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.
|
Post by philsbosoxfan on Sept 26, 2020 23:10:30 GMT -5
Yeah, I can't imagine a manager making a decision based on something like that. Game 7 of the World Series is the day after Game 6 of the World Series, but oh no there's a rainout in between so we cannot play Chavis because he had a day off. And yes, I'm being facetious here, but rainouts will happen, the built in day offs you alluded to. I can agree with the larger point of if you keep him on the bench he can get rusty and lose his rhythm kind of argument. I think that's one of the reasons Cora gave for giving Hanley the heave ho, that if he was going to spend a bunch of time on the bench he wasn't going to be the kind of player who can have that kind of time off, roll off the bench and suddenly contribute. Yeah, I can't imagine a manager making a decision based on something like that.
I gave Terry Francona that sort of advice (through Jed Hoyer and Zack Scott) and he used it.
And the way you use this particular data is to never give a guy a day off until he hits his fatigue point. For your counter-argument you picked the only possible situation -- rainout before a single game with an off day following -- where you might choose to do the opposite and sit him. Players do not have a problem seeing their name in the lineup more often. Manager do not have a problem with not having to decide who plays a given position today.
And the point here is not to maximize the lineup for a given game, but in the long run. The bad numbers on the 20+ unavoidable first days are the price you pay for the much better numbers for all the others.
The goal is to get the total best out of a player. And you always follow that principle. You sit a guy who needs a day off, even if it hurts your odds that day, if in the long run he'll be better. Francona used to do that and it would drive folks crazy, but it was smart. If Zimmer had done that in '78, we would have won the WES.
There was an infamous situation (mentioned in Francona's autobiography, but attributed to the wrong year!) where Zack Scott advised Francona to sit Mike Lowell against Chien-Ming Wang because the swing-path analytics had him as terrible against him. That was the point where Tito stopped taking analytic lineup advice.
I was doing my own lineup suggestions, and I could see that Lowell couldn't hit Wang. My advice for that series was to platoon at SS but that "nobody else is going to want to sit against the Yankees." Insulting Lowell's competitive instincts would do more damage than starting his backup would gain you.
Hmm ... please explain the downside to playing Michael Chavis every single game (up to 11 in a row) if that's what it takes to turn him into an 850+ OPS hitter.
If you do that, and it turns out that he's not that good, how many games are you losing versus an Arroyo / Chavis platoon? One? That's the max, probably. The upside is you've discovered a star, and the stats so far point at the upside. Really, the only problem I have here is that it's one thing to change how you use a player mid season, quite another when you are constructing a roster for the next year. Also, he's not doing well in this final test run but is striking out less. It seems unlikely that the Sox will go status quo at second base and he's not a very good left fielder albeit with limited experience..
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,924
|
Post by ericmvan on Sept 27, 2020 13:52:47 GMT -5
It's really hard to disagree with facts, and this phenomenon does show up for certain players, most notably with Manny Ramirez, over a huge amount of PA across his career. Sure, the average player just hits better as a generic regular than coming off the bench. But there are players who really do reset after a day off.
And as I pointed out, there's nothing weird or counter-intuitive about that except the severity of the effect. And any time that people do a thing, there will be some people who really do that thing, to a degree that might seem startling. Just look at the distribution of post counts here!
In Chavis's case, this year, when BP on a day off likely never happens, he already has ridiculous, statistically significant splits between day off and no day off the day before. A .292 OPS in 63 PA vs. .976 in 72 PA, last time I checked.
So if his regular playing time is 2 games out of 3, that's not going to tell you jack about how good he might be if he plays every day.
And yes, you can't have him play on an off day, but you can give him enhanced BP on an off day whenever possible to minimize the effect. And it changes the way you use him.
I never found any guy who had a significant effect depending on days off in a row. There's almost never a sufficient sample of the different values, to begin with.
Manny played a massive amount of games. How many samples do you have of Manny going long stretches playing limited games? You play 150 plus games and those days off mean one thing, you play 75 games and they can mean something completely different. Manny's numbers over the course of his career got better with every single day he had behind him (in a row). Each of these buckets was a season's worth of data.
Single days off happen all the time. They always reset him to being a merely good hitter (on average, over his whole career up to the time I did the study). Then he'd be better the next day, and even better the day after that, so that by the time he was playing for the 12th straight day, he was as good as anyone who ever lived. And then one day off would reset him. (His numbers did decline from too many days in a row, and it happened sooner at home than on the road.)
You're looking at the general notion of playing time frequency, which is Chris's lens. But it's wrong, at least for this guy.
And from what I know of neuroscience, you either have the ability to come off the bench and hit, or you don't. Either your "muscle memory" starts to fade from a day off, or it doesn't. If it does, it's really hard to see why the second day off would be a bigger decline than the first. If it did, if extra days off made the loss of "muscle memory" of your swing increasingly more distant, most players would be awful after a few days off
And there's an obvious limit as to how much of the memory you lose; it's just the fine-tuning of it. It's hard to see why you'd losing fine-tuning of your swing and timing from a third day off after two. And there's no evidence that getting 3 days off is worse than 2. I'm pretty sure there are guys for whom 2 days off can be a bit worse worse than 1, but given the rarity of the players and of the circumstances, it would be hard to find that in the data.
This is why managers sometimes give a guy the day off before a scheduled off day, to give him a 2-day rest from fatigue. He's needed a break, but you didn't want to reset his muscle memory with a day off. But now his timing's going to be a bit off anyway, and the extra day's not going to make it worse.
Another data point: the lack of a strong correlation between pinch-hitting skill and overall skill. Think Rick Miller, but there are many others.
|
|
|
Post by Chris Hatfield on Sept 27, 2020 17:13:09 GMT -5
Which brings to mind something that Gabe Kapler wrote about before he became a manager - the uselessness of BP in the U.S. I get that those "easy" reps can be useful for certain things, but why wouldn't you try to replicate game conditions more and face "real" pitching? Japanese teams have BP pitchers on payroll who can come in and throw hard, throw breaking balls, etc. as requested. Wouldn't that help with the whole "muscle memory" thing? It seems so easy to fix. I'm sure there are certain players where it's an in-game, adrenaline, etc. thing, but I'm sure it's not all of them.
|
|
|
Post by coke0myfavdrink on Sept 27, 2020 18:20:35 GMT -5
Which brings to mind something that Gabe Kapler wrote about before he became a manager - the uselessness of BP in the U.S. I get that those "easy" reps can be useful for certain things, but why wouldn't you try to replicate game conditions more and face "real" pitching? Japanese teams have BP pitchers on payroll who can come in and throw hard, throw breaking balls, etc. as requested. Wouldn't that help with the whole "muscle memory" thing? It seems so easy to fix. I'm sure there are certain players where it's an in-game, adrenaline, etc. thing, but I'm sure it's not all of them. I like this idea. Maybe hire a catcher and a shagger or two also and have a run through of the lineup before the game.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,924
|
Post by ericmvan on Sept 28, 2020 0:16:06 GMT -5
Which brings to mind something that Gabe Kapler wrote about before he became a manager - the uselessness of BP in the U.S. I get that those "easy" reps can be useful for certain things, but why wouldn't you try to replicate game conditions more and face "real" pitching? Japanese teams have BP pitchers on payroll who can come in and throw hard, throw breaking balls, etc. as requested. Wouldn't that help with the whole "muscle memory" thing? It seems so easy to fix. I'm sure there are certain players where it's an in-game, adrenaline, etc. thing, but I'm sure it's not all of them. Absolutely. I really want to do those platoon splits now.
|
|
|
Post by umassgrad2005 on Sept 29, 2020 15:06:49 GMT -5
Manny played a massive amount of games. How many samples do you have of Manny going long stretches playing limited games? You play 150 plus games and those days off mean one thing, you play 75 games and they can mean something completely different. Manny's numbers over the course of his career got better with every single day he had behind him (in a row). Each of these buckets was a season's worth of data.
Single days off happen all the time. They always reset him to being a merely good hitter (on average, over his whole career up to the time I did the study). Then he'd be better the next day, and even better the day after that, so that by the time he was playing for the 12th straight day, he was as good as anyone who ever lived. And then one day off would reset him. (His numbers did decline from too many days in a row, and it happened sooner at home than on the road.)
You're looking at the general notion of playing time frequency, which is Chris's lens. But it's wrong, at least for this guy.
And from what I know of neuroscience, you either have the ability to come off the bench and hit, or you don't. Either your "muscle memory" starts to fade from a day off, or it doesn't. If it does, it's really hard to see why the second day off would be a bigger decline than the first. If it did, if extra days off made the loss of "muscle memory" of your swing increasingly more distant, most players would be awful after a few days off
And there's an obvious limit as to how much of the memory you lose; it's just the fine-tuning of it. It's hard to see why you'd losing fine-tuning of your swing and timing from a third day off after two. And there's no evidence that getting 3 days off is worse than 2. I'm pretty sure there are guys for whom 2 days off can be a bit worse worse than 1, but given the rarity of the players and of the circumstances, it would be hard to find that in the data.
This is why managers sometimes give a guy the day off before a scheduled off day, to give him a 2-day rest from fatigue. He's needed a break, but you didn't want to reset his muscle memory with a day off. But now his timing's going to be a bit off anyway, and the extra day's not going to make it worse.
Another data point: the lack of a strong correlation between pinch-hitting skill and overall skill. Think Rick Miller, but there are many others.
When you were with the Red Sox what did you guys do to adjust numbers for variation and small sample sizes? Did you guys do things like that? Did you just look at averages? Did you dig deeper to see if it something else? I ask because looking at Manny's first season with the Red Sox he didn't have many games off. You'd be comparing like 142 games to like 10. Which isn't the same as comparing 76 to 76. If you start breaking that down into first game back, then second game back, etc. You are getting a bunch of small sample sizes and baseball players, even the great Manny have a ton of variances in his numbers in small sample sizes. An example Manny is Manny 7 out 10 games after day off, yet bad in three. On average that could make him average, yet that doesn't tell the whole story right? Small sample sizes are crazy hard in Baseball and averages can hide things. Like how far down the rabbit hole did you guys go? Like teams he played, pitchers he faced, ball park effect, was he coming off an injury?, did he just run into a hot pitcher, did he party more on off days, etc?
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,924
|
Post by ericmvan on Sept 29, 2020 16:13:39 GMT -5
Manny's numbers over the course of his career got better with every single day he had behind him (in a row). Each of these buckets was a season's worth of data.
Single days off happen all the time. They always reset him to being a merely good hitter (on average, over his whole career up to the time I did the study). Then he'd be better the next day, and even better the day after that, so that by the time he was playing for the 12th straight day, he was as good as anyone who ever lived. And then one day off would reset him. (His numbers did decline from too many days in a row, and it happened sooner at home than on the road.)
You're looking at the general notion of playing time frequency, which is Chris's lens. But it's wrong, at least for this guy.
And from what I know of neuroscience, you either have the ability to come off the bench and hit, or you don't. Either your "muscle memory" starts to fade from a day off, or it doesn't. If it does, it's really hard to see why the second day off would be a bigger decline than the first. If it did, if extra days off made the loss of "muscle memory" of your swing increasingly more distant, most players would be awful after a few days off
And there's an obvious limit as to how much of the memory you lose; it's just the fine-tuning of it. It's hard to see why you'd losing fine-tuning of your swing and timing from a third day off after two. And there's no evidence that getting 3 days off is worse than 2. I'm pretty sure there are guys for whom 2 days off can be a bit worse worse than 1, but given the rarity of the players and of the circumstances, it would be hard to find that in the data.
This is why managers sometimes give a guy the day off before a scheduled off day, to give him a 2-day rest from fatigue. He's needed a break, but you didn't want to reset his muscle memory with a day off. But now his timing's going to be a bit off anyway, and the extra day's not going to make it worse.
Another data point: the lack of a strong correlation between pinch-hitting skill and overall skill. Think Rick Miller, but there are many others.
When you were with the Red Sox what did you guys do to adjust numbers for variation and small sample sizes? Did you guys do things like that? Did you just look at averages? Did you dig deeper to see if it something else? I ask because looking at Manny's first season with the Red Sox he didn't have many games off. You'd be comparing like 142 games to like 10. Which isn't the same as comparing 76 to 76. If you start breaking that down into first game back, then second game back, etc. You are getting a bunch of small sample sizes and baseball players, even the great Manny have a ton of variances in his numbers in small sample sizes. An example Manny is Manny 7 out 10 games after day off, yet bad in three. On average that could make him average, yet that doesn't tell the whole story right? Small sample sizes are crazy hard in Baseball and averages can hide things. Like how far down the rabbit hole did you guys go? Like teams he played, pitchers he faced, ball park effect, was he coming off an injury?, did he just run into a hot pitcher, did he party more on off days, etc? There was no "you guys"; Bill James, Tom Tippett and I (and who knows who else!) worked separately. They liked to compare the different takes on things. I think I mentioned that I gave a green light to Jed Lowrie's disappointing junior year while Bill thought it was a red flag. So that's the only acquisition I may have influenced (aside from Carlos Pena, which was apparently pretty much me, and who got squeezed off the 40-man within months).
The usage splits (rest / consistency / travel) were almost always done over the course of a career. So all those confounds you identify so well melt away in the aggregate!
I think the only time I looked at a single season was to try and figure out Pedroia's rookie season and his terrible numbers on the road. I gave advice that apparently got passed down and taken, as the split disappearred the next year and the comments he made about the improvement matched the advice pretty strongly. But with a small data set, there has to be both a huge split, then it needs to have an explanation that makes easy sense, and then that explanation has to predict a further split you hadn't looked at. As in, if I'm right about this, then these games belong in the other bucket, and if I move them there the split should get bigger. And they do.
There are a lot of splits that have big Y2Y variations so big that you have to look at the whole career; looking at a single season will just drive you bonkers (performance against LHP by righty hitters is one of these). Another one I used with the Sox is pitcher performance by batting order position as a proxy for opponent hitter quality (that one actually helped got me hired, because I used it to predict we'd' annihilate the Cardinals starter, 105 wins notwithstanding. Every one of their guys had feasted on weak hitters). A third is the challenge / pitch-around splits for hitters that I developed to tweak batting orders -- although that came too late to be really used with the Sox; I talked about it with Zaxk Scott the winter before (while?) the economy tanked and all of the consultants got laid off.
I was well aware of small samples. My system for predicting batter / pitcher matchups, which looked for types of pitchers that a guy could hit, had a ridiculously primitive regression to the mean for the individual matchups when they were less than 32 PA, a number I came up with empirically. Less than 8 PA was meaningless -- except that hitters differed in how much they struggled in general, against all pitchers, in the first 7. So if there were a guy who projected to hit well against a pitcher because of his type (e.g., flyball pitcher who gave up a lot of homers relative to expectations based on K, BB, and FB / GB) and he was always helpless in his first 7 PA but had actually hit this guy the first time he faced him, that combo created a big green light.
(That example, BTW, is not hypothetical -- that was Doug Mirabelli against Mike Maroth on 5/3/05, where I convinced Jed to talk Tito into sitting Papi and having Mirabelli DH. Who then hit a slam off of Maroth in a 5-3 victory.)
(It goes without saying that when there was an 8+ PA sample in a head-to-head matchup, you looked at every game. I would look, obviously, for cheap hits and line drive outs, but the big thing was performance relative to teammates. Sometimes a guy who "has really hit this guy well" is a guy who got lucky and only faced him when he had an off night. And vice versa. The message here is that most of what the Statistically Correct regard as mere random variation is real differences that we don't measure well. )
|
|
|
Post by orion09 on Sept 29, 2020 19:19:05 GMT -5
I think the only time I looked at a single season was to try and figure out Pedroia's rookie season and his terrible numbers on the road. I gave advice that apparently got passed down and taken, as the split disappearred the next year and the comments he made about the improvement matched the advice pretty strongly. What was the advice given about Pedroia?
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,924
|
Post by ericmvan on Sept 29, 2020 23:03:49 GMT -5
I think the only time I looked at a single season was to try and figure out Pedroia's rookie season and his terrible numbers on the road. I gave advice that apparently got passed down and taken, as the split disappearred the next year and the comments he made about the improvement matched the advice pretty strongly. What was the advice given about Pedroia? Basically to get a better night's sleep. Google "Eric Van ESPN Stats Revolution" for a similar story about Youk. (The rest of that story, from their 15th anniversary issue, is behind the ESPN+ payroll, but what excuse do you have now that it essentially comes free with a Disney+ and Hulo combo?!)
|
|
|
Post by umassgrad2005 on Sept 30, 2020 14:11:31 GMT -5
When you were with the Red Sox what did you guys do to adjust numbers for variation and small sample sizes? Did you guys do things like that? Did you just look at averages? Did you dig deeper to see if it something else? I ask because looking at Manny's first season with the Red Sox he didn't have many games off. You'd be comparing like 142 games to like 10. Which isn't the same as comparing 76 to 76. If you start breaking that down into first game back, then second game back, etc. You are getting a bunch of small sample sizes and baseball players, even the great Manny have a ton of variances in his numbers in small sample sizes. An example Manny is Manny 7 out 10 games after day off, yet bad in three. On average that could make him average, yet that doesn't tell the whole story right? Small sample sizes are crazy hard in Baseball and averages can hide things. Like how far down the rabbit hole did you guys go? Like teams he played, pitchers he faced, ball park effect, was he coming off an injury?, did he just run into a hot pitcher, did he party more on off days, etc? There was no "you guys"; Bill James, Tom Tippett and I (and who knows who else!) worked separately. They liked to compare the different takes on things. I think I mentioned that I gave a green light to Jed Lowrie's disappointing junior year while Bill thought it was a red flag. So that's the only acquisition I may have influenced (aside from Carlos Pena, which was apparently pretty much me, and who got squeezed off the 40-man within months).
The usage splits (rest / consistency / travel) were almost always done over the course of a career. So all those confounds you identify so well melt away in the aggregate!
I think the only time I looked at a single season was to try and figure out Pedroia's rookie season and his terrible numbers on the road. I gave advice that apparently got passed down and taken, as the split disappearred the next year and the comments he made about the improvement matched the advice pretty strongly. But with a small data set, there has to be both a huge split, then it needs to have an explanation that makes easy sense, and then that explanation has to predict a further split you hadn't looked at. As in, if I'm right about this, then these games belong in the other bucket, and if I move them there the split should get bigger. And they do.
There are a lot of splits that have big Y2Y variations so big that you have to look at the whole career; looking at a single season will just drive you bonkers (performance against LHP by righty hitters is one of these). Another one I used with the Sox is pitcher performance by batting order position as a proxy for opponent hitter quality (that one actually helped got me hired, because I used it to predict we'd' annihilate the Cardinals starter, 105 wins notwithstanding. Every one of their guys had feasted on weak hitters). A third is the challenge / pitch-around splits for hitters that I developed to tweak batting orders -- although that came too late to be really used with the Sox; I talked about it with Zaxk Scott the winter before (while?) the economy tanked and all of the consultants got laid off.
I was well aware of small samples. My system for predicting batter / pitcher matchups, which looked for types of pitchers that a guy could hit, had a ridiculously primitive regression to the mean for the individual matchups when they were less than 32 PA, a number I came up with empirically. Less than 8 PA was meaningless -- except that hitters differed in how much they struggled in general, against all pitchers, in the first 7. So if there were a guy who projected to hit well against a pitcher because of his type (e.g., flyball pitcher who gave up a lot of homers relative to expectations based on K, BB, and FB / GB) and he was always helpless in his first 7 PA but had actually hit this guy the first time he faced him, that combo created a big green light.
(That example, BTW, is not hypothetical -- that was Doug Mirabelli against Mike Maroth on 5/3/05, where I convinced Jed to talk Tito into sitting Papi and having Mirabelli DH. Who then hit a slam off of Maroth in a 5-3 victory.)
(It goes without saying that when there was an 8+ PA sample in a head-to-head matchup, you looked at every game. I would look, obviously, for cheap hits and line drive outs, but the big thing was performance relative to teammates. Sometimes a guy who "has really hit this guy well" is a guy who got lucky and only faced him when he had an off night. And vice versa. The message here is that most of what the Statistically Correct regard as mere random variation is real differences that we don't measure well. )
Some interesting stuff. The way I look at it 150 games in a season doesn't equal 15 years of ten random samples of 10 games. Just because you look at his whole career it doesn't make the issues of small sample sizes go away. It actually makes the margin of error higher. Your still looking at like 2000 games versus 300. Your just adding up a bunch of 140 to 10 game samples. We still see high variances with players from season to season with large sample sizes. I give that information to my boss and he's asking me to prove it. Assign confidence and margin of error to it. Dig deeper to make sure averages aren't giving you a false narrative. Look at the standard deviation from game to game and season to season before we would every talk about something as a fact. Heck we really never do that to begin with, it's more I think this is true and this is my confidence level in it. Baseball is hard and 10 different people can look at the same numbers while getting very different results and none of them being 100% right. Not with the data in this case, with it's variance and major difference in sample sizes.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,924
|
Post by ericmvan on Mar 14, 2021 6:37:00 GMT -5
For those new to this thread, he's hit much worse in his MLB career after a day off than when he has played the day before.
Not in this thread: given the splits he has and the frequency of off days, he still wouldn't be a good MLB hitter overall. He needs to lift the performance after an off day from terrible to OK.
The hot spring (no surprise since you hit every day, a lot) inspired me to take the next step in this study. So I've now looked at Chavis' minor league career, and the same split is in 2016 through 2018, each year, although not as dramatic. It's also in 2014.
2015 was the year he was trying to simultaneously learn 3B and how to hit pitchers above the GCL level. You would presume he'd be working on one or the other every possible off day. He had 471 PA for Greenville add rather notoriously hit .223 / .277 / .405, putting his whole top-prospect status into question.
He was also DHing quite a bit because he was sharing the position with Devers. You might think that when he had a day off and was expecting to DH the next day, or when he was DH'ing for the second day in a row, he might hit somewhat better, because he would be taking a lot of of BP in anticipation of just hitting, while not having to worry about the glove.
You'd be right. He had 103 PA in those circumstances and hit .305 / .350 / .633. Half of his 16 homers came in those games, in 95 AB versus 336.
More at some point. Don't I have a study underway of another 1B/ 3B with plus power?
|
|
gerry
Veteran
Enter your message here...
Posts: 1,664
|
Post by gerry on Mar 14, 2021 11:00:29 GMT -5
Wonderful dialogue here on a wonderful day when the Sox are actually available on NESN two days in a row. Could a study be done on NESN’s erroneous thinking about how a baseball starved fan base during a pandemic, and with about 20 new players to observe, wouldn’t want to watch every single game?
Pardon the off topic rant about NESN’s callousness.
If Chavis were primarily a DH, based on your numbers, there is potential for him to hit about .900 OPS and 40HR in about 500AB. And he could backup four positions. That would also be wonderful, and somehow this doesn’t seem entirely impossible for this kid. AAA might be a good chance for him to actually practice this.
|
|
|