Two Myths I'm Ready to Debunk

[hoosiers]

And the overwhelming evidence shows that when you translate MiL stats for league, park and age of player, good MiL stats will translate to good ML performance. Not for EVERY player, but in general.

What does "in general" mean? Six in 10? Seven? Eight?

[Sports Guy]

No. Here's what I'm trying to find out:

• Several people claim that MiL numbers for a given player are a reliable (albeit imperfect) predictor of that player's ML performance.
• What I am asking is for a rational quantification of "how reliable" (which boils down to the same thing as "how imperfect").
• This is the kind of question that stats guys in other fields routinely address. This is the kind of thing that stat procedures are good for.
• To make it concrete, I want to know if a given guy has MiL performance of [plug in whatever performance level you want], what percentage of the time will he (a) have a big league career beyond the "cup of coffee" kind, and (b) have ML numbers that mirror his MiL numbers per some adjustment factor.
• All I see is normative data that addresses this for large groups. I'm asking how that translates to individuals. It's a fundamental error to assume that they are the same. They're not. For example, life expectancy for your year of birth might be 76 years, but that does not give us *any* confidence that you will drop dead on your 76th birthday.

The question I am asking is not a crazy question. This is not an obtuse, arcane question. And it is certainly not just a "semantic" issue. To the contrary, this is *exactly* the kind of info we need to have if we are to know how to sensibly treat a guy's MiL numbers.

What happens instead is that we get vague-but-blanket claims that MiL numbers are "reliable but imperfect" for an individual. Once you enter Stats World, that answer is not meaningful. The whole point of stats is to quantify things that are otherwise vague and fuzzy so that you know how to look at things better. I'm just asking for info about where stats guys use stats to figure out how much confidence we can have in MiL stats predicting an individual player's ML performance. The appropriate answer will come in the form of a number that is produced by appropriate and well-established statistical procedures. You can have all kinds of arcane debates about which procedures are best, given the properties of the problem-space, but I'm not arguing about that; I'm just asking for *any* versions of that kind of analysis. What we need is something that will tell us whether we can trust it 30% of the time, or 70% of the time, or whatever. That doesn't tell us whether a given player falls into the right bucket, but it will tell us how much weight to give these things.

So then what you need to do is go through every player who has ever had significant abs or IP at the ML level...You then need to go back and check their MiL numbers.

Seems to me that you want someone else to do your research instead of just using the info provided and common sense.

[DrungoHazewood]

No. Here's what I'm trying to find out:

• Several people claim that MiL numbers for a given player are a reliable (albeit imperfect) predictor of that player's ML performance.
• What I am asking is for a rational quantification of "how reliable" (which boils down to the same thing as "how imperfect").
• This is the kind of question that stats guys in other fields routinely address. This is the kind of thing that stat procedures are good for.
• To make it concrete, I want to know if a given guy has MiL performance of [plug in whatever performance level you want], what percentage of the time will he (a) have a big league career beyond the "cup of coffee" kind, and (b) have ML numbers that mirror his MiL numbers per some adjustment factor.
• All I see is normative data that addresses this for large groups. I'm asking how that translates to individuals. It's a fundamental error to assume that they are the same. They're not. For example, life expectancy for your year of birth might be 76 years, but that does not give us *any* confidence that you will drop dead on your 76th birthday.

The question I am asking is not a crazy question. This is not an obtuse, arcane question. And it is certainly not just a "semantic" issue. To the contrary, this is *exactly* the kind of info we need to have if we are to know how to sensibly treat a guy's MiL numbers.

What happens instead is that we get vague-but-blanket claims that MiL numbers are "reliable but imperfect" for an individual. Once you enter Stats World, that answer is not meaningful. The whole point of stats is to quantify things that are otherwise vague and fuzzy so that you know how to look at things better. I'm just asking for info about where stats guys use stats to figure out how much confidence we can have in MiL stats predicting an individual player's ML performance. The appropriate answer will come in the form of a number that is produced by appropriate and well-established statistical procedures. You can have all kinds of arcane debates about which procedures are best, given the properties of the problem-space, but I'm not arguing about that; I'm just asking for *any* versions of that kind of analysis. What we need is something that will tell us whether we can trust it 30% of the time, or 70% of the time, or whatever. That doesn't tell us whether a given player falls into the right bucket, but it will tell us how much weight to give these things.

I suggest you follow this link, select Clay Davenport from the dropdown box, and ask the guy who's spent the last 10 or 15 years developing minor league translation systems your question.

Seriously.

[rolliefingers]

What I am asking is for a rational quantification of "how reliable" (which boils down to the same thing as "how imperfect").

First, we'd have to agree on what constitutes "great, good, decent and poor" stats in both MiL and ML.

Then we'd have to adjust every player's stats (for both MiL and ML) for league, park and age, and then place each into one of those four categories.

Catching my drift?

Personally, I don't care if Rogers Hornsby's MiL stats correlated with his ML stats. I'm willing to base my assumptions on just the modern era (1993? and later). If someone has worked out the correlation ratios for everyone who's played since 1993, I'd like to see them.

[rolliefingers]

What does "in general" mean? Six in 10? Seven? Eight?

8.49231

Seriously, go find Clay Davenport's e-mail address and ask him yourself. I am not an expert on this stuff and have never claimed to be. I believe in this stuff because I believe in both scientific analysis and common sense.

[clapdiddy]

I went back and found your link to this: http://www.baseballprospectus.com/statistics/minoreqa.php. Is that what you were referring to in post #54?

I went there are read it. What I see there is application of normative data to many players to adjust for ballparks and league performance. I also saw claims that reams of data somehow prove that it's a "valuable tool". Maybe it is, I don't know. But I did not see *any* analysis tells us how much confidence we can have in expecting an individual's MiL stats to predict that individual's ML success. Now, it's not unheard of for me to be looking right at something and not see it. So, maybe I missed something. If I did, please help me out. Where is the part where it tells us how much we can trust MiL numbers to predict ML performance in a given individual's case? What is the correlation factor? How much of a determination relationship is there? I saw zilch about any of that. But maybe I was being stupid (it happens).

I'm not trying to start a fight, but do you have an example of a player or 2 who had inferior number in the minors, but managed to put up good to very good statistics in the majors?

[weams]

I'm not trying to start a fight, but do you have an example of a player or 2 who had inferior number in the minors, but managed to put up good to very good statistics in the majors?

If they do....those players discovered better living through chemestry.

[RShack]

I'm not trying to start a fight, but do you have an example of a player or 2 who had inferior number in the minors, but managed to put up good to very good statistics in the majors?

No, not at all. In fact, I have asked that question several times, simply because I'm curious. The reason I'm curious is not to somehow diss stat guys, but rather because I think it would make a fascinating story. Personally, I would expect that we can have high confidence that lousy MiL numbers telling us what we need to know. But I do think that looking at any exceptions would be very interesting, just like many fluke things are fascinating. For example, Koufax is a fluke in that his career divides into 2 very different stories, and the latter story is that he was the God of Pitching.... with only 2 pitches!. I don't think it proves anything, but I still think it's fascinating, as are the various explanations for why it happened.

EDIT: check out Matt Holliday's MiL vs. ML numbers. I know it's partially a Coors thing, but still...

[clapdiddy]

If they do....those players discovered better living through chemestry.

No doubt...and I'd be shocked if Rshack or anyone will find a player in this mold.

I would think, as a generality, teams don't tend to promote guys who don't put up good numbers...except for the Orioles.

[rolliefingers]

No doubt...and I'd be shocked if Rshack or anyone will find a player in this mold.

I would think, as a generality, teams don't tend to promote guys who don't put up good numbers...except for the Orioles.

Dude, guys with .600 OPSs in AA are the new undervalued commodity. Didn't you read Moneyball?

[RShack]

Seems to me that you want someone else to do your research instead of just using the info provided and common sense.

No, what I want is for people who routinely make broad sweeping claims to do their own homework to justify those claims, rather than demanding that others do it for them. What happens here is kinda like saying "Most elephants like to eat ketchup. Everybody knows that, so it's up to you to prove I'm wrong."

What makes it especially nuts is that the topic-at-hand is stats, and stats is precisely where people in other fields routinely address important issues about what conclusions stats do and do not warrant. I don't know why it's so hard to get simple answers to basic questions here.

[DrungoHazewood]

What makes it especially nuts is that the topic-at-hand is stats, and stats is precisely where people in other fields routinely address important issues about what conclusions stats do and do not warrant. I don't know why it's so hard to get simple answers to basic questions here.

I've pointed you in the right direction. You just have to want the answers more than you want to have another 87-page thread where you say we won't do the legwork for you.

[Witchy_Chick]

Future performance is a bell curve, a range of possibilities. Within that range they're very reliable, but we freely admit that some individuals will be out in the tails of the bell.

I don't see how this is such a hard thing to grasp, or that no matter what the standard deviation of that bell curve is (within reason) that this gives us a valuable tool.

Well, some stat-heads are more willing to admit this than others.

I brought this up a while ago in the Trax thread (IIRC): How many times does a player have to buck the stats' odds before their actual performances are taken with more weight than what the stats "say" those performances should be? When does an anomaly cease being an anomaly and become the norm?

That means that using House's 3-for-20 stint - spread out over a month, no less - to prove that he's not an ML player is the height of madness.

Lots of guys struggle when they first come up, and lots of guys have slumps at inopportune times. You have to give them more than a handful of opportunities before drawing conclusions.

BTW, Bynum and Trax were pretty much exactly what we expected they'd be. Guthrie is the only guy who really supports your argument.

I beg to differ on Trax. Prior to the O's signing him, I had no pre-knowledge of his abilities/peripherals/what-have-you. Based on many posts here after his signing, I expected him to be horrific -- and he was far from that.

I understand that past stats & peripherals can show trends, reasonable expectations, etc. BUT, many people here take them as a certainty of what players will do.

Witchy

[RShack]

I've pointed you in the right direction. You just have to want the answers more than you want to have another 87-page thread where you say we won't do the legwork for you.

Sound to me like you don't know how to justify the claims you make, and instead of saying that, you tell me to ask someone else. I didn't even ask you to justify your if-fy claim, I just asked you to point me to an article. The article you pointed me to had zilch to do with the question.

Do you understand the question? If you do, are you claiming that it's about semantics or arcane stuff? Or do you agree that it's important?

[Sports Guy]

No, what I want is for people who routinely make broad sweeping claims to do their own homework to justify those claims, rather than demanding that others do it for them. What happens here is kinda like saying "Most elephants like to eat ketchup. Everybody knows that, so it's up to you to prove I'm wrong."

What makes it especially nuts is that the topic-at-hand is stats, and stats is precisely where people in other fields routinely address important issues about what conclusions stats do and do not warrant. I don't know why it's so hard to get simple answers to basic questions here.

Well, why don't you back up your claims?