I’m not sure I understand what’s going on with Damon in LF.  What kind of weighted average of 29 and 16 results in -1?  Same thing with Swisher in LF.

Also, is Gardner 0 just because of the small sample size?

I’m not sure I understand what’s going on with Damon in LF.  What kind of weighted average of 29 and 16 results in -1?  Same thing with Swisher in LF.

I’ll re-check those in the morning.

Also, is Gardner 0 just because of the small sample size?

Yeah.

So we should be exactly 1 Bobby Abreu better on D…

Thanks, SG. Hopefully Jeter has really improved…

Don’t mean to linkjack, but this dude over at THT is optimistic on Cano:

http://www.hardballtimes.com/main/fantasy/article/smoke-and-mirrors-robinson-cano/

Teixeira’s numbers seem odd.  Was his defense really stellar last year after being slightly-below-average-to-poor the previous three?  Is he just sort of average, and we got sold a bill of goods re his glove?

Do this numbers include any adjustment for aging or possible improvement from young playes?  Actually, do young players usually improve their defense after a season or 2 in the majors? I don´t think it would matter much for the overall team defense, only 3 players in that group should see age related decline (ARod, Jeter and Damon), but a reasonable Jeter decline could mean a significant downgrade, and it is not like the Yankees would deal with it mid-season.

Actually, do young players usually improve their defense after a season or 2 in the majors?

I don’t have any links handy - I’d recommend perusing MGL’s postings at The Book’s blog - but MGL I think has studied this.  I think his current stance (it’s changed a little over the years I believe), is that defenders typically improve up until age 24-25.  Not count pitchers and catchers whom I think he hasn’t made any determinations on yet.  I’m gleaning this from some stuff he divulged when he spent a few days answering questions here, plus some comments I’ve seen from him on other blogs (FanGraphs, Neyer, etc).

IIRC, he also mentioned that 1B was one position that players kept improving on as they got older.  I’m sure there is a point where the improvement stops, but I don’t recall ever seeing a mention of it. 

I’ve thought of as well if there are other areas where we may see improvement past the age of 24-25.  E.g. if a player starts in the majors at a later age, or if after changing positions at any age if the player may improve after a year or two.  There’s also the question of if a player can truly improve their skill-set later in their career.  Depending on how he does this year, it’s possible Jeter has done it twice (though UZR doesn’t go back to the 90’s, ZR seems to indicate Jeter was worse then, than he is now).

Is he just sort of average, and we got sold a bill of goods re his glove?

Yes.

Also, is Gardner 0 just because of the small sample size?

Yeah.

I misspoke.  Or mis-typed at least.  He was 0 because of my math error.  Even regressed towards the mean, he was good enough last season in his very limited time to project as a plus going forward.  Take that with pounds of salt though.

Teixeira’s numbers seem odd.  Was his defense really stellar last year after being slightly-below-average-to-poor the previous three?  Is he just sort of average, and we got sold a bill of goods re his glove?

Zone rating thinks Teixeira’s a little better than this, he projects around +7 using that.  Also, Teixeira may be good at stuff that doesn’t get picked up by most fielding metrics, like scooping throws.  But odds are he ‘looks’ like he’s a great defender but is probably closer to average.  Still, that’s better than what was there last year.

Do this numbers include any adjustment for aging or possible improvement from young playes?  Actually, do young players usually improve their defense after a season or 2 in the majors?

Mike K. hits most of it, but to answer the question, no I didn’t include an aging factor.  In general, aging peaks by age 25, then starts to decline for most positions.  Decline is anywhere from 1-2 runs a year on average.  So we probably want to knock down everyone except Cano, Gardner and Melky by a run or two. 

My understanding is there can be some improvement in young players, especially OFers as they get used to the new sightlines of major league stadiums and the extra tiers/stands.  But I haven’t seen empirical data to confirm that.  You may also have players who move positions in the minors and may still be learning the new position by the time they get to the majors.  However, I imagine in many cases, you’re as good as you’re going to be defensively by the time you hit the majors.

However, I imagine in many cases, you’re as good as you’re going to be defensively by the time you hit the majors.

I’m curious if anyone has taken the time to study players that have “worked hard to become good defenders”.  E.g. the story with Wade Boggs was always that when he came into the league, he wasn’t a good defender.  But he worked every day on his defense, and eventually became an excellent defender (he won two GG after turning 30 after all!)(that’s sarcastic, BTW).

So I guess the question is, does the data back it up?  UZR doesn’t go back that far.  ZR misses the first several seasons of Boggs’s career (and isn’t on ESPN right now).  I’m not sure if TotalZone is available for 1987-forward.  I suppose I could use defensive Win Shares (best of what’s left?).  It’s just the type of thing that seems interesting, if there are players that have defied logic and really *have* become good defenders later in life.

Yeah, ZR only goes back to 1987 and by that point Boggs rated well.

If people can think of post-1987 examples I can see if the #s match the perception.

But it is entirely possible the Yankee defense will be average or (gasp) above average this year.

I certainly hope so.
Let’s hope that nobody else is afraid of the BIG BLUE WALL OF DECAPITATION (a.k.a. the tiny blue wall that frightened Bobby Abreu out of his socks).

I also hope that one day there will be pf/x-like data for more than just pitching.

^ I had a dream dammit!!!

I also hope that one day there will be pf/x-like data for more than just pitching.

My understanding is that Hit F/X is coming, which has a lot of good possibilities for fielding if it’s done right.

I think you are underestimating Cano.  It is interesting that in 07 when the Yankees had decent defensive play at first base Cano had his best defensive season.  In 06 and 08 when the Yankees had poor defensive first base play Cano played poorly defensively.  I hope that Teixeira is at least close to his defensive reputation and his good play improves Cano’s play.

I think you are underestimating Cano.

These aren’t my opinions.  This is what an objective set of data says.  I’m not underestimating anyone.

I do agree though, I’d expect Cano to play better than those numbers.  Standard zone rating says he should be a +3 next year, FWIW.  Then again, I’d expect Jeter to play worse too.

I think you are underestimating Cano.  It is interesting that in 07 when the Yankees had decent defensive play at first base Cano had his best defensive season.  In 06 and 08 when the Yankees had poor defensive first base play Cano played poorly defensively.  I hope that Teixeira is at least close to his defensive reputation and his good play improves Cano’s play.

Maybe, but for over half the season Cano was one of the better defensive second basemen in baseball and then had a brutal, brutal 2 weeks to a month that destroyed his defensive stats.

Yeah, ZR only goes back to 1987 and by that point Boggs rated well.

Thanks SG!  Now if you could just compare every player who started after 1987…I’ll have to give it some thought.  Jeter of course there was the big deal about how he worked on lateral movement and all, but we’ve only got a year of data so far to show he’s improving.  Hanley Ramirez I’ve heard some of the, “he worked hard on his defense”, but ditto him for a year of data.  Can’t think of anyone else right now.

My understanding is that Hit F/X is coming, which has a lot of good possibilities for fielding if it’s done right.

Not sure if you saw this on FanGraphs yet.  Looks like a good start.  With high-quality cameras (relatively) cheap now, I’m sure MLB will be looking to put them all over stadiums in the next few years.

I also expect Arod to play worse in the field.

Based on those numbers, the Yankees could go Damon-Swisher-Nady in the OF, at least occasionally (or if Gardner absolutely faceplants).  I dunno why Swisher’s projection is -8 runs instead of -11, though (he was -11 both times he played a substantial amount.  The positive number from 2006 is an obviously small sample fluke… why not just toss that?).

11 runs is 1.1 wins right?  14 runs (Gardner’s projection) is 1.4 wins, for a total swing of 2.5 wins.  What’s the win differential between them offensively, again?

I dunno why Swisher’s projection is -8 runs instead of -11, though (he was -11 both times he played a substantial amount.  The positive number from 2006 is an obviously small sample fluke… why not just toss that?).

The small positive in 2006 is not really skewing the numbers, it’s the regression towards the mean (league average) that’s helping him.  He’s probably closer to that -11 than a -8 if you believe the scouting reports.  Then again, given the margin of error here there’s probably not a ton of distinction there.  Figure he’s a win worse than an average CF defensively.

CAIRO says:
Gardner: .249/.332/.323, 65 BR over 650 PA.
Melky: .263/.319/.375, 69 BR over 650 PA.

CHONE says:
Gardner: .258/.341/.345, 73 BR over 650 PA.
Melky: .280/.345/.402, 80 BR over 650 PA.

So figure there’s somwhere around a half win offensive difference between them.

Umm…

But what about Gardner vs. Swisher?  I’m guesstimating 2.5 wins separate those two defensively (Gardner would be ~25 runs better in the field)... what about on offense?

But what about Gardner vs. Swisher?

CAIRO says:
Gardner: .249/.332/.323, 65 BR over 650 PA.
Swisher: .236/.347/.439, 87 BR over 650 PA.

CHONE says:
Gardner: .258/.341/.345, 73 BR over 650 PA.
Swisher: .247/.360/.454, 92 BR over 650 PA

So figure there’s something like 20 runs of offensive difference between them, or about two wins.  So it’s very reasonable that Gardner’s defense closes the gap.

Interesting.  So Damon-Gardner-Swisher, with Nady as 4th OF looks like the best alignment.  Provided, of course, that Gardner’s defense really is that good and he doesn’t just get the bat knocked out of his hands.

If I’m reading the splits numbers right, Swisher is actually a touch better vs. LHP, so a platoon with Nady doesn’t make sense.  Damon does have a bit of a platoon split, though.  So maybe Nady plays LF against LHP?  That builds in some rest for Damon.  You’d take a hit on defense, of course.  Damon would then be available to PH for Gardner late if needed.

You could also use Nady instead of Matsui against lefties. Having Nady and Swisher makes the team much more flexible.

Hmm, thinking about it some more…

Maybe against LHP they could go Damon/Swisher/Nady (or Nady/Damon/Swisher) in the OF.  It depends on whether Gardner can hit LHP.  The sample size at the ML level is tiny.  Are minor league splits available?  I don’t see them on baseball reference…

Good point about Matsui (though he’s respectable against LHP).

Matsui’s career OPS against LHP is roughly .800.  Damon’s is .750 (though Damon plays defense and runs the bases well).  Nady’s is about .850.

So… you could play Nady instead of Damon for a ~.100 OPS upgrade, with some loss of defense & baserunning, or you could swap out Matsui and gain .050 points of OPS and lose nothing at all defensively, and gaining a little something on the bases.

Figure to use the worst defensive CF when Wang pitches, no?

Provided, of course, that Gardner’s defense really is that good and he doesn’t just get the bat knocked out of his hands.

If forced to wager a few bucks on these propositions, I’d go with his defense really not being that good and he does just get the bat out of his hands.  I like the kid alright, but I’m just not feelin’ it.

Give me Swisher-Damon-Nady with lots of Cabrera and Gardner for late inning defense and rest for the regulars, including some of Damon at DH to rest Matsui.

Thing is, if I know I have to use Molina sometimes, I think I want it go be against LHP, because he’s utterly useless against RHP (and close to ok against LHP).

Then, depending on his health & effectiveness, you may want to DH Posada, because he hits LHP even better than Nady.  Health permitting, I’d be tempted to go Nady/Damon/Swisher in the OF, Molina C, Posada DH.  Against LHP only.  The perfect storm for this scenario has Wang on the mound for the Yanks.

Are minor league splits available?

From MILB.com.  Only one year’s worth and a little over a 100 PA, but he batted better against lefties last year.  While I don’t think he has a reverse-split, I also think it shows he isn’t overwhelmed by lefties.  Of course, I don’t know if any of the lefties there have the stuff of Jon Lester.

Thanks, Mike.  Considering he slugged .500 in those 100 PA it seems a tad flukey.  Most hitters have a platoon split, so I’d expect him to have one as well.

Right, then.  My idea lineup against RHP:

Jeter
Damon
Teixiera
Rodriguez
Posada
Matsui
Swisher
Cano
Gardner

Against LHP:

Jeter
Damon
Rodriguez
Teixiera
Posada (DH)
Swisher
Nady
Cano
Molina

If Po isn’t up for DHing when he’s not catching, then I’d go

Jeter
Damon
Tex
ARod
Swisher
Nady
Matsui
Molina
Cano

I know the real Yanks will not bat Jeter in front of Damon, but I think it’s better.

SG:
I always thought of Melky as an above average fielder. Is this telling us otherwise or is there an information gap due to him not playing more than 4 years?

“Give me Swisher-Damon-Nady with lots of Cabrera and Gardner for late inning defense and rest for the regulars, including some of Damon at DH to rest Matsui.”

I like this, especially given the injury concerns of Matsui. Having a better pitching staff may help alleviate some pressure on the defense so they can afford to have Swisher/Damon/Nady providing better offense.

I always thought of Melky as an above average fielder. Is this telling us otherwise or is there an information gap due to him not playing more than 4 years?

No system can tell us definitively if someone is a good or bad fielder.  I think Melky’s an above average defender too, but he’s one of those players where there’s a fair amount of disagreement in various systems.  Normal zone rating thinks he’s a plus defender.

You could have sample size issues here, you could have a problem with how some of his chances are being categorized, or you have a distribution of chances that favors one system or another.

I also wonder if after years of watching Bernie in CF, our bar for an above average defender is lower than it should be.

Off topic, but just following up with the idea of nonlinear interactions discussed in the last thread. This is a very, very toyish exercise, and I had a incredible desire to procrastinate and not do what I was supposed to do:

Consider 2 types of players, Player A is an all or nothing guy, batting .200/.200/.800, with an ops of 1. Player B never takes his bat off his shoulders except to foul balls, so that his line is .000/.500/.000

A lineup made of 9 type A players would score an average of .75 runs per inning, very good, but less than I thought a lineup of players opsing 1 would do. A lineup made only of type B players would score an average of .72 runs per inning, a little less so one could claim that a type B player, all other things being equal, is less desirable than a type A player.

But the thing is that a lineup of 8 As and a B will always average more than .75 runs/inning, no matter where the player B is batting in the order. Best case scenario, if he leads off the inning, we can expect .87 runs in the inning. This tells me that if a team has 8 type A under contract, a type B will have more value for that specific team.

As a side note, I always thought that teams wanting a “true leadoff” player or something of the sort was just stupid, just pick the best player available, but it seems that there are some cases where a specific ability is more desirable.

The thing is, most teams looking for a “true leadoff hitter” aren’t planning to bench one of their nine legitimate 3-4-5 hitters to get him into the lineup.

I know, and this comes from the very sketchy and toyish design of the exercise, but the point remains that, given who are the other hitters in your lineup, different skills may be more desirable, and not only from a lineup construction point of view, i.e. who bats leadoff or cleanup, but from a run production point. Also, even if 200/200/800 hitter really existed, would he be a legitimate 3-4-5?

Since I had some more time to kill, just to show how OBP is more important than slugging, if we had a type C player, a three true outcomes kind of guy with a line of .150/.400/.600, also opsing 1, a lineup of 9 such players would score an incredible 1.25 runs per inning (all previous calculations were done in the proverbial back of the envelope, so there may be some error, but the design is a simple Markov process so if anyone would like to repeat, it should be easy).

What the hell is up with Jeter? Awful to bad to abysmal to average?

Or are you tinkering with the numbers because you hate him? 

What the hell is up with Jeter? Awful to bad to abysmal to average?

Or are you tinkering with the numbers because you hate him?  LOL

SG doesn’t hate Jeter.  That would be “playa hatin’”, or so I understand.

The thing is, most teams looking for a “true leadoff hitter” aren’t planning to bench one of their nine legitimate 3-4-5 hitters to get him into the lineup.

Also, isn’t a “true leadoff hitter” usually just a speedy singles hitter? The idea being that he is more likely to score when he gets on due to his speed/baserunning skills. Or are we talking about an <ideal leadoff hitter who has both above average OBP and base running skills?

I know he doesn’t hate him, hence the lol-face. Isn’t it a long-running joke on this site that SG “hates” him?

I don’t know how to do a back-of-the-envelope Markov chain - if you care to post a sketch I’d be interested.  I would have written a little toy Monte Carlo.

Also I wasn’t sure what Bs do - they walk 50% of the time?

Also, a general question about the OPS=OBP+SLG formula - I thought that there was a standard, different formula for which this is only a shorthand approximation - is that a mix-up with a differently-named stat?  A glance at the web suggests the latter - odd.  This argues for 1.73*OBP + SLG, presumably environment-dependent of course.

Every sabermetrician—heck, everyone who reads sabermetrically-oriented blogs—hates Derek Jeter.  They curse him every time they strike out with yet another chick.  They have his face on a dartboard (in their mother’s basement) and use it for target practice every time he’s voted to the all-star team.  Ditto for Alphonso Soriano, David Eckstein and Darrin Erstad.  They also keep framed photographs of Adam Dunn, Jack Cust, Nick Swisher and Ken Phelps (I said it) next to their computers (in their mother’s basement), and routinely lay flowers in front of these.  In between Star Trek episodes.  When they’re not masturbating to hot alien chicks. 

That’s me, anyway.

If they routinely lay flowers, why do they need - oh, never mind.

My only question is which crappy defensive SS wins the Gold Glove next year in the American League now that Michael Young will probably move to 3B.

Rafael Palmeiro.

Jed Lowrie

Tits Magee

Tits Magee

good lord i did not intend to post that twice. website’s fault.

Rilkefan:
Players B walk 50% of the time, thats all.
If all players in the lineup either homers a with probability a, walks with probability b and makes an out with probability 1-a-b, you can make the 13 states chain where each state is determined by the number of outs and man on bases, and calculate the probability of transitions and run expectation,like:

E(1out, 1 man on base)= a[2+ E(1 out, 0 man on base)]+b(E(1 out, 2man on base)) + (1-a-b)E(2outs, 1 man on base).

Solve the linear system and get E(0,0) (Note that you can solve the 4 by 4 system with variables E(2,0), E(2,1), E(2,2) and E(2,3) independently, and plug the answers in the previous LS.)

The MC would work, and it would be easier to compute the run expectation for mixed lineups, but I hate programming, even a simple code like that.

Also, a general question about the OPS=OBP+SLG formula - I thought that there was a standard, different formula for which this is only a shorthand approximation - is that a mix-up with a differently-named stat?

I think you’ve already pretty much found it.  I remember about 10 years ago I started reading Neyer on ESPN, and OPS was a new-fangled stat then.  I recall the discussion on how modified OPS (IIRC, at the time some called it OPS+, before OPS+ is what it is today) was better than straight OPS.  Originally I saw some writings of using 2*OBP+SLG, and then 1.8*OBP+SLG.  OPS was the original (I believe Bill James) creation, and since then others have worked to refine it to give OBP greater weight.

SG can probably say better, but I don’t think modified-OPS has ever gained much traction, even among the really sabrmetrically inclined groups.  OPS is usually close enough that if you want a “quick” stat that’s fine, and if you want a more accurate stat look up the players’ wOBA.

even if 200/200/800 hitter really existed…

You’ve never heard of Dave Kingman?

even if 200/200/800 hitter really existed…

You’ve never heard of Dave Kingman?

Rob Deer also put up that line one year, no?

Even the 2*OBP+SLG is very much dependent of the enviroment.
A lineup of players batting 100/400/400 would score an astonishing run per inning, .98 to be precise, (and you wouldn’t call any player a superstar) and they have the same OPS+*/?? as the 200/200/800 players.

The point I found most interesting in the exercise is that different skills are desirable in different situations. It is an extrapolation of the data, and it may be intuitive generally but at least to me it was not all that sabermetrically intuitive, maybe if your team has a very good SLG you should value adding OBP more than if you have a team with a very good OBP, in which case you should pay a premium for SLG.

Even the 2*OBP+SLG is very much dependent of the enviroment.

I think that’s one of the reasons wOBA and similar metrics are used more.  IIRC, wOBA is basically Linear Weights (Base Runs) expressed as a rate.  And Base Runs I believe is accurate to within a few runs (on a team level), so it is pretty good at describing how teams go about creating runs.

maybe if your team has a very good SLG you should value adding OBP more than if you have a team with a very good OBP, in which case you should pay a premium for SLG.

Yeah, I think having some balance is a good thing.  You can’t score runs if guys don’t get on base, but even then it is hard to score if if guys behind them aren’t getting XBH.  Ideally of course you have players that are of the 300/400/500 (avg/obp/slg) type, where they do everything well.

I think that’s one of the reasons wOBA and similar metrics are used more.  IIRC, wOBA is basically Linear Weights (Base Runs) expressed as a rate.  And Base Runs I believe is accurate to within a few runs (on a team level), so it is pretty good at describing how teams go about creating runs.

The point about environment still stands though, because those linear weights are run-environment dependent.  A HR is only worth 1 run in an infinite run environment, but it’s worth more in a lower run environment.  The reason you can use wOBA and LW is that the run environment doesn’t change all that much from year to year or team to team.  In 2008, the range for R/G was 3.95 (Giants) to 5.56 (Rangers), but most teams scored somewhere between 700 and 800 runs (17 teams, AL teams scored more than NL, obviously).  So the “adjusted OPS” is still pretty close for most MLB situations.

OPS is the main stat I look at, guess I should start to switch.

SG can probably say better, but I don’t think modified-OPS has ever gained much traction, even among the really sabrmetrically inclined groups.  OPS is usually close enough that if you want a “quick” stat that’s fine, and if you want a more accurate stat look up the players’ wOBA.

I think Rob Neyer was trying to push SLOB at one point, (SLG times OBP).  But yeah, OPS is generally close enough if you’re doing something like player comparisons, especially if you don’t have the time or inclination to calculate or lookup a player’s wOBA. 

If you want to instead look at how many runs a player contributes to his offense, then you should move on to something like linear weights, which is why I think modifications to OPS never really took off.

I think Rob Neyer was trying to push SLOB at one point, (SLG times OBP).

Well, OBP x SLG make a lot more intuitive sense than OBP + SLG if you’re trying to model the interaction of the two.  The problem is that an individual player’s OBP doesn’t interact with his own SLG, it interacts with his teammates’.  Isn’t OPS really just a simplification of (OBP x (league SLG)) + (SLG x (league (OBP))? 

Even the 2*OBP+SLG is very much dependent of the enviroment.

Following the above, the simple-minded reason that weighting OBP more works a little better is that the league SLG is usually a good bit higher than league OBP.  So what depends on the environment is the multiplier you should use for OBP.  1.8 or 2 works pretty well in the current environment, but 1 would probably work better for the American Association in 1891.