fielding runs [Archive] - Baseball Fever

538280

08-17-2005, 08:26 PM

This is a post I made in the thread "Derek Jeter's Fielding" in the Yankees forum:

Fielding Runs are based on the player's fielding statistics at each position compared to the league average for this number of innings played. This is the basic formula:

PFR/(PO-SO for team)-LFR/(PO-SO for league*Player Innings

PFR is the player fielding rate, while LFR is the league fielding rate. The rates at each position are as follows:

1B:.2(2A-E)
P, 2B, SS, 3B: .2(PO+2A-E+DP)
OF:.2(PO+4A-E+DP)
C:.2(PO-SO+.4(A-CS)-E+DP+PB/2)

There are other factors added in to adjust for the handedness of the team's pitchers, and the amount of double play opportunites for the player.

To adjust for the handedness of pitchers, the basic formula is multiplied by this equation:

(1+ADJ*YF*DLHP)

ADJ is the adjustmen figure appropriate for the position and statistic. No adjustment is needed for catchers or center fielders. These are the adjustment figures for each positon and stat:

1B:n/a PO, -.4 A
2B:.23 PO, -.27 A
3B:-.22 PO, .34 A
SS:-.1 PO, .14 A
LF:-.16 PO, n/a A
RF:.09 PO, n/a A

YF is the year factor, necessary because this factory steadily increased in importance for 1910 to 1970. Before 1910 the YF is 0, so no adjustment is needed. The adjustment can be calculated for each year from 1910 through 1970 by subtractting 1910 from th ear in question and then dividing by 60. After 1970 the YF is always 1.

DLHP is the difference in the precentage of left-handed pitching from the league average.

DP opportunities were estimated from hits, BB, HBP, and HR allowed, plus errors committed. Using a multiplier for homers to account for DP opportunites lost did not improve the estimate in years where there are actual data: the average error in these years was around 2 percent. The formula for calculating DP opportunites is:

.662(H-HR+BB+HBP+.575E)

Individual DPs were divided by the team DP opp. divided by the league average DP opp. fro catchers, there is an aditioinal defensive calculation made to rate them on other defensive aspects. I won't go into that here, though, since we're talking about a shortstop.

I answered the question about how they are calculated, and I'll answer the others as best I can:

2.Positions that demand more defense (SS, 2B, C, 3B) tend to have more fielding runs than the outfielders and 1B, which is to be expected. You can't really compare defensive value between different positions by using fielding runs. Since fielding runs is part of Batter Fielder Wins or TPR, that really is a good thing about it, because it gives a positional adjustment for positions from which less offense is expected.

3.They really aren't all that accurate compared to other fielding metrics, but they are a decent metric if it's all you have access to.

4.The 2005 ESPN Baseball Encyclopedia has fielding runs listed for every player in MLB history, and in every season.

darkplague17

08-17-2005, 09:20 PM

Thanks for the info.

They really aren't all that accurate compared to other fielding metrics

So what are accurate metrics?

SABR Matt

08-18-2005, 01:22 AM

Careful...there's more than one type of Fielding Run.

The more modern FRAA/FRAR used by BP are quite a bit different than the ones described in this thread and involve team level adjustments and so forth.

Currently the favored fielding statistic is UZR...but that's proprietary information that can't be duplicated and hasn't been totally and completely explained beyond the theory and basic outlines.

Ubiquitous

08-18-2005, 01:29 AM

UZR explained by MGL himself part 1 (http://www.baseballthinkfactory.org/files/main/article/lichtman_2003-03-14_0/) and part 2 (http://www.baseballthinkfactory.org/files/main/article/lichtman_2003-03-21_0/), it can be duplicated. One just needs:
Retrosheet, freely available
database software
Excel

SABR Matt

08-18-2005, 11:32 AM

Ubiquitus...you canot duplicate UZR...not for all years in which it has been calculated...no one but an anointed few have access to 1993-1999 PBP

And I'm glad he's explained it more thoroughly than I had been able to find...those reports did not comeup on any net searches I attempted.

Ubiquitous

08-18-2005, 11:41 AM

But you can duplicate 35 odd years of baseball, and you can duplicate his work. Which is the most important aspect. The nuts and bolts are explained and one can do duplicate his work. Whereas there are other metrics that all we see is the end result and a brief description.

darkplague17

08-18-2005, 12:09 PM

good stuff, where can I find UZR ratings for players today?

Ubiquitous

08-18-2005, 01:12 PM

You will have to do them yourself.

csh19792001

08-18-2005, 03:16 PM

good stuff, where can I find UZR ratings for players today?

http://baseball-fever.com/showthread.php?p=359113#post359113

misterdirt

08-19-2005, 10:52 AM

UZR explained by MGL himself part 1 and part 2, it can be duplicated. One just needs:
Retrosheet, freely available
database software
Excel

Unless I am seriously missing something I don't believe the detailed zone information that MGL used to calculate UZR is available from retrosheet. He used STATS data which is not freely available.

SABR Matt

08-19-2005, 12:32 PM

I thought I recalled something about the zones he uses being more granular than the zones available in Retro files.

misterdirt

08-19-2005, 03:38 PM

In post 57 to the discussion of his Part 1 article MGL says he uses STATS data transposed to the retrosheet format (the old project scoresheet zones). But most retrosheet scoring doesn't include even the detail from the project scoresheet zones.

RGreener26

03-30-2006, 01:50 AM

Tango Tiger

03-30-2006, 05:16 AM

538280

03-30-2006, 05:29 AM

UZR by MGL (Lichtman), followed by PMR by Pinto (Baseball Musings), and John Dewan's Plus/Minus (Fielding Bible).

After that it's a chasm, and then it's all those without a PBP system, like Win Shares and BP's system among others.

Just curious, do you by any chance know how exactly BP calculates their fielding metrics? They don't seem to tell us anywhere on their site.

Tango Tiger

03-30-2006, 06:51 AM

argh

03-30-2006, 09:37 AM

What I don't get is why so many of the defensive metrics (even UZR and PMR, although I have yet to see many recent UZR's, has MGL released any?) disagree on certain players in certain years. There seems to be a much larger variance in agreement than say, metrics that measure offensive value.

I remember several years ago, someone (forgot his name) released DRA, which he said was quite accurate without pbp data. What happened to that?

SABR Matt

03-30-2006, 10:19 AM

Personally...although I know PMR is something that's had a lot of work put into it...I find its' defensive ratings to be extremely bizarre most of the time. Not trying to be rude, but the 2005 ratings for third basemen for example make no sense at all to me. They've got Adrian Beltre below average and significantly behind such luminaries as MELVIN MORA...(there are other bad fielders he's got ahead of Beltre...I don't remember them all off the top of my head).

And ultimately...my problem with UZR, PMR et al is that they don't really measure arm strength...just range.

digglahhh

03-30-2006, 10:43 AM

The most I've seen is in one of their annuals... I think 2002. Maybe it's more recent, I don't know.

The problem is that these non-PBP systems always try to estimate things that we can determine without a doubt with a PBP system. We have PBP data going back to 1960.

We can't always take the flyball/linedirve distinctions as gospel either. Many of those calls are subjective. Just intuitively, the positive connotation associated with a linedrive would lead us to consider a borderline fly/liner a linedrive if it falls in for a hit, and a flyout if it is caught. I think, as a result of this, hit %s on line drives and flyballs are upwardly and downwardly skewed, respectively.

Sometimes, for entire teams in the Retrosheet data, there will be like 8x as many linedrive singles as flyball singles. Intuitively, this can't be true. There is also substantial variation from team to team. Some teams will have no groundrule doubles while others will have, literally, forty. The Retrosheet data comes from more than one source, so we mix subjectivities.

A couple of us have just completed an exhaustive Retrosheet project that lasted a couple of months and I learned a lot about the data, that points to its inaccuracies, overall its still pretty good. If you were to sort the data as an Excel file by the play events, you would start to see how some events seem to be over-represented.

Tango Tiger

03-30-2006, 11:37 AM

I didn't discuss the subjective FB, GB, direction, location, but rather, simply the handedness of the batter. There's no reason to estimate that.

For the FB, GB, LD, Pops, sure there's some issue with maybe 5 or 10% of the data. But, again, why estimate this using traditional means, when you are far better with actual observations?

Chisox

03-31-2006, 05:59 AM

Hey everyone, new to the site! I was looking at Baseball Prospectus' Fielding Runs Above Average (FRAA) statistic and looking for any unusual findings. It actually has Torii Hunter at a -5 in 2003, -2 in 2004, and 4 in 2005. They then go on to project him as having a negative FRAA in the upcoming season.

Just curious to how this could be? I know a lot of players get overhyped, (Bobby Abreu winning a Gold Glove, etc.), but how is he so bad? Watching him, it appears as though he has nice range and a superior ability to catch the ball. Is he really that bad? In addition, how can a player face such a large swing from year to year? Unlike hitting, I would think fielding would be unable to suffer from slumps.

Also, what is the consensus on the best defensive statistics out there in addition to FRAA?

Thanks for any help.

I have no idea on how FRAA is calculated, but let me explain what I think might be the cause of the sudden rise in Hunter's rating.
Living in Minnesota all my life (except for two years in Arizona when I was one and two), I've gotten to see more Twin's games than I could possibly count, many more than my own team. Based on that, I can offer some of my perceptions. It at least seems like there are a huge number more hits that make it into the outfield than other parks, and the batters are able to hit them to all parts of the field, and certainly to the gaps more than most places, or at least the visitors can. My theory on this is that the Twin's have had ridicoulesly low walking pitchers on the team who don't strike out batters, either. Quick trivia question: which active pitcher has the lowest BB/9 ratio for their career? Answer: Brad Radke, far lower than Greg Maddux. Carlos Silva just set records for fewest walks. (I don't have the numbers available to me now, I'm just going by memory.) The balls always seem to be right over the plate almost saying "HIT ME", and no Twin's pitcher blows anything by any hitter.
The Metrodome with astroturf was very, very fast. Balls would hit out their and take off for the fence. Any ball that was hit with any authority in the gaps could easily make it to the fence. I saw countless balls that looked like lazy liners over the infielder's glove travel the near 400' that would easily be tracked down at any other park. As a result, the outfielders had a real headache (and workout) getting to balls that should have been routine. This wouldn't affect hits much, I don't think, but it would definitely have an effect on how many bases the hits would be worth. The astoturf was replaced (either before or after the 2004 season) with more natural feeling/looking/properties in general turf that should have slowed down the balls considerably.
One thing that could easily affect Twin's fielding stats is the dome itself, or roof. The roof is an off-white canvas with MANY holes in it. Just under the roof (probably "inside" the roof would be better) are shadows. Looking up at the ball can be VERY difficult, particularly if you lose site of it. Of course, you have lights and fans all over the place up there, as well. I would love to hear from someone who played there felt about it. Did Hiddengem play there at all?
I don't know if the stats would bear what I just said out; Matt or Tango might be able to know that, but it makes sense to me. BTW, in the games that I saw last year (which were not nearly as I used to see), I thought Hunter lost a step, which would make sense since he's older.

Mr. Red

04-30-2006, 06:08 PM

What is the best way to determine a fielder's value? And what is the formula? Thanks in advance.

538280

04-30-2006, 06:36 PM

Well, there are lots of ways and tons of metrics. Fielding runs is definitely not the one you'd want to rely on though.

Mr. Red

04-30-2006, 06:50 PM

Which one should I use then?

Ubiquitous

04-30-2006, 07:47 PM

none and all.

Mr. Red

05-01-2006, 06:30 PM

Can someone please provide a non-ambiguous, helpful answer? Thank you.

538280

05-01-2006, 07:12 PM

All right, some pretty good ones are Defensive Win Shares and FRAA (fielding runs above average also it's cousin FRAR, fielding runs above replacement, not the same as Total Baseball's "Fielding Runs"). They're pretty good, a good sketch, but suffer because they lack sophisticated PBP data like other metrics. Definitely much better than RF or FR, because they adjust for defensive context. FRAA and FRAR are available for every player in history at BP's website.

DWS aren't available for every player in history year by year, but the Hardball Times has them for some recent seasons. For historical seasons, you can check this list I posted here on BBF a long time ago for how DWS evaluates pretty much every player in history who played a significant amount of time:

http://www.baseball-fever.com/showthread.php?t=35689&highlight=Defensive+Shares+Letter+Grade+Ratings

For more advanced metrics that I believe use PBP data, there's UZR (Ultimate Zone Ratings), which I believe are available online somewhere but I'm not sure where, and DRA (Defensive Regression Analysis). Here's an article that explains that as well as some results for some players:

www.hardballtimes.com/DRA_2003.pdf

There are also regular zone ratings, which are easier to find than UZR but not as good.

Mr. Red

05-01-2006, 07:26 PM

Whats the formulas for FRAA and FRAR? Thanks.

pizzacutter

05-01-2006, 07:26 PM

538280

05-01-2006, 07:27 PM

Whats the formulas for FRAA and FRAR? Thanks.

BP (as usual) doesn't let us know what the formulas are.

538280

05-01-2006, 07:29 PM

Mr. Red,

The truth is that there aren't any really good fielding statistics, at least none that everyone can agree on. There's range factor, which is a measure of how many balls a fielder gets to, but getting there is only half the fun. Good reliable meaningful fielding stats are something that have eluded SABRmetricians since the beginning of the discipline.

John Dewan just published a book on the subject (The Fielding Bible), although I haven't read it. Perhaps that might answer your questions more thoroughly?

That's not really true at all...the world of statistical fielding analysis has increased tenfold in the last few years. We're now able to evaluate fielding fairly confidently, though of course still fielding stats aren't nearly as straightforward or sure as hitting stats.

pizzacutter

05-01-2006, 07:49 PM

538280

05-01-2006, 08:01 PM

Fairly said that there are plenty of measures floating around, but show me three guys who all agree on one of them and I'll show you a guy who's been cloned twice!

There has been a good amount of agreement that most defensive metrics for the most part are credible. They give us a very, very good idea of how valuable a player was on defense. The only problem is that they're not as straightforward as hitting stats. But, they still are accurate, and getting more so every day. In 15 years I suspect we may be able to reach the point where we can statistically evaluate fielding almost as confidently as we can statistically evaluate hitting. Progress in that area is being made all the time, and with DIPS theory we're getting more accurate answers.

Taco De Muerte

05-01-2006, 08:21 PM

I use a bunch of stats when evaluating a players defense - UZR, Plus/Minus, defensive WS, FRAA, etc. I used to like Rate2 - But Now I believe it's a piece of junk. I've moved on to UZR and Plus/Minus

Ubiquitous

05-01-2006, 08:24 PM

That's not really true at all...the world of statistical fielding analysis has increased tenfold in the last few years. We're now able to evaluate fielding fairly confidently, though of course still fielding stats aren't nearly as straightforward or sure as hitting stats.

Fairly confidently? Only in the minds of those creating the metric. The recording of fielding events is so inexact I can't see how any of them are reliable. Now possibly they might point you in the right direction, player A good, player B bad, but then again they also might point you in the wrong direction. So much of the recording of fielding events is a matter of opnion and even much more is not even recorded in a form available to most people.

BoSox Rule

05-01-2006, 08:41 PM

UZR isn't available to the public because MGL has been working for the Cards for about 2 years. However, he posts numbers on Baseball Think Factory occasionally. One day I think SABR Matt's work will be the best defensive work we have.

SABR Matt

05-01-2006, 10:49 PM

well...PCA defensive analysis is a step in the right direction in the field of SEASONAL fielding analysis ober the other seasonal metrics (FR, WS, TPR)...at least IMHO it is...though some here disagree.

But I won't be producing top-of-the-line defensive analysis on the WHOLE until I can encorporate some PBP information for modern players in some way...

UZR/ZR fail one important test IMHO...they're not top-down methods and defense is a team sport, making top-down analysis of it mandatory IMHO. I think PBP analysis can be incorporated into a top-down model...blend the best of both the seasonal and PBP fields into one approach.

538280

05-02-2006, 05:25 AM

Fairly confidently? Only in the minds of those creating the metric. The recording of fielding events is so inexact I can't see how any of them are reliable. Now possibly they might point you in the right direction, player A good, player B bad, but then again they also might point you in the wrong direction. So much of the recording of fielding events is a matter of opnion and even much more is not even recorded in a form available to most people.

All right, maybe somewhat confidently. It is true that defensive metrics have been able to agree on most things, though there still are some players where you can find here they're rated very good, here very bad, and here only okay. But, the metrics actually do agree quite often.

misterdirt

05-02-2006, 07:15 AM

Tango's fan survey is probably as accurate as the sabremetric numbers based fielding analysis at this point.

SABR Matt

05-02-2006, 07:27 AM

Assuming the fans he's surveying know a great deal about baseball...I suppose it's possible.

I am of the continued opinion however that it's very close to impossible for a human being to watch all of the fielders in enough detail that they can visually see who the good fielders are and who the bad ones are because the apparent visual difference between a good fielder and a fair one is so miniscule as to be unrecognizeable on a per-game basis, and even if you could identify it, it comes in the form of differences between making a play look routine (which would therefore make it forgettable in the human brain which tends to recall emotional extremes and not routine events) and making it look extraordinarily difficult (Jeter is the master of making the play but making it look ridiculously difficult in doing it on routine grounders into the hole or up the middle) and one step further...making it look like a routine base hit which we're also not able to recognize as a mistake by the fielder.

misterdirt

05-02-2006, 08:18 AM

Tango's survey is of Baseball Think Tank posters, a relatively sophisticated group of baseball watchers that know the difference between the flashy play and letting balls go through the hole or up the middle. Plus the whole idea of a survey is to aggregate opinions from a lot of people that watch a lot of ballgames. It is not meant to take the place of numerical analysis but to supplement it. Even the best numerical systems admit they can't measure the defensive contributions of catchers and first baseman accurately. They also differ markedly in evaluating right fielders. Most don't attempt to Park Adjust figures and the others use Park Factors designed for hitters. Even with all the work that has been published in the last year there are still many problems that need to be addressed.

west coast orange and black

05-02-2006, 08:47 AM

pizzacutter: ...show me three guys who all agree on one of them and I'll show you a guy who's been cloned twice!

haha. this leads into something that i have been meaning to ask.
if several different methods provide no clear-cut player as being the best in a particular defensive area, on what do you guys rely to settle the tie?

SABR Matt

05-02-2006, 08:48 AM

Wait a second here...I will cheerfully admit that rating catchers is difficult and still lacking...but my ratings of defensive first basemen are IMHO quite accurate indeed.

misterdirt

05-02-2006, 10:33 AM

Wait a second here...I will cheerfully admit that rating catchers is difficult and still lacking...but my ratings of defensive first basemen are IMHO quite accurate indeed.

First of all you haven't published your ratings or your system so they don't enter into the discussion. Second, how do you account for errors saved by good plays by the first baseman on bad throws by the infielders? This is a major part of the defensive value of a first baseman.

SABR Matt

05-02-2006, 10:42 AM

actually...not really, misterdirt.

a) First basemen don't really save that many errors...even the great ones...because their infielders are pretty competent and don't make a huge number of throwing mistakes relative to their chances to throw the ball.

b) Most first basemen are pretty good at catching bad throws because that's all they've been doing their entire lives...catching the baseball. That's what makes up a huge chunk of playing baseball even as a kid.

c) Differences in picking out bad throws in the dirt or saving high throws from a good fielder to a bad one MIGHT be on the order of a dozen or two (tops) per year...compared to the two or three hundred other plays they make, and some of those differences are caught in the analysis of the entire INFIELD'S defensive value when you work with a top-down approach that includes unit anlaysis (infield, outfield, catcher, pitchers). That's not to say I might not be missing a small element of fielding that could help us distinguish between Todd Helton and Keith Hernandez (Hernandez and John Olerud were both world famous for reducing error counts...Todd Helton is not despite being extremely skilled at everything else a first baseman does) that could be accounted for in the PBP era by doing something like the study done by TangoTiger on defensive catching in which he looked at the rates which the catcher allowed WP, PB, SB, CS, etc with each different pitcher he caught on the mound compared to what they did when they were pitching to any other catcher and found the differences (he ended up concluding however that over the course of the catcher's ENTIRE CAREER...he might have saved on the order of 10 runs...if he were really really good...I suspect a similarly stunted magnitude would be found in a detailed analysis of error rates.

Tango Tiger

05-02-2006, 11:29 AM

You have systems that uses just "seasonal" data, and you have systems that use "play by play" (PBP) data.

Systems that use seasonal data knows the number of outs (A, PO), but they don't know about the number of hits allowed by each player (but they know for the team). They also don't know the general direction that balls were hit in. They try to estimate these two by looking at the handedness of the batters faced (or handedness of their own pitchers), as well as trying to infer the gb/fb ratios.

Systems that use PBP data know for a certainty the handedness of the batters/pitchers. The GB/FB is fairly reliable, though there could be disagreement between a line drive and fly ball. The general direction of the ball hit is fairly reliable, though the exact slice has some room for error.

By definition, an intelligent PBP system would outshine a seasonal system. UZR, Pinto's PMR, and Dewan's Plus/Minus all are in a general agreement. They each have their own little thing that they also consider or not. For example, PMR doesn't include the distance of the ball traveled, but it uses the hardness of the ball hit (which of course is subjective). However, when you look at things over a period of years, all these breaks will even out. All these systems have the same data from the seasonal systems, plus the PBP.

ZR is kind of a poor-man's PBP system. It's one major shortcoming is that it doesn't consider the handedness of the batter, while the seasonal systems do try to infer it. For this reason, the ZR is sort of a half-and-half, a worst-of-both worlds. It's reliability would be anywhere from slightly below the seasonal systems to slightly above them.

***

I disagree with Matt's contention of the "failing the important test". They are fairly top-down, though not explicitly. Pinto's PMR is the most top-down of the PBP systems. In any case, the top-down approach is not a requirement.

***

My scouting survey is of hardcore fans (not just BTF, but from all over the web. SOSH for Sox fans, OriolesHangout, USS Mariner, Batters Box for Jays fans, etc. ) These are all crazy fans like you and me. I find great comfort in what the fans say:
http://www.tangotiger.net/scouting/scout2005_winners.html

You can click the "complete results" link. When you click on a player's name, you also get to see the most similar players to that player. Personally, I find it fascinating and illuminating.

Tom

Tango Tiger

05-02-2006, 11:33 AM

Which reminds me. Only with my survey did fans come out with such certainty on an obscure rookie to proclaim him one of the best fielding players in all of baseball: Betancourt of the Mariners. He outshined all shortstops, and almost all players. And Mariners fans have seen Vizquel, Rodriguez, Ichiro, Cameron. It's not like they don't know great when they see it. And they elevated Betancourt into that company.

However, there is not a single system in this world that would have proclaimed such a thing after 50 games with such certainty that the fans did. Maybe by the end of this year, UZR, PMR, and Plus/minus will come around. But, Fans already know.

SABR Matt

05-02-2006, 11:38 AM

Excellent point about Betancourt.

I'm a rabid Mariner fan as anyone around here knows...and Yuniesky Betancourt is comparable defensively to Ozzie Smith IMHO. And I don't think that's an absurdist thing to say.

misterdirt

05-02-2006, 11:47 AM

throws from a good fielder to a bad one MIGHT be on the order of a dozen or two (tops) per year..

That's actually quite a lot of value. A good play that creates an out and keeps the runner off first base (or possibly second or third on a bad overthrow) saves about the same amount of runs as a good play on a ground ball that keeps the batter off first base and creates an out. Since 85% of first baseman are between +-10 plays a year on ground balls according to the Fielding Bible, a swing of a dozen plays on throwing errors saved WOULD be a majority of the first baseman's value. And a + dozen plays saved in a year for a good full time first baseman over a bad one would not be extreme since it would be less that 2% of the throws that an average first baseman handles.

that could be accounted for in the PBP era by doing something like the study done by TangoTiger on defensive catching

I agree, that's exactly how this should be studied but nobody has done it yet. Hence, my original comment that there are a lot of problems that need to be addressed. I believe that such a study would show a much greater difference between first baseman than Tango's did for catchers since the run value of an error saved is greater than the run value of a passed ball or wild pitch. I also personally believe that the there is greater variation in the abilities of first baseman to prevent errors than there is in catchers to prevent passed balls. But that's what studies are for, aren't they, to prove or disprove opinions. Apparently Dewan thinks there is some merit in studying this as he is having his game watchers count good and bad plays on throws as part of his Defensive Misplays Study starting this year.

Tango Tiger

05-02-2006, 12:19 PM

Matt, I guess I should have noticed by your icon where your allegiance lies. And it's not absurb for you to say, since virtually all Mariner fans described him in lofty terms.

***

Dewan measures Scoops here:
http://actasports.com/sow.php?id=85

However, he didn't track opportunities. Eckstein would give Pujols far more chances to make a scoop than Everett or Betancourt would (I'm guessing). Looks like there's Pujols and Minky, and then the rest.

SABR Matt

05-02-2006, 01:06 PM

I don't buy that conclusion that first basemen are +/- 10 plays...the range delta-averages (average - actual) by PCA for plays made by first basemen (and average is fully adjusted based on the actual plays made by the entire infield...the bases allowed by the infield and the tendencies of the pitching staff) is more like +/- 60...not +/- 10. I know the fielding bible is being raved about right now, but if their conclusion is that first basemen only make 10-20 plays that matter...I don't even slightly buy it.

Tango Tiger

05-02-2006, 01:21 PM

I think the range was +/-15 to 20 bases for the 1B. This is also in-line with what UZR found.

I am in complete agreement with this. +/- 60 plays for a 1B? I don't even slightly buy it.

SABR Matt

05-02-2006, 01:36 PM

I should check that again...I was pulling the +/- 60 from memory only...

but +/- 20 plays just seems awfully awfully low to me. UZR/Fielding Bible's PBP et al. make certain assumptions about what a "zone" is that I don't think truly represent how a fielder plays his position...and you and I will continue to disagree on whether a top-down approach is necessary in fielding analysis, but fielding is not now, nor was it EVER an individual effort...players rely on each other on nearly every play they make and teams don't function on a player by player basis when playing defense. A top-down approach is the only way you can "see" this fact.

SABR Matt

05-02-2006, 01:49 PM

oh wait a second here...

No...my 60 figure came from plays above the MARGIN...not average.

*headdesk*

OK...maybe the +/- 20 isn't that far off from what my seasonally driven analysis would suggest. Sorry...LOL

It still feels low to me...but I'm guessing that's just because I think in terms of margins not relative to averages.

I'm as big a fan as any of encorporating PBP data...but I still think in order to convert that data from "plays" to "wins" properly you need to start from a top-down approach...I'm guessing you (Tango) would use a win expectency matrix, but I fear that would create weird biases for players on teams with a lot of high-leverage situations or very few such...and there may be other problems with doing that as well.

Mr. Red

05-02-2006, 02:00 PM

To rephrase my original ?, I'm comparing Cobb and Ruth for a math paper, what method can I use and obtain the formula? Thanks.

misterdirt

05-02-2006, 02:17 PM

I think the range was +/-15 to 20 bases for the 1B. This is also in-line with what UZR found.

3-Year plus minus rankings ranged from +53 for Teixeira to -28 for Sexson but Pujols was ranked 3d at +25. So as I said before 85% (actually 90%) of the first baseman averaged less than +-10 per year over the three years studied. Ryan Howard was a +19 for 2005 but didn't have the other 2 years.

Tango Tiger

05-02-2006, 03:43 PM

No, I wouldn't use win expectancy.

***

You don't need to use the top-down approach. The interdependence of fielders can be handled just like here:
http://www.tangotiger.net/catchers.html

This would work perfectly for a 1B, especially if either he or his fielders come and go alot.

***

dirt: if 90% of +/-10 plays, that would probably give you 1 standard deviation = 6 or so plays I'd guess. 3 times 1 SD will give you pretty much the whole range, so that gives you +/- 18 plays.

In my scouting report of 2003, I showed this:
http://www.tangotiger.net/scouting/scout_Results1B.html

The range was in the +/- 20 run range, which means +/- 25 plays.

***

Red: just use the data at baseballprospectus.com .

misterdirt

05-02-2006, 04:31 PM

You don't need to use the top-down approach. The interdependence of fielders can be handled just like here:
http://www.tangotiger.net/catchers.html

Tango - Did you read Matt and my earlier posts? We don't agree on much but the one thing we did agree on and already discussed was that the best model for evaluating first baseman would be modeled after your catcher study.

dirt: if 90% of +/-10 plays, that would probably give you 1 standard deviation = 6 or so plays I'd guess. 3 times 1 SD will give you pretty much the whole range, so that gives you +/- 18 plays.

Gosh! You know the probabilities of the normal curve, who'd have thought. Your statement adds nothing to what I've already stated. I used 85% on purpose because Matt said that first baseman's variations on catching infielder's throws would not be an important part of his overall fielding rating. I wanted to show that for a significant number of first baseman, (half of 85%), it would likely be the major portion of his rating. This is using Matt's low estimate that handling throws might be +- 12 plays a year.

Have you considered doing your catcher study for first baseman? Having the data would be a whole lot better than us sitting around and speculating on what it might be.

SABR Matt

05-02-2006, 06:29 PM

Hold on a second here...

Never would I suggest that the best model for rating the TOTAL fielding skill of a first baseman would be to do it the way it was done in the catcher study. That will only work for rating his ability to prevent throwing errors. It won't help you figure out how rangey he is...and that method cannot and should not be adapted to rate that fielder...or ANY POSITION'S total fielding controbition aside from possibly catchers as you showed...and that's because catchers and pitchers are so inherently related.

A first baseman cannot be rated for his range or throwing arm...nor can a SS or a LF or a 2B be rated for those skills or any other in the method Tango chose for his catchers study.