Tangotiger Blog

Monday, July 22, 2013

Has any team’s worst fielding-outfielder played centerfield?

Most of the time, I'll guess say 60% of the time, a team's best fielder will be in CF.?  Another 25% of the time, he'll be in RF, and 15% of the time in LF.  But, I have to figure that the chance that the team's worst fielder will be in CF has got to be close to 0.

So, I'd like to see results to that effect.  First, in terms of observations for any given year, what is a team's UZR by LF, CF, RF, and are both LF and RF at least 10 runs ahead of CF.  (Remember the positional adjustment.  Since UZR centers everyone at 0 at each position, we need to adjust that.)  And in those cases that it does exist, is it simply random variation, or did a team truly put their worst fielder in CF?

Thinking about it, maybe old man Bernie Williams and old Man Junior would fit the bill.  I guess in those cases, the manager's hands may have been tied by something beyond his control.

Anyway, would love to see someone tackle this issue.

Similarly, that the SS is the worst of the three infield positions (2B, SS, 3B).  I'd expect that to occur with a bit more frequency, because those positions are less fluid than the outfield.  Beltre and Rolen won't move to SS, regardless how bad fielding the SS is, and how natural they'd make that transition.

Monday, June 24, 2013

Fielding positional adjustment

In response to a Bill James reader, I said:

Bill, I would just like to correct the reader who said: "Yet most analytical systems that use a positional adjustment would actually penalize Phillips because he shifted to the "easier" position of LF."

This is not a "penalty". From 1991-95, he played mostly toward the easier positions, whereby he totalled 30 runs above POSITIONAL average according to TotalZone, but his positional adjustment was -12 runs, for a total of +18 runs.  In the 4 years prior to that, he was only 10 runs above POSITIONAL average, but his positional adjustment was +8 runs for a total of... +18 runs. 

The positional adjustment exists because the player's fielding accomplishment is being compared to the average for that position.  So, in order to align the average corner OF to the average 2B/3B, we need to estimate that the average 2B/3B is about 10 runs better per year with the glove compared to the average corner OF, to compare them to a common baseline (and I use Willie Bloomquist if I need to provide a face).

***

The best way to present fielding, is to do fielding relative to position PLUS positional adjustment.  That's how you do it.

You do NOT do "hitting above position".  I know some of you out there do that, but it's a not-good way to do it.  And I've seen all the arguments.  Nothing you say will change my mind on the issue.

If you think like a scout, they won't say "he's a 70 hitter... for a SS, but a 50 for a 1B".  We don't adjust a hitter's OBP based on what position on the field he plays.  It's not harder to hit because you happen to play SS than 1B.  (Catcher and DH are exceptions.)

The idea of the positional adjustments is to show that the player's fielding talent does not usually change if you move him to 2B or LF.  His talent is his talent.  (Though in some cases, say Ryan Braun, it does.)

(And yes, I know that the Philips example was cherry-picked.  But if I used a non-face example, it would look empty.)?

Tuesday, June 18, 2013

Substitute penalties

?We noted in The Book that pinch hitters face a DRASTIC drop in talent level.  Andy followed that up to see which managers got hurt the most with ill-advised pinch hitting.

Indeed, an average hitter being used as a starter will have a wOBA of .341, while an average hitter pinch hitting will have a wOBA of .320. That's a 21-point drop in wOBA when pinch hitting (or, more precisely, a 6% drop), completely corrected for everything I can think of. For those who have read The Book, this is a familiar result. And indeed, we still find that the largest cause of this is a much higher strikeout rate (23% higher than when starting), though it's also true that fewer batted balls are hits.

Interestingly,substitute pitchers have the mirror effect: they increase their talent level (by improving their wOBA by some 25-30 points), riding mostly on their improved K-talent (17% higher as reliever), but also with improved BABIP talent (17 points lower, and thereby giving DIPS a black eye).  This was explained also in The Book, and later expanded in the blog a few years back under the "Rule of 17".

Also in The Book, we noted there was a DH penalty.  I looked at this in the blog as well, showing there was a 16 point drop.

Somewhere on the blog, I also showed there was a penalty for non-starting hitters, who were not PH or DH (i.e., late-inning fielding subs).  I think it was close to 10 point drop in wOBA.

While we may have exhausted the penalties for subs on the hitting side, Pizza comes along and looks at penalties for fielders who move positions on the fielding side.  He relies on the cruder fielding measures, and he finds no difference.  But, my prior is that there should be some difference.  So, if MGL has it handy, or Ben at BIS, I'd love to see how the advanced metrics look here.

And while we're there, also look at UZR for late-inning fielding subs.  Surely they should be a bit worse than as starters?

Thursday, June 13, 2013

Minor league catcher framing

Max continues his look by applying his great work to the minor leagues, and finds that the skill can be established there as well.

Catcher framing numbers in the minors seem to predict those obtained in The Show fairly well: a weighted correlation coefficient computed on the data points shown above, comes out at a healthy 0.58. How does that compare to other statistics?

Max: what was the average number of opps per catcher??

Tuesday, May 28, 2013

Who do UZR and DRS not agree with?

James gives us the top 10.  I think if you just went on the sniff-test, you'd probably side with one of DRS or UZR on 6 or 7 of the players, and you'd side with the other on the rest.

Hence, if you are going to make this political, you'd show which ones you disagree with to disparage one, and which you agree with to amplify the other.?

Monday, May 20, 2013

Brandon McCarthy: one excellent fielding catcher nearly as important as seven excellent fielders

He was asked a question he's never been asked:

On whether he’d rather have an excellent pitch framer and an average defense behind him, or an average pitch framer and an excellent defense behind him: “I would probably always rather have…I don’t know what way I’d answer that question, because when you have someone where it’s not working back there, you’re super annoyed. And every time you turn around and the ball, someone doesn’t get to it or something happens, then you get annoyed with that. So I want them both. I would still probably say defense, but I could go back and forth on that.”

If McCarthy had Beltre and Simmons and Trout and on and on (seven of them), that would be a much bigger advantage than Lucroy or a Molina.  I think you can make an easy case for the excellent pitch-framing catcher to ONE excellent fielder.  I doubt you can make a case for the catcher v two excellent fielders (but I guess you can try).  But, one catcher against three, much less seven, excellent fielders?  No, I think I'd have to stop you right there. 

I'd be interested to see you guys work out the problem based on the percentage of PA that are BIP, as to how little a pitcher has to rely on his fielders in order for the excellent catcher to be more important than seven excellent fielders.

Kudos to Ben to get him to talk about it.  ?

Thursday, May 16, 2013

WOWY catcher-framing, since 1988

Finally!  As you know, I'm pretty big on the With Or Without You (WOWY) method, especially as it relates to catchers.  I did it for WP, PB, BK, SB, CS, PK, and I added:

I'm not including blocking the plate or framing the pitches, though that last part might be doable ?

It took some ten years, but Max has not only taken on the challenge, but really went all-out.  First, he did the right thing by comparing the PITCHf/x method to the Retrosheet method.  Basically, how much do you lose by not knowing the exact location of the pitch?  Seems to be: you don't lose much, especially as your sample grows. 

With that established, we have the treasure-trove of data going back to 1988 via Retrosheet.  And Max gives us the career and seasonal numbers.  Just tremendous effort, and very well-done.  One of the best research pieces of the year.

(Max: will you be posting all your data? 

As well, why did you choose "per 5000 pitches" to represent a season?  Just taking one catcher at random, Piazza was averaging 8.4 called pitches per inning, so if we go with say 1200 innings to represent a regular catcher season, that's 10,000 pitches.  Wouldn't that be the better seasonal-standard?)

 

Wednesday, May 15, 2013

Catcher article to end all catcher articles

Whoah, one of the best combination of sabermetric-speak and non-sabermetric?-speak I've ever seen.  Too much good stuff to quote, so I'll just take this quote from catcher Chris Stewart:

"Within the last two or three years it's taken over as one of the highly sought-after skill sets for a defensive catcher," he says. "The sabermetrics stuff coming out, they put a value on it. We actually have a number for it. It's not a 'This guy's good' or 'This guy's bad,' it's like, 'This guy's this good' and 'This guy's this bad.'"

 

Tuesday, May 14, 2013

Marvin Miller’s FBI file

And yes, be prepared to read the word "communist".?

Adding metrics of different uncertainty levels

Colin brought up the following:

It turns out that there is in fact a very large body of work about how to add two quantities of varying reliability. It involves estimating the amount of confidence (or uncertainty) you have in each estimate. It has use in things like polling, for example. And it can certainly be applied to measures of fielding; we’ve used it to regress our FRAA, for example, to account for random error in our estimate of fielding opportunities. It’s nontrivial to apply this kind of thinking to metrics like DRS and UZR that aren’t built from the ground up to be used this way; there are a lot of sources of potential systemic bias in those metrics that make it much more difficult to regress them based on the amount of statistical power you have. But there is absolutely a way to combine quantities of differing reliability together such that you’re putting more weight on the more reliable estimates. And if other WAR implementations were to do so, that would fix the problem Heyman is bringing up.

This is something that MGL has repeatedly addressed with regards to the UZR component of WAR.  I seem to remember him making a long article out of it at Fangraphs, and I know he's mentioned it a few times at our old blog.  (I'm sure someone out there can google it for the rest of us.) So, MGL is in agreement with Colin.  It's just that very few people are listening to MGL on this sub-issue.

The basic idea is this: with hitting stats, we know the outcome.  And while we know the number of opportunities, we don't know the quality of those opportunities, but we can easily estimate it (by looking at the parks, the opposing pitcher and the opposing fielders).  And in any case, the quality of those opportunities are going to be in a tight range.  Even if Larry Walker faced mostly bad pitchers at Coors in the high scoring era, the quality of his PA might be a wOBA of .400 for an average hitter, and even if Jose Cruz (the dad!) faced mostly good pitchers at the Astrodome in a low-scoring era, the quality of his PA might be a wOBA of .250.  Or, if we go to two decimal places: .40 and .25.  (And that's about as exagerrated as I can make it.)  And that's for 4 PA per game.

With fielding stats we also know the outcomes, but we have very little confidence in the quality of those opportunities.  First off, we have nine fielders.  At least six of them won't even be involved at all.  And it's possible that none of the nine are involved at all.  So, the chance of making an out could be .99 or .01, and we don't know for any of the fielders what those numbers are.  That's why we love zone stats, so we can at least start to zero in on the fielders who MAY be involved.  Maybe now we're at the .10 to .99 range for the SS/3B, and we're at the .00 to .01 range for everyone else.  But it's still quite a range.  And we've got 27 BIP each game to worry about.

Colin's point is that we're very unsure about the quality of those opportunities on the fielding side, compared to the quality of those opportunities on the hitting side.  So, when we start adding up the various components, we shouldn't just do a straight add.  He is right.

Where it REALLY gets complicated is that now, the seasonal numbers won't add up at the career level.  Indeed, the game-by-game numbers don't add up to the seasonal numbers.  If you were to apply a methodology at the game level, chances are virtually every player will be close to 0 runs.  But, when you have a collection of games, you add up the numbers first, and then you reduce it based on the uncertainty level of the metric (which depends on the sample size).  And so, once you get at the career level, you apply the same process.

And no one does that (though they should).

Basically, the point then becomes at what point do you decide that you want things to add up, and when don't you want things to add up.

***

Which brings us back to MGL's preferred methodology, and that is to basically just be interested in the true talent level of the player.  Regardless of two hitters who both have the same wOBA in 2013 and the same UZR in 2013, if one guy has shown a propensity for being great, then count his 2013 performance in a more favorable light.  To put it plainly in an exteme example: Felix's perfect game is more real than Armando's or Humber's.  So, regardless that they all got their 27 outs in the eyes of everyone (whether in real time or after the fact), Felix's 27 outs are more indicative of his true talent.  His perfect game is worth more.

MGL basically bypasses the discussion of the measurement error by merging the issue of the measurement error with the issue of random variation of performance.  Hence, he simply presents the true talent level of the player at point time T.  Ultimately, he doesn't care about giving out lottery tickets to players.

Which goes back to perhaps Colin's greatest line, and it went something like: a game was played where 54 outs were recorded, and we learned nothing new about the players.

***

Anyway, so there are a few angles for people to tackle the issue, and they need to decide what it is they are interested in.  What they are EXACTLY interested in.  Otherwise, we're all just running in place on the issue.  Except for MGL, who has wisely figured out over a decade ago at least, to just focus on true talent, which is really what the front office, the scouts, and everyone outside of fandom cares about.

?

Thursday, May 09, 2013

Andrelton Simmons

Mark once again highlights Simmons'... well, highlights.?

Saturday, May 04, 2013

Intro to FIELDf/x

?This was an article that originally appeared in the Hardball Times Annual 2011, co-written by Greg, and posted on Alan's site (pdf).

Wednesday, April 17, 2013

Multi-position player

?I'd like to hear some suggestions as to what to call a player that plays multiple positions (Zobrist, etc), other than "utility".

Tuesday, April 16, 2013

Crowdsourcing BIP

Matt would like to do it.  Sky tried it a ?few years ago, and the results looked promising.  And here's a related thread.

I also know we did it as a crowd, together via a chat.  It was a Phillies World Series game, but I don't remember which one.

Anyway, the key is to determine whether to presume the player's initial position, or not.  And, I think it's easier to presume his positioning.  (That is, don't give the positioning skill to the fielder, but to the pitcher.) 

Monday, April 01, 2013

Defense Value

?I didn't realize (or remember) that Fangraphs combined the fielding relative to average and the positional value.  This is the best way to put defense of all players along the same singular scale.  As you'd expect, the leaders are filled with C, SS, and CF, 2B, 3B.  Brett Gardner's LF performance is given its due, as is Ben Zobrirst, and if you keep paging through, you'll see Heyward.  This is what the positional adjustment does: it puts everyone along the same scale, so that you can have a singular list.

You see something similar when you look at the Fans Scouting Report: Ichiro, Heyward, and Alex Gordon are the first names to pop up from the non-premier positions.

Friday, March 22, 2013

Positional Utility leaders

?The McEwing list is out.

Thursday, March 21, 2013

Positional adjustments

This is a good post from Dave, that talks about the positional adjustments.

It bears repeating that the positional adjustments I use are guides, and not hard-and-fast numbers.  The important part in Dave's article:

Mark Smith wrote about the evolution of the catching position two weeks ago, noting that the recent trend has catcher offensive performance spiking while caught stealing rates are steadily falling, which might suggest that teams are trading have been more wiling to trade defense for offense behind the plate in the last few years. Just from an anecdotal perspective, you could find support for that idea in the crop of starting catchers being handed jobs this year — Jesus Montero, Wilin Rosario, John Jaso, and Tyler Flowers are all bat-first catchers that have poor-to-awful defensive reputations, but they’re all penciled in as big league regulars for 2013. Of course, there are also some pretty terrific defensive catchers at the beginning of their careers as well, and there have always been bat-first catchers in Major League Baseball. Just pointing at those four doesn’t prove that teams are definitely shifting towards offense behind the plate.

In a nutshell, if teams have decided to trade defense for offense at a position, then the positional adjustments need to be altered.

It's important to remember what the positional adjustment does, and again, quoting Dave:

For those who aren’t aware, the positional adjustments are a necessary component because defensive metrics calculate a player’s performance relative to his peers at that position, but an average defensive shortstop is almost certainly a superior defender to an average defensive second baseman. Same for center fielders versus corner guys, and third baseman versus first baseman.

If for example you have a league in Canada that is filled with LHH, teams may decide to shift their talented fielders from 3B to 2B.  Suddenly, 2B might be as important a fielding position as SS, and 3B might be slightly more important than 1B.  And we can tell based on the fielding talent of the players we find at those positions.

The problem is that we don't know the fielding talent at those positons because every fielder is being compared to his positional peer, meaning the average at each position is... well, average.  We need some independent way to put them on the same scale.

I did this based on players who play multiple positions, and seeing how their performance compares to their peers.  This is no different than say park factors or league factors.  If you get better numbers as you go from park A to park B, or league A to league B, or position A to position B, then that means your competition got worse in situation B compared to situation A.  (Change in your own talent notwithstanding.)

Another way you can do this is look at offense at that position.  If you see a sudden rise in hitting at a position, that COULD be because you traded defense for offense.  But, that's not a given.  If ARod enters the league, you are not trading defense for offense necessarily.  If he's a stud fielder AND a stud hitter, then where else would you put him?  Same for Junior, Trout, Mantle, Mays, etc.  So,  you have to be careful in ASSUMING that an increase in offense is because of a decrease in defense.

You could look at the Fans Scouting Report.  That's a good way, as long as you can feel comfortable you can handle the inherent observer bias.  After all, it might be hard to put the fielding talent of 3B and 2B on identical scales, given that they don't show their skills in the same way or the same frequency.

You can look at how players move along the spectrum, and see if players move more or less frequently from position A to position B to position C over their careers.

There's a dozen ways to look at this problem to try to find a solution.

Friday, March 08, 2013

Rawlings, SABR, and a trademarked fielding metric

Well, this is new:

As part of the multi-year collaboration beginning with the 2013 season, SABR will develop an expanded statistical resource guide that will accompany the Rawlings Gold Glove Award ballots sent to managers and coaches each year. In addition, SABR will immediately establish a new Fielding Research Committee tasked to develop a proprietary new defensive analytic called the SABR Defensive Index™, or SDI™. The SDI will serve as an “apples-to-apples” metric to help determine the best defensive players in baseball exclusively for the Rawlings Gold Glove Award and Rawlings Platinum Glove Award selection processes. The collaboration also installs SABR as the presenting sponsor of the Rawlings Platinum Glove Award.

I'm more interested in the proprietary aspect of this metric.  Who owns it, who's going to develop it, what are the data sources being used, etc.  Is it going to be black box, and so a matter of "trust us", or is it going to be transparent? 

By and large, black boxes don't sell.  It requires an enormous faith onto the person or group developing it.  For something like UZR, we've got MGL, and PECOTA/ZiPS/Oliver has Nate/Dan/Brian.  But, they've at least given us enough background to kind of go along with how it's working.

If this new metric is going to be more closed than that, then I don't see why anyone would buy it.  But if it's going to be more open then that, then bring it on!  I'd love to see it.

Monday, February 25, 2013

Does outfield alignment matter?

Dave is asking if it really makes a difference if you put the better fielder in the corner, and the worse fielder in center.  It really depends how big the gap between your fielders are naturally.   I think it's sufficient that anything other than somewhat close fielders should have the better fielder in CF, and let the other guy set his ego aside.

To me, Trout/Boujos is more interesting, as it reminded me of Beltran/Cameron.  In the Mets case, Beltran was there for the long haul, while Cameron was not.  So, even if you wanted to argue that Cameron was the better fielder, I'd argue for Beltran in CF because you don't want your mainstay to be swapping positions every year.

Friday, February 22, 2013

Best fielders, ever

Nice job from James.