However, it's almost certainly the case that his splitter frequency while setting up was optimized by Koji and his catchers via trial and error, in accordance with game theory. If your best pitch is an off-speed pitch, you try to throw it as often as possible, but not so often that it becomes too predictable. The way you do that is to throw it more often and back off when you start to see diminishing returns. So there was presumably a small data set that showed that throwing the splitter at something approaching or approximating the closer rate while setting up was not only not more effective, it was less effective. The data set, if it exists at all, is surely such a small sample as to be completely meaningless. The analysis of any such data set was surely rife with confirmation bias. Yes, you have mentioned this already. Farrell is no scientist, of that I am certain.
|
|
|