THE BOOK--Playing The Percentages In Baseball

Another find is Akinori Iwamura.  He was a perennial Gold Glover in Japan, and has vaulted himself in a class just a shade under the big three (Rolen, Beltre, Chavez), and with the future (Feliz, Zimmerman, Inge, Crede). 

Now, Kaz Matsui was also a Gold Glover in Japan, but was horrible with the Mets.  (The Mets’ “scouts” thought so much of Matsui that they moved Reyes from SS to 2B to accomodate him… Reyes is this year one of the best fielding SS in MLB.) Matsui is now redeeming himself at 2B.

See where I’m going here?  Who the heck knows what the scouts were thinking, and how much pressure they had from the Wilpons or the GM du jour.  Fans could see why Iwamura won those gold gloves, and couldn’t see why Matsui did.

Tango,

Do you think that most of these votes come from fans watching on TV or actually attending games?  If it mostly comes from TV-watching, as I suspect, I wonder how fans judge “first step”, which isn’t generally visible on TV.

It also looks a little strange that all the top “First Step” guys are speedy baserunners—I would have thought that some not-very-fast guys would have a great first step.

Here are the positional totals:

Pos All In Fi Sp Ha Re St Ac
CF 60 59 69 73 60 53 46 50
SS 59 58 60 59 56 60 60 58
2B 56 56 58 57 53 57 50 57
3B 54 55 54 49 52 53 58 52
RF 50 44 49 54 44 50 62 55
LF 43 39 47 52 41 41 40 43
1B 41 45 36 32 48 41 40 44

C 42 46 34 31 47 47 49 47

Categories are:
Instincts FirstStep Speed Hands Release Strength Accuracy

Catchers have their own extra traits (experience in various movements and calling pitches) that adds significant value that is not captured here.

***

John, there is of course potentially great bias, since fans may not really be able to separate something as nuanced (simply because they aren’t really looking for it).  Here are the top First Step scores for guys with a speed of under “70”. 

FirstStep Speed Pos Player
86 66 3B Iwamura, Akinori
85 68 SS McDonald, John
84 66 3B Zimmerman, Ryan
83 60 3B Rolen, Scott
83 68 3B Inge, Brandon
82 64 3B Chavez, Eric
81 61 2B Ellis, Mark
81 69 2B Hudson, Orlando
81 61 SS Greene, Khalil
79 63 CF Jones, Andruw

And if I ask for speed score under 50, I get these guys:

FirstStep Speed Pos Player
73 43 3B Crede, Joe
70 48 1B Loney, James
68 45 1B Pujols, Albert
63 49 2B Valentin, Jose
61 44 2B Pedroia, Dustin

If we had an independent body, like scouts, we can see how much bias First Step is influenced by Speed, and discount that.  For example, let’s say the scouts tell us that Joe Crede has the best First Step in the business.  While the Fans were able to recognize Crede with a 73, which is very high, they probably knocked him down a few notches because he’s not a speedster.

What we need is for MLB to be more systematic in their scouting evaluations of players.  For the most part, if not the entirety, MLB teams take “notes” on a Lotus Notes system, that is used reactively, not proactively.  That is, the data is not stored in a database that allows for analysis.  It’s simply a “logging” system, of dates and observations.  That’s a great first start of course.  But, you need the next step.

If they had that, they could cross-check against the Fans’ report, and be able to validate the results or find biases (which can be corrected if they are systematic, as you are suggesting here).

http://baseballprospectus.com/chat/chat.php?chatId=357

JWR (Chicago): Who should win the NL Rookie of the Year this season-- Braun or Tulowitzki? Who looks better over the next five years?

Nate Silver: My first instinct is certainly to say Braun on the RoY question, but what’s interesting is that the DTs perceive an absolutely enormous difference in fielding acumen between the two; Tulo is at +21 (!) while Braun is at -17 (!). I think I’d want to see some independent confirmation of those fielding numbers before I could make any judgment.

On a scale of 0-100, with 50 the average, and 1 SD = 17 (note, 1 SD = 20 in each trait, but overall 1 SD = 17), Tulo is an “81” (one of the best fielders in baseball), and Braun is “40” (fairly below average for a 3B).

To translate those numbers into runs, the “41” point gap is roughly equal to 30 runs in a neutral position (over 162 games).

Tulo’s 81 compares to the average SS of 60, so that +21 points is roughly +15 runs.  Braun’s 40 compared to the average 3B of 54, making him about -10 runs.  Throw in the roughly 5 run positional adjustment, and you have a 30 run difference between the two (over 162 games).

Maybe Iwamura is the Yuni Betencourt of 2007.  The stats don’t rate him that well.  He’s dead last in THT zone rating among AL qualifiers, though STATS zone has him in the middle of the pack.

Betancourt, by visual accounts, has lost something.  And the Fans’ see it that way too.  Here’s how he did in 2005, 06, 07:

In Fi Sp Ha Re St Ac
87 89 78 90 91 81 85 2005
86 85 77 84 85 77 74 2006
77 78 72 74 57 73 23 2007

His problems stem as to what he does once he has the ball in his hands. His accuracy is abysmal, especially for a SS (second-worst among all SS in baseball).  His release/footwork is severely down from 05/06.  Looks like obvious mental lapses, or just being plain lazy.

Whatever UZR, PMR, Plus/Minus has him will likely not show this downgrade.  To the fan, Betancourt has done some real obvious changes, that has severely changed his talent level.  We’re talking about a 15 run downgrade per 162 GP, comparing 2007 to 2005/06.  Even if the PBP systems do show that drop, it will come with a huge uncertainty level.  That is, a 15 run drop in UZR is not a 15 true talent drop, but likely no more than 3 or 4 runs.

However, seeing that Betancourt had an enormous talent level in 2005/06, and has dropped it substantially in 2007, it still mean that he has the potential to get back to his 05/06 level.

UZR will simply note any difference as being part of the sampling error, and will by virtue of the years, estimate Betancourt’s talent as somewhere around whatever the UZR numbers show, but closer to his 07 UZR numbers, whatever those are.

58 RF Upton, Justin

Fans were telling me how tough it was with Upton, since he looked terrible as an infielder and good as an outfielder.

Should this be B.J. Upton?  Justin’s never played a professional inning as an infielder.

My bad. 

The data is correct about Justin.  The comment is correct about BJ.  And BJ is #14 in the “can least agree upon”.

Given the list, I should not have made the comment, since BJ was not in the top 10 list to begin with.

Funny, I’ve been asking myself “Why is Trot Nixon still in the outfield?” for a while now.  The Indians actually had him taking ground balls at 1B earlier in the year, but then decided it would be easier to bench him.

Do you have the “ALL” best in the post or did I miss it?  I just see the “ALL” worst.

The “ALL” will be unveiled with fanfare that will be so overwhelming… kidding.  I’m waiting for each of the bloggers to do a writeup on the 9 players, and then I’ll load up all the files.

However, I did show the top 3 at each position…

Tango,

Any update on Lastings Milledge?  He didn’t get that much playing time this year, but I’m curious if the fans like his defense any more than last year.

http://www.insidethebook.com/ee/index.php/site/comments/lastings_milledge/

Let’s see… let me pull up his record:

All In Fi Sp Ha Re St Ac
52 24 54 78 49 49 71 46 2007
45 17 45 77 18 49 76 62 2006

That puts him at about -5 runs for a CF, which would be about +4 runs for a corner OF.  So, Mets fans saw a bit of improvement from last year: they know he can now catch a ball (important for an OF), but his throwing accuracy is down.

I looked for guys like Milledge: minus instincts (40 and under), decent first step (60 and under), plus speed (60 and over), and I got this list:

In Fi Sp Ha Re St Ac Pos Player
24 54 78 49 49 71 46 CF Milledge, Lastings
30 44 67 42 37 65 41 CF Hall, Bill
33 44 60 37 39 53 43 RF Encarnacion, Juan
23 49 83 37 8 0 13 CF Pierre, Juan
28 44 67 12 9 15 16 3B Ozuna, Pablo

I feel bad for Bill Hall.  He ranked very highly last year, when he was playing SS.  They tried to Soriano-him, and Fans didn’t think the results were pretty.  I’m not sure if this is an IF/OF observation bias, or he simply is not cut out for the OF.

If you have Juan Pierre, there’s really no other place to play him.

Milledge should be a corner OF, on the Mets (Beltran, Chavez) or otherwise.

Interesting to see McDonald rated highly at short. Personally I have a sneaking suspicion that he carries quite a bit of tangible defensive value. I know his UZR numbers have been essentially average over the long term, but I have to wonder given his usage prior to this year, as a late inning defensive replacement, and sometime spot starter at a variety of defensive positions make him hard to accurately measure. It seems to me it’s hard to get on any sort of roll either offensively or defensively if your getting yanked in/out of the lineup.

In terms of McDonald’s first step quickness, he won’t win many 40 yard sprints, but I note he considered quite a good raquetball player. The type of first step quickness and agility of this sport, I would think , translates quite well to the shortstop position.

I’m a little surprised Aaron Hill was not in the top 3 in second base rankings.

It is all very tight:
All Player
81 Ellis, Mark
79 Hudson, Orlando
78 Phillips, Brandon
76 Roberts, Brian
74 Polanco, Placido
72 Hill, Aaron

Note that each “ranking point” is roughly 0.7 runs.  The average 2B is a “56”, so Hill is 16 points above, or roughly +11 runs.  Ellis leads at +17 runs.  Certainly, if someone wants to say that Hill is a better fielder than Orlando Hudson (and the Jays had em both), that can be supported to a certain extent.

Tango, have you previously discussed how you convert category rankings into overall rankings—some skills are more important than others and it depends on position.  And how do you convert overall points into runs—do you convert via SDs to an already existing scale, like UZR?

Right about the SD.  It works out to around 0.7, so that the Fans’ SD and UZR SD match.  This is further supported by looking at the easiest positions: LF/CF/RF.  The overall CF is 60, and the overall corner OF is 47.  That’s a 13 point difference.  I’ve shown in the past where of guys who play both CF and a corner, that there’s a 9 run gap, and 9/13=0.7.  So, it all aligns itself fairly well.

If I apply the 1pt = 0.7 runs rule, and I force the catcher as a “67”, I get the following spectrum:

Pos Points Runs
“C” 67 9.3
CF 60 4.6
SS 59 3.5
2B 56 1.3
3B 54 (0.1)
RF 50 (2.4)
LF 43 (7.1)
1B 41 (9.1)

All 8 54 0

I’ve always had it as around +1.0 wins C, +0.5 CF/SS, 0 2B/3B, -0.5 LF/RF, -1.0 1B.  And the above aligns itself to my simple rule of thumb.  It’s pretty darn close, anyway.

The catcher thing, where I have to force a “67” even though fans rated them as “42” (25 point difference or 17 run difference), I’ve really got no choice in.  Catchers are such a separate pool, that I usually like to keep them out of it, or accept the reality that it’s such a unique position.

(There is an enormous gap between RF/LF these days, and this can either be real, or an observational bias.  It should be noted that half of the best fielding LF have fewer than 60 games played, and that’s not the case at all for CF/RF.  LF may be the place where the worst fielders now find themselves.)

***

As for each position, I do have a scale per position, whereby the “speed” portion is heavy for CF and light for C.  I should probably publish those weights at some point, and you guys can debate/decide if they make sense.

(There is an enormous gap between RF/LF these days, and this can either be real, or an observational bias.  It should be noted that half of the best fielding LF have fewer than 60 games played, and that’s not the case at all for CF/RF.  LF may be the place where the worst fielders now find themselves.)

Can this be attributed entirely to the RF’s having better arms (the Release, Strength, and Accuracy categories) with both groups being essentially the same in the others?

Err, forgot to include, how do the two groups do without their arms?  Is there any difference between how LF vs. RF “field”?

Can this be applied in the IF?  I’m thinking SS vs. 2B vs. 3B.

I have the positional totals in post 3.  2B/3B are, for each of their traits, a bit lower than SS.

The RF is better than LF for each of their traits (including speed!), though far better in the arm categories.

People may be surprised to know that more balls are hit to RF than LF.  We know the arm requirement.  So, it’s possible teams have responded to this by putting their weaker fielder in LF.

Tom,

This is actually something I noticed mid-season:

http://www.hardballtimes.com/main/article/d-fence/

Now, half-a-year worth of fielding statistics isn’t worth too much, but little has changed as I’ve re-run the numbers throughout the year. Here’s what I remarked in that article:

“Actually, it is pretty bad considering the dearth of quality left fielders out there. Other than Carl Crawford (a bland +1 thus far), I’m not sure there’s one good defender playing in left. I mean, c’mon—Craig Monroe? Emil Brown? Matt Diaz? Aren’t you supposed to hide the bad fielders in right?”

Matt Diaz actually is a decent fielder.  Still, there are probably more bad left fielders now than I’ve seen in the past.  The National League has Duncan, Dunn, Carlos Lee, Josh Willingham, Pat Burrell, Wily Mo Pena of late, Bonds, and Luis Gonzalez. 

Its as if they are saying “Hey, we want a DH in our lineup too”

Matt Diaz is a “34”, meaning he’s about -5 runs as a LF.

Interesting insight about the NL/DH.  Don’t forget Alou, when he’s healthy.  This means that Ibanez will be traded to the NL soon.

Talking about Iwamura:

http://www.tbo.com/sports/rays/MGB2DAOB17F.html

Given that you’ve decided you want both Iwamura and Longoria on your team, it’s a question of which guy fits better where.  As far as I’m concerned, 2B and 3B are roughly even in terms of fielding impact.  I don’t have a scouting report profile on Longoria obviously, but I’ve got Iwamura, and he’s outstanding in all facets of the fielding game.  His weakest trait (a 61, which is still above average) is his arm strength, which would make it ideal for 2B.

When I did a quick sim, his top fielding comps are: Ellis, Hudson, Hill (3 best fielding 2B in baseball), Punto (2B/3B), Chavez, and two SS (McDonald, Cabrera).

So, it would seem like a pretty good move.

What do the folks think about Jacoby Ellsbury? I only ask this because I’ve noticed a surprising number of defensive misplays this season ( including two tonight) and his zone rating as a CF is terrible (small sample size). I do put some stock in the fact that he has great speed and has a tremendous defensive reputation in the minors, but the early returns objectively and subjectively (to me) are not anywhere near as strong.

http://www.hardballtimes.com/main/article/ten-things-i-didnt-know-last-week45/

Studes said: “Emil Brown has looked good, too”.  However, he did horribly in the Fans’ Scouting Report.  Below him were Adam Dunn and Jay Gibbons.  And that’s it.  Above him are Luis Gonzalez, Kory Casto (whoever that is), Raul Ibanez, Chris Duncan, and Carlos Lee.  His Range Factor compared to the league is excellent.  He may be getting the job done, but he’s not “looking good” in doing it based on the company he keeps.

***

Ellsbury wasn’t on the ballot, because he only had 7 games played at the time.

Well, he’s looked good to me.  I’m still allowed to have an opinion, right?

Just checking…

Of course, that’s the point of the report, to cull the opinions of all, to the point where we can disregard the opinion of one.  That’s why I keep saying it’s completely irrelevant what I see, and what some other single observer thinks.  You, me, the guy reading this post right now… we simply have such a high uncertainty level around our observations, that we can safely ignore the opinions of one. 

On the Cardinals board, some of the fans there at mid-season were raving how Chris Duncan has really turned it around, fulfilling Larussa’s promise.  And yet, when I get the ballots back, there’s Duncan near the bottom, just like last year.  That’s why I can’t listen to any single person.  I have no idea where they’re coming from, what they are seeing, and how strongly they feel about it.

How do the fans see Holliday?  I know they thought he was bad last year (he scored a 31, which is around the same level that Dunn/Brown/Gibbons scored in 2006) but his zone rating on THT is 2nd best among qualifying left fielders.

A slightly above average LF, across the board.

Until I can finally get the HTML files uploaded on my site (definitely this week… just gotta find the time between the time my kid goes to bed, I sleep, and he wakes up), here’s the full results on Google:
http://spreadsheets.google.com/pub?key=pkimQBCeCjbjEqvnXANBEPQ

If UZR gets published, maybe the more resourceful readers here can get something cool going with that, the Fans’, and THT/BIS/STATS.

Note that the “ALL” is based on a weighted average of the 7 categories that is NOT position-specific.  The weights that I have used are:
1.26 Instincts
1.47 First Step
0.98 Speed
1.12 Hands
0.84 Release
0.77 Strength
0.56 Accuracy

Those weights were based on each of the 7 positions’ weights (unpublished, so far).  For example, for “arm strength”, it might be a 2.0 for catcher, but 0.2 for a 1B.  (Numbers for illustration only.) “First Step” seems to be something that all positions need to some extent, which is why it is so high.

So, when you look at the “ALL” column in the above link, it’s based on the above weights.  But, if you are going to compare to something position-specific, like UZR, then you need to use better weights.  (Like underweighting the arm categories for LF/CF, and overweighting them for 3B, etc.) Like I said, I’ll publish the position-specific weights at some point soon.  Just a pain that my home computer is at home, when most of my net postings are not done at home.  It’s hard to get as much done as I’d like.