THE BOOK--Playing The Percentages In Baseball

I’m amazed that Paul LoDuca didn’t score lower.  “Inept”, I think, is the word best suited to describe his fielding this year.

I can’t remember if this has come up in past years’ discussions, but are the components that go into the Overall rating weighted differently by position?  I was anxious to see how Hanley would fare here (and I can not wait to see his PMR chart), and I’m simultaneously surprised and completely in agreement.  I think I had generally the same ratings for his components, but would never peg them as coming out to a 48 Overall.

In Hanley, you’ve got a guy who absolutely has the foot speed and arm strength to conceivably get to any ball necessary and make the throw—but he also has a lack of (looking for the right word...) dexterity, perhaps, that causes miscues on routine balls, and a tendency to make worse throws when he has time than when he’s working at speed.  I’m glad there’s some relative disagreement in the ballots, because I can absolutely believe people saw more of one Hanley than the other. 

But when it comes down to it, I just can’t imagine having a shortstop with 28 Hands and 23 Accuracy come out as a 48 Overall defender, even in the context of position-neutral.  I hope I’m making that clear—I know we’re going on skill sets and performance by players across the spectrum, but because of the position he plays, I just feel the sum is less than the parts.

I explained the position-neutral weights in a recent thread.  It was something like 1.5 weight for “first step” and 0.5 for “arm accuracy”.  And everything else in-between. 

Those would be the average weights for the 8 positions.  Obviously, some positions need arm accuracy much higher, and others need them even lower.

I really have to publish the position-specific weights, and will try to do so tonight.

Hanley being a 48 means that he has overall an average skillset, without looking at the need to leverage a specific skill for a specific position.

Once I publish the position-specific weights, we can see how he looks at each of the 7 positions (catcher excluded).

***

If you are looking for the most average fielder, it’s Adam Kennedy, Andy Laroche, and Chris Thigpen.  You can get to them all from here:
http://www.tangotiger.net/scouting/sim2007_6318.html

Kennedy has really plummetted from recent years.

***

Ryan Braun.  Every year, there’s one or two guys who are almost completely unique:
http://www.tangotiger.net/scouting/sim2007_8034.html

His closest comp, and he’s not even that close with a comp score of “53” (where most players have guys who comp in at “80” and above), is Reggie Abercrombie.  Basically, make Abercrombie slower and a bit smarter, and you get Ryan Braun.

Braun needs to become a LF almost immediately.  And a 1B very soon.

Here are the position-specific weights for each trait.  The average for each position is 1.0 (rounding errors, otherwise):

Ins ___ Fir ___ Spe ___ Han ___ Rel ___ Str ___ Acc___ Pos
1.3 ___ 0.3 ___ 0.3 ___ 1.3 ___ 1.3 ___ 1.3 ___ 1.3 ___C

1.6 ___ 1.6 ___ 0.4 ___ 1.6 ___ 0.8 ___ 0.4 ___ 0.4 ___1B
1.6 ___ 1.6 ___ 0.8 ___ 1.6 ___ 0.8 ___ 0.4 ___ 0.4 ___2B
1.5 ___ 1.5 ___ 0.7 ___ 0.7 ___ 1.5 ___ 0.7 ___ 0.4 ___SS
0.9 ___ 0.9 ___ 0.5 ___ 0.9 ___ 0.9 ___ 1.9 ___ 0.9 ___3B

1.0 ___ 2.0 ___ 2.0 ___ 1.0 ___ 0.5 ___ 0.3 ___ 0.3 ___LF
1.0 ___ 1.9 ___ 1.9 ___ 1.0 ___ 0.5 ___ 0.5 ___ 0.2 ___CF
0.9 ___ 1.9 ___ 1.9 ___ 0.9 ___ 0.5 ___ 0.5 ___ 0.5 ___RF

Feel free to offer your opinions.  Just make sure each line averages to 1.0.

***

This is how Hanley does at each position (though don’t pay attention to catcher):

Hanley
C 41

1B 43
2B 47
SS 46
3B 46

LF 54
CF 54
RF 54

As you can see, he’d make a fine corner outfielder.  As a SS, he’s -10 runs relative to the average SS.  As a corner OF, he’s +5 runs relative to the average corner OF.  The gap between the average SS and average corner OF is a bit less than 15 runs (I figure about 10 runs).  So, it makes some sense to get him out of the SS position.

Here’s Ryan Braun:
C 31

1B 37
2B 40
SS 38
3B 37

LF 46
CF 46
RF 45

As you can see, another fine outfielder.  He’d make an average corner OF, and he’s about 10 runs below the average 3B.  But, the gap between the corner OF and 3B is 5 runs.  So again, he needs to move.

BJ Upton, the elder:
C 62

1B 62
2B 67
SS 63
3B 64

LF 71
CF 70
RF 70

His skill set show him to be an excellent 2B, about +10 runs, and a super corner OF, about +15 runs.  That’s around a breakeven.

However, Fans made comments to me that it was very strange to rate him, as they didn’t see him the same way as 2B and in the OF.  So, he’s probably overrated in the measure in the IF.

Derek Jeter:
C 61

1B 52
2B 55
SS 54
3B 58

LF 53
CF 53
RF 54

He’d be a couple of runs above average at 3B, and about +5 in the corner OF.  3B makes a bit more sense for him.  At SS, he’s about 5 runs behind himself at 3B, which is breakeven.  PRetty much, you should put Jeter anywhere.

Rickie Weeks:
C 34

1B 38
2B 42
SS 42
3B 40

LF 52
CF 51
RF 51

About -10 runs as a 2B, and a few runs above average in the corner OF.  Move him already. 

Joe Mauer:
C 86

1B 73
2B 76
SS 76
3B 80

LF 68
CF 68
RF 70

Definitely catcher.  When it’s time to finally move him, 3B will be his spot, but let’s see what happens to his body by then.

That’s some really interesting stuff tango.  That type of data could really help teams decide what position is the best (and next best if he fails) for every player.  Those teams with this type of info would be able to find players and maximize their production at certain positions that other teams might not see with their own eye/subjective defensive analysis...without explaining the whole process, how did you come up with the position-specific weights?

Also, this type of data could be taken a step farther (I think MGL said the Cardinals did this) and you could assign a value to each trait per position (some traits would obviously be more valuable at different positions) and then come up with a linear weights number from that...this would allow you to compare to/and use in conjunction with UZR when calculating a player’s total value.  This would be especially important for those players with 3 years or less of experience.

I’m curious why LF has a slightly higher weight for speed than CF & RF. I’d figure speed would be much more important for a CF.

With middle infielders, why are hands and release/footwork essentially flipped? I would guess being the pivot on the DP goes into release/footwork, and a 2B is probably the pivot a fair bit more often...I was expecting release/footwork to be the highest-weighted component for 2B.

Think of things relative to itself.  In LF, you don’t care about the arm, but you care more about it in CF and alot more in RF.  So, even if speed is more important in CF than in LF, the arm is also more important in CF than LF.

If you want to drop the speed portion of LF down to say 1.8 from 2.0, then you’ve got to bring something else up (since the overall average must be 1.00).

However, and this is important, the CF makes about 30-35% more plays.  So, even if on a per play basis, the relative speed of the LF is more important (relative to HIS own other skills), the CF will apply it much more.

You could in effect multiply each of the numbers in post 4 by the following:
5: 2B, SS
4: 3B, CF
3; LF, RF, 1B

To get the overall impact on a game-by-game basis.  Those numbers, 5,4,3 are the balls in play each position is responsible for.  Similarly, you can multiply by 5/4, 4/4, 3/4, to try to keep things on an index of 1.0 scale (more or less).

From that standpoint, the LF speed, per play, is 2.0, but when weighted by the number of plays (three-fourths), comes out to 1.5 For the CF, he sticks to 1.9.

Hope that makes some sense…

As for the 2B, about 20% of PA are with a runner on 1B and less than 2 outs.  If you filter out K and BB, you are down to 15%.  If you filter out FB, LD, IFP, you are down to 7% or so. 

Just as a speedster can’t steal 1B, a secondbaseman can’t pivot unless there’s a groundball and a DP situation in effect.  Getting on base for the batter, and getting to the ball by the fielder, are far more important than anything.

Here’s Nate looking at several fielding stats with respect to Braun:
http://www.baseballprospectus.com/unfiltered/?p=609

It’s the same thing with Braun’s supporters, just as Roger Cedeno’s supporters, and Chris Duncan’s supporters, and Wily Mo Pena’s supporters, and Rickie Weeks’ supporters, and any other bad fielder who has some of the requisite tools, but not all: that just a little extra effort, a little extra coaching is all you need.  But, how often does this happen?  I mean, by definition, you should expect half of players to be below the median.  Are we all expecting that everyone below the median line would be so clustered to be just a shade below median, so that we can say they improved, but remained below the median?

Hope is not a plan.

The Brewers happen to have the 2B with the widest range of skillset (Weeks) and the 3B with the widest range of skillset (Braun).  That is, take the standard deviation of their 7 traits, and those two are at the top.  So, what does this mean?  This means that they have the skillset that can be the most leveraged, that putting these guys out of position will be the most costly.  And the worst position for Weeks and Brain is the infield.  Weeks needs to go to LF or CF.  Braun needs to go to LF or RF.  Corey Hart is ideal for CF. 

I don’t know how much they may have messed up Bill Hall.  His 2006 scouting report made him an ideal infielder (and expected a good outfielder too).  His 2007 scouting makes him look like a horrible infielder.

Has any team ever had so many players out of position?

That makes a lot of sense, thanks.

Just to get a bit deeper on the middle infield weights, can you explain why second basemen get a 1.6 for hands and 0.8 for release while shortstops get 0.7 & 1.5 respectively?

This is how I started the process:
Ins Fir Spe Han Rel Str Acc
4.0 4.0 2.0 4.0 2.0 1.0 1.0 2B

I begin with a “1” to the trait I deem the least important for that position.  I figured arm strength and arm accuracy is the least important for a 2B.  I relied on intuition and data.  For example, here’s the overall totals for 2B:

In Fi Sp Ha Re St Ac
56 58 57 53 57 50 57

So, of all players who are in fact 2B, they scored the lowest in arm strength.  No surprise there.  The rest of the numbers are fairly close.  Hands is the second lowest, so you figure I should give it a “1” or a “2”.  I gave it a “4” because I figure that a secondbaseman needs to be able to catch the ball.  If you look at the bottom 2B, you will see that where they rank the lowest is in the Hands category.

If I look at the standard deviation:

In Fi Sp Ha Re St Ac
15 14 13 18 16 11 16

You see that Hands is the highest.  This either means that we really don’t care about Hands, and so, they can have great or soft hands (the way you wouldn’t care about the speed of the catcher), or we really care about Hands, and it’s tough to find high quality here.  My intuition pushes me to the latter.

So, I go through the list, and give a 1, 2 or 4 for each trait.

This is how I see the shortstops:

Ins Fir Spe Han Rel Str Acc
4.0 4.0 2.0 2.0 4.0 2.0 1.0 SS

I bump the strength up compared to the 2B, and I swapped the hands and release, figuring that it’s more important for the SS to have the good release than the good hands.

The lowest rating of the 7 traits for the SS is hands at 56.  All the others are 58 to 60.

The S.D. was lowest for Speed, but I reasoned the opposite of the Hands for 2b: you need to have speed, so that’s why the SD is so low, and that’s why I think it’s important.

This was my basis for each of the positions:
In Fi Sp Ha Re St Ac Pos
4 1 1 4 4 4 4 C

4 4 1 4 2 1 1 1B
4 4 2 4 2 1 1 2B
4 4 2 2 4 2 1 SS
2 2 1 2 2 4 2 3B

4 8 8 4 2 1 1 LF
4 8 8 4 2 2 1 CF
4 8 8 4 2 2 2 RF

I would certainly welcome opinions to bump the numbers up or down.  They will of course not matter for the majority of players.  They will matter for guys like Weeks and Braun.

Tango, remember that subjective defensive evaluation system where people would rate each play as routine, moderate, challenging, and very difficult for range, hands, and throw?  Well couldn’t you apply similar trait-adjusted weights to each trait to come up with defensive linear weights numbers for every player using a type of subjective defensive evaluation system that used to be discussed on fanhome?

Yup, we sure can.  I suppose I can come up with a quick survey asking people to rank the most important to least important traits at each position.  Probably after 10 or 15 of you guys answer, we’ll have something fairly reasonable, likely more reasonable than my own numbers.

I thought so.  I’d be interested in helping with something like that.

Sean Forman was nice enough to send me the fielding numbers for minor league players.  Here’s what I looked for:
- all 3B with a career 500 chances or more, since 1992 (Braun had 735)
- who also played at least 1 game in MLB

That gives me a total of 194 players.  The fielding percentage of Braun is 189th best.  Aramis Ramirez is the only one close to him to give Braun fans hope.  (Second to last was Corey Hart.)

The only thing I notice is that many of the players (including Braun) would see little difference in performance between LF, CF, and RF.  All the numbers are close to each other for the players you gave examples of...seems like there would be a larger distribution for the OF positions.

Also, I thought MGL once told me this as a rule of thumb:

First basemen and LF are generally 11 runs above average compared to the league average batter, RF are 8 runs above average, CF’ers are average, third baseman are 2 runs below average, second baseman are 6 runs worse, SS are 9 runs worse, catchers are 14 runs worse and DH’s are 5 runs above average.

This is obviously just offensively, but if so, the gap is not 5 runs, but over 10, for Braun moving from 3B to LF offensively. Overall, the move to LF is probably the best move for Braun, but I don’t think that he will be gaining THAT much more value by moving to the corner OF.

He actually said it in this interview:
http://yankeefan.blogspot.com/2006/02/interview-with-mgl.html

I still thought he told me 3B were -2. Hmm.

I’ve revised my numbers slightly to account for the extra balls in play for CF, relative to the corner positions.

As for how to do the conversions, I don’t at all subscribe to the point-of-view in #17.  I go with the theory that you have pools of players, and that anyone can be a DH, and therefore, it is, to use MGL’s favorite word, preposterous, to count the DH at 5 runs while counting 1B and LF at 11 runs.

My spectrum is as follows in wins:
+1.0 C
+0.5 SS, CF
+0.0 3B, 2B
-0.5 LF, RF
-1.0 1B
-1.5 DH

You can search this blog for several threads on the subject.

Did you edit your previous post to have the revised numbers listed now?  I’m just curious if Braun’s numbers changed.

Here’s the Fielding Bible results, of which the Fans makes up 10% of the score:
http://actasports.com/sows.php

Comparing by position:

C - Molina: was #3 by the Fans, using the position-neutral numbers.  Would go to #2 otherwise.  A great choice.

1B - Pujols: was #2 by the Fans, so an obvious great choice.

2B - Hill: a good choice.  He was 6th with Fans, but very high.

SS - Tulo: was #6 by the Fans, but very close to #1.  You can definitely make a case for him.

3B - Feliz: was #4 with Fans, but close to #1.  An excellent choice.

LF - Eric Byrnes: nothing special about him, and a poor choice according to the Fans.  But, alot of the top LF didn’t even qualify to make the ballot (needed 500 innings by Sept 15 or so).

CF - Andruw: #4 and close to #2, but far away from Ichiro.  Would have been a good choice, if there wasn’t such a standout there already.

RF - Alex Rios: #4 and close to #2, but far away from Shane Victorino.  Same comment as above.

Did Braun’s numbers change that much or is he still about the same in any OF position?

Braun comes in at an average corner fielding OF (and 1B), and -10 runs in CF (and 3B).

He gets a 5 run boost by being converted into the OF.  He could survive as a CF for a year, but that’s not really what you want him to do.

The problem is that his skillset is so unique, I’m not sure my process can do him justice.  Even when I take out his three arm categories, he ends up with only 5 comps.  And his Release is by far worse than any of them.  He’s got Jerry Hairston as one comp (CF) and Vlad (RF) as another.  The other three (Kinsler, Frandsen, 2B, Lopez, SS… Lopez is the worst fielding SS in the league) all can at least throw somewhat.

He’s the bottom of the barrel, there’s no two ways about it.  There’s as much reason to believe in him at 3B as there is to believe in Weeks at 2B.

This means that the plan is “hope”.

I just found Justin’s post about switching Reds players last year:
http://jinaz-reds.blogspot.com/2007/11/where-should-they-be-playing.html

This is exactly how I would do it…

Very interesting application. Before I read Justin’s conclusions, I pasted his defensive chart for the Reds into a spreadsheet and played with it myself. If the offensive marginal value is included, and maybe convert the defensive chart to runs runs above average at each position, would be a neat way of finding the optimal lineup for each team. The Reds look like they were already optimized, as many teams probably are, but reading last years discussion of the Brewers, they were clearly a misaligned team.

Here are the weights of a similarly proposed system:
http://www.lookoutlanding.com/2008/8/9/590377/exploring-the-defensive-sp

You can compare that to my values in post 4 or 11.  I’ll check it out a little later today…