THE BOOK--Playing The Percentages In Baseball

The one I always like to go to first is Derek Jeter:
http://www.tangotiger.net/scouting/sim2006_5406.html

His top comp is the equally-sabermetrically-maligned Michael Young.  Jeter and Young are clear “Fans against MGL/Dewan”.  His top comps are all over the place: 5 SS, 3 3B, 2 2B.  5 RF, 2 LF, 1 CF.  Two catchers. 

Contrast this with Tejada:
http://www.tangotiger.net/scouting/sim2006_5888.html

His top 6 comps are OF (5 RF, 1 LF).  His next 4 comps are 3B.  Fans see him as a prime candidate to move to 3B or RF.

Another interesting one is Junior:
http://www.tangotiger.net/scouting/sim2006_4305.html

His comp list is littered with catchers!  His evaluation shows him as being a very slow runner, but a good arm.  The rest of his top comps are 1B.

(I wouldn’t pay too close attention to the Catcher position, as they would certainly be the most difficult to evaluate, not to mention evaluate without position-bias.)

In any case, Junior should be playing 1B, not CF!

Yowza, here’s Rally Monkey’s ZR-based list:
http://home.comcast.net/~briankaat/MLBzoneplus.htm

Check out Betancourt at SS.  I guess when you look at Fans’ and ZR/UZR (PBP) metrics, it gets down to one thing:
1 - Fans are fairly consistent, even if consistently wrong
2 - PBP metrics have alot of uncertainty, though their mean is better than the Fans’ mean

It’s the classic high-risk, high-reward tradeoff.  Me?  I prefer T-Bills to junk bonds.

It’s the classic high-risk, high-reward tradeoff.  Me?  I prefer T-Bills to junk bonds.

And that’s why you’ll never be rich. 

Seriously, congratulations on pulling this together again.

According to the THT team page, the Mariners are +12 on fielding groundballs.  Rally’s list has the M’s regulars as 11 runs below average.  Something does not jive w/ those number’s.

Two completely different methodologies with different goals.  You can’t compare the two.

Studes, if I’m interpreting your data right, the +12 for the Mariners is 12 more groundballs turned into outs than would be expected based on batted ball types, right?

Watching pretty much every game this year, that feels about right.  Beltre and Betancourt are both good, Betancourt being the better good of the two, but I think there’s some diminishing returns with having those two next to each other.  There were a handful of plays this year than one or the other made where I’m pretty sure either one could have had a legitimate play on, and they weren’t routine plays.  These were balls on the fringes of both zones, but both guys were in position to make the play.  As such, I think their overall values will be a bit lower than their true talent level. 

Lopez is pretty mediocre, probably a couple of plays below average.  And Sexson’s bad.  +10 to +15 plays is in the range of what I’d have claimed without looking at any numbers. 

And while I enjoy Rally and Dial’s work on ZR, I’m pretty confident saying that the fact that they have Betancourt scoring so lowly exposes a flaw in the system.

More likely a flaw in the scoring system, since they simply take ZR as their basis. 

***

Do you know that the NHL has SIX official scorers to track PBP, in addition to the official scorer, game timekeeper, and penalty timekeeper.  Perhaps one day, when MLB generates more revenue than the NHL, and the public yearns for more data from the MLB than NHL, MLB will rectify this situation.

The list Tango linked to does not have the M’s regulars as 11 below, but one below (as long as we’re talking groundballs)

Sexson -6
Lopez +2
Betencourt -6
Beltre +9

Those are projections based on 4 years of data - so you can’t compare to what Hardball Times has for groundballs, not to mention Stude’s warning that there are different methodologies.

I think we’re in agreement with 3 of the 4.  Maybe Betencourt is a fielder who looks good but doesn’t make all the plays?  I wonder how UZR and John Dewan will rate him.

Sometimes the pbp systems don’t even agree with each other.  Andruw Jones below average, which is no surprise (though he was a zone rating god his first 3-4 years).

But MGL recently posted that he had Jones above average (+9?) for 2006, which was surprising since UZR had him as a bottom 3 NL CF around the mid point of the season.  Must have been one great 2nd half.

Rally, I’ve got 221 Mariner fans in my corner, who think YuBet is the greatest thing since sliced bread!

In the other corner is Dewan (and likely UZR) that doesn’t think Betancourt is hot stuff (and, is likely one of the worst SS in the league).

Now, if YuBet was a long-standing member of the Mariner fan base, who gets brownie points for once being good, and just hanging on, that’s one thing.  But, last year as a rookie, 30 Mariner fans thought the same of him.

***

The major problem with PBP data is the classification of each ball.  There is no question that some balls are routine, that should be handled 99%+ of the time.  However, *no* PBP system has the sufficient parameters to establish a particular ball in play as being a 99% routine play.  The best they’ll come up with is 95%.  Now, we may think “95%, 99%… pretty close, just off by .04”.  But, the difference between a great SS and an average one *is* .04 outs per play.

Therefore, I have a huge issue with the scoring system, that there must be a certain amount of uncertainty around that scoring.  That perhaps there are biases that apply more to YuBet than to other shortstops.  After all, most SS don’t play in Seattle, and most don’t play against those pitchers. The physical scorers who record the plays may not necessarily score other teams, or in the same way.

If Dewan/UZR want to come out and say “you know, these scoring biases, the GIGO biases, will affect 10% of the results, which should be thrown out”, or some such, that’d be good.

Rally, I somehow read Sexson is -16, that’s how come I said -11, sorry.  I understand ZR and THT’s GB #’s are totally different methods and goals, and did not realize you had you were using 4 years of data (although the use of MaxOfYEAR and the Wch description should have clued me in).  My point was, at least based on my own (albiet biased) observations, I don’t see how Betancourt is a -6 SS, or even not above average, and I personally believe the +12 is largely a result of Beltre and Betancourt (and is dragged down some by Sexson).  He’s not pulling a Jeter or Eckstein and making routine plays on crazy dives and running balls out to make them, he’s making long running grabs in no-mans land, tossing the runner out quickly, and making it look easy and effortless, play after play.  That said he may not be as good as the fan scouting indicates, but I think he probably is.

Betencourt may well be a great fielder.  I’m sure there are many situations where pbp fielding estimates get things wrong.  There are biases in the data that we know (ex. Manny and the monster) and biases we haven’t figured out.

I’m not digging in my heels and claiming he’s a bad shortstop.

Given the choice between data and scouting reports, I’d probably weight the two evenly and call him average right now, as he’s only played for a year and a half.

Two years from now if he’s still getting below average ZR’s I’ll have more confidence in that data.  But you never know, he might wind up +10 next year and we can all agree on his greatness, much like Jose Reyes (looks great) finally had a good defensive year in 2006.

What is interesting is how little faith we have of others’ observation today, but we’re more than be willing to accept the observers claims of 30-100 years ago.

Check out this article from a bit over a year ago on Betancourt.

This is what Rohn thinks:

“He’s one of the best shortstops I’ve seen. Ever. I compare him to Ozzie Smith. I know that’s putting a lot of pressure on him, but instincts-wise, he’s one of the best I’ve ever seen.

“There’s not a lot we can do for him. Just let him play. I’m not smart enough to help him. He’s too good.

I guess the question I have to answer myself is how many sub-zero UZR years would I need from Betancourt, that lets UZR more than trump the Fans.  Right now, I’m at 80% Fans, 20% UZR.

I have Yuniesky at -5 in UZR.  -3 in range and -2 in errors.  Last year at SS, he was -9 per 150 or -3 total in only 48 games.  As always, who knows?

Tango, are you simply averaging all the individual ratings to come up with a “total player rating?” Is there any attempt to weight the ratings, given that some of them are clearly more important than others in terms of a fielder’s ability to save/prevent runs?  For example, what the heck is “hands” for an OF, and who cares about that (for an OF)?

OK, I see you do weight the categories.  How so?  And what the heck is a “globe” (globe gloves)?  You Canadians have some weird humor!

When I do position-specific, I weight.  For example, for catchers, the speed trait is the lowest, and the arm is the highest.  For OF, the arms are lowest, and the first step/speed the highest. 

However, in the reports on the site, I think they are unweighted.  No biggie, since I present the data, and the reader is free to weight as he sees fit.

“Hands” would be catching the ball.  If you’ve seen Vladimir play OF, you’d know that Angel fans care about that.  I’d guess at least half the OF errors are catching errors.

***

Btw, I believe Rally Monkey ran a regression of ZR and the Fans against UZR, and came up with a high correlation.  Maybe he can point us to the relevant article.

***

Yes, the Globe Gloves, since it’s a WWW of fans.

FWIW, Betancourt ranks about average in Dewan’s system again this year.  Andruw ranks highly this year.

Kotsay is another one whose PBP ratings seem to belie the naked eye.

I was about to mention Vladimir when I saw your post.  He probably dropped 4-5 balls this year, where my guess is the average OF drops 1.  He’s always been a bit of an adventure in RF.

It wasn’t me - might have been David Gassko.

Kotsay doesn’t surprise me - I’ve seen a lot of balls fall in front of him or in the gaps.  He used to rate very highly on pbp, but the last two years he’s had a lot of injuries and he’s getting older.  He does have very good instincts on flyballs near the wall, so I would take him any day over Steve Finley, who has none.

I think Milton Bradley is the best CF on that team now.

Ken Arneson, A’s fan, told me last year that Kotsay had an injury near the end of 2005, where he was “hesitant”.  I don’t know what he’s like this year.

If there’s one thing that is clear to me about OF, it’s that all-out hustle (Erstad, Rowand, Jenkins) is a huge plus.  If part of Kotsay’s value is that he used to always hustle like crazy, and that now, because of that injury, he’s a bit more hesitant, that may be enough to turn a guy from fabulous to average (though hard to believe).

UZR has been in love with Kotsay, up through 2004.  Coincidentally (or not), the love affair ended last year.

***

With respect to Fans, there is a clear bias in star fielders, with accepting they’ve lost a step.  Just this year was the first time that Junior rated as a below average CF, while every other year, he was above average.  UZR has been on to him since forever.

***

Andruw Jones did drop sharply this year, though still above average, according to the Fans.  Same with Torii.

***

As for the A’s, they have Payton, Bradley, and Kotsay, each a CF.  It will be interesting to see how the PBP metrics sees these guys, especially in games where they all play at the same time.

Btw, I believe Rally Monkey ran a regression of ZR and the Fans against UZR, and came up with a high correlation.  Maybe he can point us to the relevant article.

***

Rally’s right - that was me, though I didn’t look at Fans. Here’s what I did:

http://stats.mostvaluablenetwork.com/general/a-detailed-comparison-of-defensive-metrics/

My finding was that 2/3*ZR + 1/3*Range (my defensive metric) will give you a .87 correlation with UZR.

This is what I was talking about

Scouting Report has been updated, and finalized.  Not much movement, but a few more ballots came in. 

Ballot counts have been something like 500, 800, 1100, 1400 for the last 4 years.  I like how it’s gaining traction, though I got a 200 ballot bonus from the ussmariner.com readers.