THE BOOK--Playing The Percentages In Baseball

HOORAH FOR DELPHI!!!

As a met fan I can definitely tell you it was an adventure with him out there.....he misplayed alot of balls.  I do think he’ll be better as time goes along....but not in CF....i think he’s gonna be RF or LF honestly especially with the mets since beltran isnt going anywhere soon.

Tango,

I live in the Jersey area, and I listened to some of those fans calling into radio stations.  Milledge hit the MLB with a bang, and the fans loved him. They couldn’t foresee possibly disliking him in the future, they liked him so much.  Then he went through a stint where he dropped anything hit to rightfield, and forgot how to hit.

Milledge played 56 fans. I bet your fans are basing their judgments on 4 or 5 bad plays.
And that performance is as unreliable as his batting performance of .240/.310/.380.  If you believe Silver’s projection of .285/.355/.455 for *next year*, why don’t you think he has little shot of improving his fielding?

You’re 100% right that the “wisdom of the crowds” eliminates *some* sources of error (e.g., inconsistency), but not others. I’ll trust the scouts and coaches that have been watching Milledge for years before your Mets fans who’ve seen him for 56 games.  And you should too.

Best,
-cdm

Nate is not saying he will improve his hitting.  He is saying his performance numbers will be improved.  That’s not the same thing at all.

When Jeter goes 0-for-April, and then he hits .320 the rest of the way, he didn’t “improve” necessarily.  He may have performed exactly the same way, and changed nothing from April to the end of the season.  It’s understanding the difference between sample and true talent.

So, for Milledge, let’s see.  I’ll see if I can get someone to give me some pro scouting reports on him.  (Then again, those scouting reports where not high on Betancourt and Francoeur before they came up.)

Thanks, but I’ll go with the fans, in general.

That’s quite a stinkaroo of an “Instinct” rating on Millege’s scouting report. 17? I suppose the optimistic Millege fan might hang his hat on that, because attributes such as instincts, first step, positioning, etc. are thought to improve with seasoning. But then again, the word “instinct” implies something that’s innate, you got it or you don’t. And guys Millege’s age typically should be at or nearing their peak defensively. If/when he’s posting numbers like Silver says, he may very well be worse than he is now.

There’s always exceptions, of course.

Tango, off the top of your head, any notable players similar to Millege defensively--younger guys high on “tools” (speed, arm) but low on “skills"--who have managed to become good defenders, according to the fans?

If you click on his name, you will see his comps.  I haven’t done anything historically, otherwise.

My guess is that you can pick out some players in past scouting reports, like Roger Cedeno, and you’ll probably get decent comps that way.  Maybe Vladimir is another decent comp of a guy with speed, and arm, but who sometimes looks lost out there?  I remember hearing the same thing about Encarnacion. These guys don’t just become good.  Why would they become good, and not other players?

And, shifts to CF are done by good corner OF, and not guys who can’t catch a ball. 

***

jianfu: you don’t have to put a fake email address.  That field is optional.

Tango,

Of course, by “improve” I meant what you mean: regression to a “true” performance rate.  Are you really saying that 21 fans viewing no more than 56 games (84 chances) can allow for a reliable estimate of his “true” fielding performance level?

And no one said that he would be good. They said credible.  I read that as below average, but not so bad that he shouldn’t be playing the position. The Mets fans think Milledge has worse instincts and hands than Jason Giambi.  That just screams “Small Sample Size!”

Lastly, Milledge isn’t a corner outfielder. He shifted to the corners for the Mets because CF was full.  He played CF for 3 years in the minors at a “credible” rate, according to BP fielding stats (~ -3 FRAA).

The guy had 2 bad months. His “true” rate is better than your Mets fans are giving him credit for. I’ll bet $20 on it: your Mets fans will come around next year, provided Milledge plays enough for him to settle down, and for them to see him play.

-cdm

Credible means below average, and it also means he can last there 2-3 years without people being in a panic every month.  That is, he can be 2 to 5 runs below the average CF, but he can’t be 10 to 15 runs below the average CF.

Right now, I’d have him as being a credible corner OF at this point, which would make him a -10 to -15 in CF.

Yes, I am saying that 21 fans viewing a few dozen games will provide a “reliable” estimate of his true talent.  Not in 100% cases of course, and Mets fans are pretty tough.  We’ll see if their love affair changes.

There are issues in interpositional comps, so at the least comparisons should start with positional comps.  Among the 165 OF that were evaluated, Milledge ranks 153 in instincts and 157 in hands.

Among the 165 OF with poor hands and poor instincts (score under 30), and good speed (score over 60), two are CF: Kenny Lofton (13 ballots) and Reggie Abercrombie (8 ballots). 

In his first 30 or so games, Abercrombie was shifted to the corner OF at the end of the game 10 times.  He came in as a PR or PH 23 times, and subsequently took the field only 6 times.  Five errors in 65 “full” games.  Fans and manager in agreement: not credible in CF.

The other guys who compare on these criteria:
- Podsednik: tried in CF, moved to LF
- Termel Sledge: plays only corner OF
- Ozuna: IF converted to LF
- Jason Ellison: started in CF, moved to corner OF

It’s the same story every year since baseball began: hope that a guy with speed can make it in CF.  Hope is not credible.

I have the results of the pro scouting reports.

ARM:
Professional scouts - average strength, average accuracy
Mets Fans - average release, strong arm, above-average accuracy

So, Fans are more optimistic than the professionals here.

RANGE:
Professional scouts - above average range
Mets Fans - average jump, great speed

It sounds like the Mets Fans and Professionals are in the same ballpark here.

FIELDING:
Professional scouts - slightly above average
I’m not sure what “Fielding” means, over and above the other two items that the professionals ranked.

Mets Fans say that he has horrible instincts and horrible hands.

So, professionals are more optimistic in this category.

Overall, the professionals have Milledge as a bit above average fielding-wise.

Overall, the fans have Milledge as a bit below average fielding-wise.

(Note: all discussion here does not include the position.)

To be a CF, you’ve got to be alot better than that.  Based on the pro scouts, it looks like Milledge would be able to play a credible CF (maybe 2-3 years without panic).  Based on the Fans, Milledge would last less than 2 years as a CF, and likely under a year.

***

We’ll see next year how the pros and Fans evaluate him.  My guess is that the pros will move closer to the Fans, than vice-versa.  But, let’s see.

I should also mention that it was Mets scouts who agreed that Kaz Matsui could play SS (which prompted the Mets brass to move their then-future star Reyes to 2B).  Mets fans thought that Matsui was one of the worst middle-infielders ever.  The brass relented the following year, by flip-flopping the two.

How bad must the Mets scouts have been (or if they were good, how utterly irrelevant), if they actually thought Kaz Matsui at SS was a good thing.  And Kaz was heavily scouted.

Thanks, but I have more faith in fans.  Fans tell it like it is.

Tango,

You presented data supporting my position, and then rejected it without justification.

I agree that there is a fair bit of signal in the fans data, and a ton of noise in scouting reports.  And, in the case of Milledge, I think the fans are reporting something real: Milledge had an AWFUL year defensively.

But is that predictive?  Tango, you would never take 80 PAs as a strong predictor of “true” talent or future success.

If Milledge played as poorly in his minor league career as he did last year, he would have been moved to 1st base.  But he didn’t.  He has played CF (poorly) for the last 3 years.  Clearly he was better at CF in the minors than he was at LF last year. 

This indicates that last year was a low point from which he is extremely likely to rebound.  This is a statistically sound principle that is overwhelmingly supported by the empirical evidence.  Can this really be contradicted by the fans perception of 80 attempts?

I clearly made my statement with “my guess” and “let’s see”.  This is not like when I talked about Chase Utley’s signing as being genius by Pat Gillick.

***

There is a reason that 80 PA as a batter and 80 PA viewed by a fan is different.  The uncertainty level of 80 PA of performance is extremely high.  The uncertainty level follows the binomial distribution.  1 SD is .053 OBP with 80 PA.  So, a guy who hit .350 OBP is 95% sure to be somewhere between a .250 and .450 OBP guy, absent any other information.  That’s pretty much useless.

But, if you watch Miguel Cabrera’s first 80 MLB PA, or Tim Raines first 80 MLB PA, or any good hitter, you really don’t care if their performance OBP is .250 or .450.  Your prior for them will have a much smaller uncertainty than 80 performance PA.

That is, based strictly on scouting, your prior for Cabrera’s first 80 PA at age 20 might be an OBP of .370.  That uncertainty level might be 1 SD = .020, so that we’re 95% sure that he’s a true .330 to .410 OBP, after only 80 PA.

In reality, his first 80 PA was a .300 OBP.  His career OBP to date is .384 (after 2400 PA).

***

The difference between performance numbers (binomial) and scouting is that the uncertainty level in performance stats approaches zero as the PA approaches infinity.

Scouting, however, starts with a small uncertainty level, and never reaches zero.  It might say start with an uncertainty level of 2 SD = .300 after 4 PA (as opposed to binomial of 2 SD = .500), and then very very quickly go down to 2 SD = .040 after 80 PA, and then pretty much stay at 2 SD = .030 and that’s as low as an uncertainty as you can get, even with infinite number of PA.

Performance numbers gets to 2 SD = .040 after about 600 PA, and 2 SD = .030 after 1000 PA.

So, give me 80 PA evaluated based only on scouting, and I’ll be as uncertain as 600 PA based only on performance numbers.

Give me 150 PA or 1000 PA or 2500 PA based only on scouting, and I’ll be as uncertain as 1000 PA based only on performance numbers. 

After that level, performance numbers become much more important.

That’s why you can’t equate “80 PA” in terms of uncertainty with scouting and performance numbers.

For fielding, the numbers are even less uncertain for scouting, and more uncertain for performance numbers.

How many ground balls did Scott Rolen have to field, before you knew you were witnessing one of the best fielding 3B of our generation? 

How many performance numbers did it take?

***

The question therefore goes to pros v fans.  I’ll equate 125-1000 more-than-casual fans to 25-100 good fans to 5-10 hardcore fans to 1 pro. 

***

All non-binomial numbers strictly for illustration, and all pulled from my a$$.