THE BOOK--Playing The Percentages In Baseball

Didn’t we talk about this in the other thread and found that the going rate for a FA win is over $4M?  Maybe we didn’t but that’s what my research shows.

I’m not sure.  I’m just using an old number from Tango (2.5).

And of course, as I always say, if the spread of FA salary is large enough, based on players’ perceived (by the teams) marginal run value (or actually whatever it is based on) AND if teams are terrible at evaluating talent, then a smart team should be able to get a marginal win for subtantiallty less than the “going rate.” IOW, the going rate does not necessarily mean much of anything when evaluating smart or dumb contracts.

Basically the large the spread in FA salaries and the more inefficient teams are at evaluating talent, the more a smart team is able to get bargains and the less they have to pay below the going rate for marginal wins.  This is a very important concept.

Which is worse, Andruw or Torii Hunter for $12 M?

Sounds like a game of “dumb or dumber”.

Jones is a much better hitter, about +2 wins above Hunter.  They are a wash fielding-wise.

Hunter is probably +3 to +4 above replacement, and Jones is +5 to +6 above replacement.  I guess I’m in the minority, and I think Jones at 10-12 million is a good deal.  I give him +2 wins above the average hitter, +1.5 wins above the average fielder, and +2 wins for the average player above replacement.  That’s +5.5 wins.  That’s worth about 12 million.  Today.  Chop off about half a win a year, so say he’s +5 wins next year, +4.5 after that, and +4 after that.  So, 3 years, 13.5 wins, or about 30 million$.  With just this little thought, that’s what I’d give him.  3/30.

Tango, where do you get 1.5 wins above average?  I know you are including a positional adjustment, but I think that is way too high!  The average CF’er over the last 5 years has been an above average hitter (+1 in lwts).  Actually that is in Superlwts (baserunning included).  If Andruw is an average CF’er, I would find it hard to believe that you are giving a CF’er +15 runs in fielding talent, but they are around even in hitting.  That would mean that teams really screw up where they position players.  I know you don’t like using average offensive value at each position as a proxy for positional adjustment, but given that they should be around the same as a “real” positional adjustment (other than for catcher, since that is so specialized and not fungible with other positions) and given that your appoximation for “relative positonal value” based on players who have switched positons is admittedly crude, I don’t know how you can use anything else BUT offensive values to estimate positional adjustments.  In any case, 1.5 wins is ridiculously high for an average CF’er.  If that were the case, then almost everyone who can remotely play CF SHOULD play CF.  If an average CF is 1.5 wins above an average generic fielder, what is a RF and LF?

Or are you assuming that Andruw is still a true +10 or +12 in CF in UZR?  I think that is not even close.  His UZR’s over the last few years are barely average.

While Andruw indeed is a much better hitter than Hunter and it is not even close (as you said, around 2 wins), I think that he is almost a win better in defense, and a few runs, almost a half win, better in speed.  So that brings the offensive gap down from 2 wins to .5 to 1 win.

If you have Jones at 1.5 wins in defense (and again, you have to tell me at least what that means and from where you get that), what is Hunter?  I think you have this all wrong.  Since you have Hunter overall 2 wins worse, I guess you have them equal in defense and baserunning.  I repsect that there is some “disagreement” even among good analysts, but that is ludicrous (Jones being 2 wins better) for the reasons stated (defense and baserunning), regardless of where you get the 1.5 wins in Andruw’s defense from.  Other than bias in favor of Andruw who probably once was a great CF (when he was 19-22 years old), I don’t think there are too many people who will not say that Hunter is a better (I think MUCH better) defender.  His UZR’s have been much better over the last 4 years.

So what defensive numbers are you using to say that they are equal defensively and why are you ignoring baserunning (albeit less than .5 wins, but still worth a mil or so)?

There is another huge problem with assuming that an average CF is worth 1.5 wins in generic defense.  I amassuming that is what you mean by those 1.5 wins.  Or even 1 win (and then assuming that Andruw is .5 win better than the average CF’er).  Tell me how you can overcome this problem.  If I find a no-name or minor league player that is an average defender (and given the ages of minor league players, I don’t think it would be all that hard) in CF but a terrible hitter, replacement level, you would say that he is worth 1.5 (or 1 win) over replacement because he is an average defender in CF (and a replacement hitter).  They would then be worth almost 3 mil on the FA market.  That is ridiculous!  I can get all the replacement level hitting average defending CF’ers for nothing from the minor leagues!  Guys like Logan and Gathright are all BETTER than average at defense in CF and are not even as bad as replacement hitters on offense.  According to you (again, assuming that is what you meant by the 1.5 wins), they would be worth millions, which is absurd.

I am sure that an average defender in CF is worth .5 wins in defense after a positional adjustment, at best and probably closer to zero.  SS is another story.  I think you got hung up on your last study of players who swtiched positions, which seemed to suggest that OF’ers may be more important than IF’ers.  I don’t think that is right.  I don’t think you can find too many SS and 2B that are average in defense and better than replacement level in hitting for free, but that you can find plent of outfielders, including CF’ers.

Let me see if I understand this.  We expect, through the natural order of balance, that all positions (except maybe catcher) should all have an off+def value of around the same.

We certainly know this is not true in high school, and mostly not true in college, somewhat not true in A, AA, possibly not true in AAA, yet, by the whole balancing act of reaching the pinnacle, MLB, we know expect equilibrium to be reached, such that the average CF and average 2B, off+def, are equals.

Now, we know that the average hitting CF is around average.  So, for this equilibrium to take place, the average CF must be an average fielder.  Right?

I’m going to give you a list of CF, and you are going to tell me if they are better than average, or worse than average fielders.  (Not at CF, but compared to all fielders.) Here we go:
Ichiro, Kotsay, Edmonds, Matthews, Beltran, Andruw, Wells, Anderson, Cameron, Torii, Patterson, Granderson, Taveras, Sizemore, Rowand, Baldelli, Sullivan, Reed, Damon, Figgins, Pierre, Gathright, Coco.  That’s 22 CF.  Which ones would you put in the below-average fielding-wise, pile?

Brady Clark, Junior, Byrd, Abercrombie, Lofton.  5 CF, all are likely below average.

Can you tell me that out of those 22, you can find 8 of them as being below-average, overall-fielding-wise?

And, even if you do, surely the 14 best CF are much better than the 14 worst CF are bad, no?

To me, there’s no question, that equilibrium has not been reached.  There should be more Endy Chavez, Gary Pettis, and the like.  My research clearly shows that the fielding contributions of the average CF is on par with those at SS.

In essence, but certainly not to the same degree, CF is like a QB in the NFL.  There is certainly no equilibrium point in the NFL, because not just anyone can play QB.  As a better example, in the NHL, there are 3 forward positions (LW, C, RW).  However, if you were to draw up the list of the top 30 forwards in the NHL, 15 of them would be centers.  Or 25 of the top 50 forwards.

Same thing here.  Since CF has an average hitting value, then if I were to draw up a list of the top 49 fielders in all of baseball (excluding catchers), we’d expect 7 of them to be CF, if we have equilibrium, and CF is the average fielding position. 

I however believe that you’ll find almost double that amount, and an equal number for SS.  You’ll probably get 7 or 8 at 2B and 3B each, and 2 or 3 at LF, RF, 1B.

I don’t really get you argument.  You keep asking about these players’ “fielding ability” including all fielders.  That is circular.  I have no idea how they would do at SS, etc.  Please read this next paragraph carefully.

You are missing a point here Tango.  The only way to use a positional adjustment other than the difference in average hitting is if almost all players can go from one positon to another and your positional adjustments still hold.  If that is the case, and you can show that with the data, then you can use whatever the data show.  If not, then you HAVE to use the offense to make the adjustments since making adjustments any other way is by definition trying to compare quantitatively apples and oranges (or apples and shoes).

Now, even if you can show that fielders who move from CF to SS retain the same UZR that STILL does not mean that their positonal adjustments are the same.  What if there are only a relatively few CF that can play SS and that if you moved a random CF to SS, he loses 10 runs in UZR?  What then?  What are your adustments?  What if only a few say, 5%) of all CF can move to SS and have the same UZR and vice versa (5% of all SS can move to CF and retain the same UZR)?  What is your positional adjustment then?  Are the two positions equal even if one positon hits a lot better than the other?  What about if a CF who moves to SS stays the same in UZR, but a SS who moves to CF gains 5 runs (a strange put plausible imbalance)?  What then?

Your CF’ers are obviously great athletes with great speed.  So what?  Does that make them good, great or even adequate SS?  I have no idea.  You want me to look at that list and say off the top of my head, “My gosh, they are all great ahtletes, they must be able to play SS?” I have no idea if that is true!

Even if we concede that CF’ers are the best athletes on the field (using some combination of subjective and objective criteria), does that mean that they must have the highest positional adjustment?  Not if they can’t play SS adequately!  Just because someone is a “great athlete” does NOT mean they can play any particular position adequately.

What about catchers?  Where do they fit in in your positional adjustment scheme?  Most of them are obviously not good athletes.

IMO, there are only two things you can do to generate positional adjustemnts.  The ideal way, and they way that really only makes sense is to figure out what happens if you take a player at one positon, any player, not just the ones who are suited to that other position, and move him to another postion.  What then happens to his UZR?  Because the only reason for positional adjustments is just that (moving players around the defensive spectrum).

In my research and yours on UZR and players moving positions, I just have not seen a whole lot of solid evidence that CF is equal to SS and there are not enough players who have moved to make inferences about “any old” player.

Right, my premise is about the ease of moving players around.  Catchers are excluded because of the non-ease.  My research shows a huge dropoff when a non-catcher plays catcher.  Experience here is golden.  Not so at the other positions.  Ask Soriano.

And my basis is two-fold.  One is the interpositional changes.  I think we can all agree that it’s rather easy to compare CF to LF and RF, and there is an abundant of such player movements.  There is also alot of LF/RF to 3B/2B.  There is also alot of 3B/2B to SS.  And LF/RF/3B to 1B.  So, we can construct something fairly strong, while also accounting for the selective sampling (only certain players get to play multiple positions), to show that the fieldign spectrum has SS and CF at the top, followed by 2B/3B, then LF/RF, then 1B.  To put CF *below* 2B/3B is in my opinion, and to use your words, preposterous!

The *only* saving grace you have is that the 2B/SS/3B must be RH throwers.  To that end, that puts a crimp in what I’m saying.  Putting a LH thrower at 3B, who may be pound-for-pound as good as a RH 3B would probably cost 10 runs.  That is, a guy who is +5 as a RH 3B may be -5 as a LH 3B.  And who knows, that may be conservative.

If instead we limit the pool to RH throwers, then, it’s possible that CF drops from the SS level to below the 2B/3B line.

The other basis is the Fans’ Scouting Report.  Jsut as in my interpos comparisons, the SS and CF are neck-and-neck for best fielders.  And it’s not even close.

Again, if CF and SS were equal, then you should be putting a lot of those CF’ers who don’t even play full time at SS.  They would be average or above overall at SS, since the average SS hits around 11 runs worse than the average CF.

Why does that not happen?

Guys like Gathright and Logan would be making 8 mil a year at SS and guys like Crisp would be 12 mil a year players.  Damon would be a 20 mil a year SS!

I agree that a positonal adjustment should be made based upon what any given player would do at each postion, but as I said before, I think it is really hard to get a handle on that, and from what I recall from your research the numbers were by no means definitive.

I certainly agree that SS and Cf are in the same ballpark.  As to why the average CF’er hits so much more than the average SS, I don’t know.  Those things fluctuate all the time.  Maybe we are just in a high CF/low SS cycle now.  Or maybe it is an inherent bias/inefficiency in the way scouts and teams look at the fielding spectrum and fielders.

One questions.  Actually set of questions.  What would happen if all of the positions had such unique skills that if a player moved from one position that he is good at to any other position, he always gets worse (sort of like catcher)?  Would there be a positional adjustment?  How would you calculcate it?  Would you then use offense for that?  If there were more or less players who were suited to each position, wouldn’t there then be a difference in each position’s average hitting talent, even if no one position was inherently harder than another?  Or are you automatically defining a position as harder than another if less playes are suited for it?  Couldn’t that be what is happening between SS and CF?  Even though CF requires more (or the same) “athleticism overall” as SS, there are still fewer players with the skills to play SS?  If one position hits less than another, unless the teams really screw things up, isn’t it a fact that there have to be fewer players available with the skills to play that position?

Or what if the skills that make one suited for a particular position correlate with good or bad hitting, independent of how athletic you have to be at that position?  That is partially the case, is it not?

Lots of tough questions to answer I think, which is why teams simply need to weight all of their options rather than try and automatically compare the value of one player to another when they play different positions.  In fact, it is really not necessary for a team to care about positional adjustments, as far as I can tell, so it is no more than an intellectual exercise anyway.  Personally I don’t care who is the better player, Sizemore or A-Rod or Pujols.

Well, the simple answer here is that you’re both wrong! Mitchel is correct in saying that the adjustment should be based on offensive numbers, but it should NOT be based on the average lwts at a position, but rather the replacement-level. And then you’ll find that a replacement-level CFer is indeed worse than replacement-level players at most positions, and be able to adjust accordingly. This removes the bias of having a few great players at a position, and tells us what we want to know anyways—a player’s value above replacement, not average.

Yes, David you are 100% correct!  Positional adjustments, at least in terms of valuing players, are completely determined by the offensive replacement level at each position for an average defender of course (or, more correctly, the total offense plus defense of a replacement player at each position).  That is true by definition, since the only way we CAN value players is knowing how much over replacement AT THEIR POSITION they are worth.  It has nothing to do with how much harder one position is than another, although in a perfect world, I suppose that is going to determine replacement level at each position.

But I am NOT wrong about using average offensive values at each position in the major leagues!  That is what determines replacement level, or at least you can get one from the other! In fact, that is the best way to determine replacement level.  It is too cumbersome to try and find out the offensive levels of freely available talent.  If the average offense at SS is -9, then replacement level is going to be around -25 to -30.  If the average major league CF is +1, then replacement level is going to be -15 to -20.  That is why using average offensive levels in the majors is by definition THE way to determine positional adjustments and NOT some method of determining which position is harder than another.

I’ll add one thing in your favor.  Average offensive values for each position in the major leagues does not of course completely determine replacement level.  As you said, a few really good players can screw that up.  Maybe using median is better.  But either way, it will get you close to estimating replacement level at each position.  In fact, it might be the best way (using mean or median in the majors) since, as I said, it is not that easy (on a practicle level) to determine replacement level by looking at freely available talent.

But you are right that positional adustments in the context in which we are discussing them (basically valuing players) are completely determined by the total offense and defense combined of a replacement player at that same position and that does not necessarily have anything to do with how hard a position is, although obviously it is strongly related.

Not much to disagree with there, and certainly using the median would help, but you can determine replacement level by looking at freely available talent. It’s difficult to define FAT, but once you do that, all you have to do is take their projections and use those for your positional adjustments. It’s possible that this isn’t worth the effort, especially if using the median gets you most of the way there (not sure that it does), but certainly center fielders should not be marked down; I think we can all agree that CF is not a hitter’s position.

I’ll see “you’re wrong”, and raise you a “double wrong”.  If you treat each position as MGL is trying to ask, that you would lose so much value switching (like to catcher), then fine, you would use the repl level.  However, that’s not how it works, especially in the outfield.  There is no such thing as a “replacement level CF”, without considering the OF in totality.  You have replacement level OF, and you need to do positional adjustments between CF/LF/RF.

The same applies for 2B/SS/3B.  Each position in its group draws on the exact same pool of players.  In some cases, a player belongs to multiple pools (say like Soriano).  Even Adam Everett would belong to both pools (IF and OF), because his fielding traits are perfectly suited for the OF (if Manny is in the pool of OF, then pretty much every single 2B/SS/3B is in the pool of OF also).

Like I said, the LH OF throwers are NOT part of the IF pool.

Now, if each player was suited only for one particular position, then you have an NFL-type situation, and you compare Manning to the repl level quarterback, and you compare TO to the repl level receiver, etc.

Btw, I think that 2B are the least-compensated position, salary-wise, right?  So, I think teams are well-aware that not all positions are “equals”.  Of course, you can argue that teams overpay for hitting and underpay for fielding.  However, I don’t think it works out quite that way.

***

And how to handle the DH?  The off level of DH is *lower* than 1B and RF.  Rather ludicrous, isn’t it?  That if Ortiz was being compared as a 1B, he would look *worse*!  The repl level of the DH is every single MLB.  That’s the pool to draw from.  For 1B, the pool is almost every single MLB.

Stop treating each position as if it were sacrosanct, like an NFL QB.  It’s much more like the NHL C, LW, RW.  If you can play C, you can play LW or RW.

And in the NHL, a forward (C, LW, RW) is pretty much a different pool than a defenseman.  It’s rather rare for a player to play in both groups.  That’s completely unlike MLB, where IF/OF swaps are done, practically every week for each team.

Let’s deal with the reality of the similarities and non-uniqueness of MLB, compared to the NHL and NFL.

Tom,

I don’t necessarily disagree, but your defensive positional adjustments are too biased. A guy who plays a lot of time in both CF and LF will by definition be at least a decent CFer; otherwise he wouldn’t play that much. I agree that there’s no way the positional adjustment for CF should be to subtract one run from lwts, but defensive positional adjustments are not the way to do it. Unless…

I suspect the error bars would just be too high this way, but what if you looked at only guys with a small sample size at one position and a large sample at the other? Then you’d be looking at essentially emergency replacements for teams that don’t have a backup plan. Guys like Trot Nixon in CF. Then, you’d be looking at a more or less random group of players. If somehow you could get the error bars low enough, I’d be much more willing to accept those.

Agreed.  When I did the adjustments, I broke down the guys into “primary CF”, “primary LF”, “neither”.  I then looked at how each of these groups did at both CF and LF.  I showed how there’s a definite “experience” bias, so that guys who are primary at one position lose about 4 runs when switching to a “similar” position (like one OF to another), and if they switch to an unsimilar position (like OF to IF) they lose 8 runs.

There is plenty of data to work with (for me anyway, since I’ve got the UZR data), so it’s just a matter of spending the required time.

I haven’t looked at the handedness issue, and it may turn out that what keeps the CF from being on top is that all the good-fielding LH throwers in baseball are competing for one spot (CF), while all the good-fielding RH throwers are competing for four (2b, ss, 3b, cf).

While I think it’s interesting to know how many runs a player gains or loses when shifting positions, I agree with MGL (and David, I think) that this isn’t relevant to a player’s value.  Suppose by chaining this data you could establish to everyone’s satisfaction that the average first baseman would be -60 runs at SS.  Does that mean the average SS is “worth” 60 more runs on defense than a 1B?  No.  The average SS’s (or repl SS’s) defensive value over a 1B is exactly the difference in their offensive production—nothing more or less. 

Let’s say that a repl 1B is +15 runs on offense vs. a repl SS, and both provide approx. average defense at their position.  I think Tango would accept that assumption.  The two players have equal total value—zero—by definition.  The 1B is +15 runs on offense, so mathematically he is -15 runs on defense.  It doesn’t matter that the 1B would really cost you more runs if forced to play SS (of that the SS might be more than 15 runs better playing 1b)—that’s irrelevant.  The value of something is what we pay to get it, and the “cost” of acquiring average defense at SS is a loss of 15 runs on offense. 

I don’t see how the fact that players can play multiple positions changes this.  We value a player based on what position he does play; if he moves, we use a new benchmark: repl level for that position. 

* * *

One small caveat:  someone at BPro (Nate Silver?) did research suggesting that replacement shortstops—and only shortstops—provide somewhat below average defense.  I thought that was interesting.

Continuing with your 60-run example: if the average fielding SS, when moved to 1B, would be +60 runs, that becomes very relevant.  What happens?  If you have two SS in your system, one of those guys gets shifted to 1B.  This will happen until you get to the point where you reach equilibrium.  That is, you shift over so many SS to 1B, so that the gap in their fielding matches their gap in their hitting (the other way).

The presumption, among you guys, is that this equilibrium point has been reached.  That all the shifting has been done.  There’s no reason to think that it has.

Let’s get back to hockey, because this is the most relevant to the discussion.  50% of all star forwards are centers.  Would you compare your center to the “replacement level” center, or to the “replacement level” forward?

If in the NBA half of the great players are centers, and the bit players are forwards, do you compare your center to the repl level center?  No way.  It’s understood that you have a pool of players to draw from.  You are compared with that pool.

And, also do the same for HS baseball.  You wouldn’t, would you?  That’s because equilibrium has not been reached.  Your best players are concentrated at key positions.

"The presumption, among you guys, is that this equilibrium point has been reached.  That all the shifting has been done.  There’s no reason to think that it has.”

Well, AN equilibrium has been reached:  the average and repl level offense for positions is pretty stable.  So the market has established a stable price—in the form of runs lost on offense—for current MLB-quality defense at various positions.  We don’t know if that equilibrium is perfectly set, but I’d say the burden is on you to show there’s a market failure (and maybe you can).

For example, it sounds like you think there are excellent defensive CFs whose defense is so valuable they should be playing over some current CFs.  Shouldn’t this be a testable proposition, by identifying some backup OFs you feel meet that description, and then assessing their def and off true talent?  Admittedly, that gets tricky if you’re talking about AAA players.....

Not necessarily over current CF, but over current LF and RF.

***

We haven’t reached an equilibrium, but a stable level.  It’s not equilibrium.  (May be semantical, anyway.)

More to the point, discuss why you would (or would not) compare a college CF to a college replacement level CF.  And LF to LF, etc.

And then, repeat that for High School.

"We haven’t reached an equilibrium, but a stable level.  It’s not equilibrium.  (May be semantical, anyway.)”

What happened to your faith in the “wisdom of crowds?” The allocation of MLB players to positions, given their relative def and off ability, has been pretty stable for decades (except for increase in offense, and presumably some decline in defense, at SS).  These outcomes reflect the collective judgment of hundreds of scouts, managers, and executives (and even players and agents) at the minor and major league levels over many years, in very competitive environments.  They could be wrong, but I’d think you would give it a bit more deference. 

In any case, who are the backup CFs who should be regular LF/RFs?  What are their UZRs and linear weights?  You may be right, but given the higher offensive production at the corners, I’m skeptical.....

* * *

“More to the point, discuss why you would (or would not) compare a college CF to a college replacement level CF.”

I would, but I actually don’t understand the point you’re making!

Of course I agree 100% with David and Guy (and myself) on this.  The bottom line is if I start with a team of replacement players, which is the baseline because I know 2 things, one, how much it will cost (say, 500,000), and two, the hitting level of a replacement player at each position (with a little effort), I can value players at any position on an even keel, simply by looking at how much better or worse they are than those replacement players.  That is the ONLY way to value players.

Tango’s point, and it is valid, is that I can do better by shifting players around the defensive spectrum.  I can’t do that if teams are 100% efficient in doing that already.  Tango is suggesting, I think, that they are not, and that there are plenty (or at least SOME) of players who are not playing the “correct” position.  I don’t know if that is true or not, and I don’t know that we can ever figure that out from the UZR data, because there is too much selective sampling and there are too many unique skills at each position that are not fungible.  Even though baseball defensive skills may be more fungible than football ones, certainly they are not 100% fungible - not even close.

Bottom line I think is that even though it is not perfect and it precludes us from doing the optimal thing, using some offensive measure (major league mean or median, or actually going through the troublke of figuring out replacement level) for each def. position is probably the best way of doing the positional adjustments.  Trying to use the UZR data to do that will yield some unworkable results I think.  For example, if you insist the CF and SS have the same adjustment, whatever that is, since major league SS are -9 in offensive lwts and CF are +1, you HAVE to say that a replacement level, average defensive SS is 10 runs less in offense than a replacement level, average defense CF.  I don’t think that is a tenable position.  If that were the case, then you would be much better off having replacement level CF, because you would likely be paying too much for anything above replacement level in CF.  If that is an inefficiency in baseball, so be it. 

IOW, if the best you can get in freely available talent in CF is -20 offensive lwts and SS at -30 (both average defensively), which I think is the case, then go with the CF, right?  Tango would say that you are getting a bargain in CF to the tune of 1 win.  Or something like that.

As far as other levels, like HS, or even LL, where the best hitters and fielders are at the same position, I would have to think about that.

And I think that the first thing we have to do is to EXACTLY define what we mean by a “positional adjustment” and what the purpose is.  As with many things, the whole argument could be centered around the fact that we are not defining our question properly.

Tango, how are you defining “positonal adjustment?” Precisely.

Guy, “wisdom of crowds”: is that the same crowd that would put Mark Belanger as a #2 hitter, and no one would think that’s odd?  Not *every* crowd is wise!

***

Ok, forget MLB.  How would you evaluate a high school CF, in terms of his production, and compare him to a high school RF?

Are you going to concede that the average HS CF is much better than his corner buddies?  (off+def)

Are you also going to concede that the average HS SS is much better than his 2B partner? (off+def)

***

As for defining the positional adjustment.  Let’s suppose that we have the “average” fielder, a guy who is in the middle of the pack in all characteristics.  I’ll come up with some names, and you choose whoever it is that fits that best: Chipper, Trot Nixon, Jason Lane, Ryan Church.  Someone like that. 

Let’s say that this player, if you put him at any of the 7 positions, will make an out 80% of the time, in his “zone of responsibility”.  So, just define each position’s zone, so that this is true.

Now, take the actual average fielding CF.  Clearly, they will make an out more than 80% of the time, on average, in their zone.  Let’s say this number is 85%.  The question then is: how often will be make an out in the typical LF zone?  Let’s say this number is also 85%.  It doesn’t have to be, since the types of balls won’t be the same.  Let’s make it easy for us, right now.

The next question is: how often does the average LF make an out in his zone.  Let’s say that the average LF will make an out 77% of the time in his zone.  It should of course be less than 80%, since the average LF is clearly one of the worst fielders on the team, and not one of the best (on average).

So, that’s our positional adjustment.  The average CF is +.08 better than the average LF on balls in zone.  And, that’s what we care about.  If the average LF has say 400 balls in his zone, then this average CF would get to .08 x 400 = 32 more balls than the average LF.  (All numbers for illustration purposes only.) So, about 25 runs.

Well, we’ve got a problem here.  Why in the world is the LF such a bad fielder? 

If this is high school, we’re not worried.  There’s not enough ball hawks out there.  We expect an imbalance.  We expect our HS CF to be like the QB.  But in MLB?  It shouldn’t happen.  (Theoretically, if Darwinism was alive and well.)

So, that’s what you have to do.  You have to figure out how a poor, average, and great player at each position would do at each of the other positions.  Or, similarly, have this player, this Chipper Nixon Lane player, play at each position, and treat him as the baseline, and compare everyone to this baseline.

(Handedness, as noted, is an issue.  Perhaps a big one.)

You need a common baseline to determine if you are indeed in equilibrium.

Endy Chavez of the Mets might be the perfect example here (if we look at his career record, not just this year).  He is a replacement-level hitter.  A guy who is just about the minimum production you’d want, if he was an average fielder.  He is in opinion a great fielder, one who is something like +20 to +30 runs above Chipper Nixon Lane.  It matters naught if the average CF is +10 or +25 above the Chipper Nixon Lane line.  Because, Endy would find a home somewhere, be it CF, LF, or RF.  Once that line is drawn, that Chipper line, that’s where you compare the players.

Now, granted, it is much easier to draw the line for LF/CF/RF, and another one for 2B/SS/3B.  And, it’s tricky to compare these two lines.

But, at no time do I even consider that every position has reached equilibrium.  Nor is it required.

***

Finally, let me ask you one last question.  Suppose that the average hitting 1B is +30 runs.  And suppose you have someone like Minky, a guy who is not a good hitter, but is a great glove.  If you compare him to the avg 1B, he would look good against the average 1B fielder (say +10) but horrible hitting wise (say -30 relative to the average 1B).  So, he’s -20 as a 1B.  But, if you move him to 2B, he might be -5 relative to the average 2B fielding-wise, but be +5 relative to the average 2B hitting-wise.  OVerall, against 2B, he’s 0.

So, what is he?  -20 or 0?  See, your other presumption here is that the guys at the end of the spectrum, guys who have nowhere else to go to get “into balance”, aren’t outliers.  On the fielding spectrum, the 1B is next to the LF, and say there’s a 6-run gap fielding-wise between the two.  But, what if the hitting gap is 15 or 20 runs?  Clearly, the average 1B would have more value than the average LF.  Eventually, you’d have a rebalancing, so that the good hitting 1B find a home in LF and RF, at the cost of defense (Manny, for example).  But, again, no reason to think this is all in balance.

"So, what is he?  -20 or 0?”

He’s -20 when playing 1b, avg at 2b.  Clearly, he should move to 2b (in your scenario).  What’s the problem there? 

And I’m sure such players exist, because each team has to make the best assignment it can within a narrow range of choices.  A team with a great 2B might rationally keep Minky at 1B for a season or two.  Plus, teams make mistakes.  But the question isn’t whether some individual players are at the “wrong” position.  It’s whether there’s a systemic market failure of some kind, that leaves certain kinds of players at the wrong positions, or some players not playing when they should be. You say “again, no reason to think this is all in balance,” but where’s the evidence for that?

* * *

Question for you:  let’s say the average SS would be +15 at 1b, but the average 1B would be -50 at SS.  What’s the “positional adjustment?”

***

Mark Belanger OPS+:  68
Endy Chavez OPS+:  67

You take Endy in LF, I’ll take Belanger at SS, and let’s play.....

So, how much do you pay Minky?  400,000?  4 million$?

***

As for the systematic bias: it seems that your evidence is that it’s always been done like this, that MLB is the final state, so a balance will exist.  Yet, I fail to see how it’s even possible that it will be exactly average every single year.  And, if it’s not exactly average, how far from average is it?  And, what happens in those years where the better hitting players are at SS and not 2B (and, yes, it’s happened).

I’ve provided evidence already, both in players who play multiple-positions, as well as from fans who evaluate players.  In both cases, CF/SS are neck-and-neck.

***

Belanger/Endy.  MLB has evaluated Belanger and Endy as being “perfect” at SS and CF.  This actually is evidence in my favor, since I’m arguing they are both at the top of the fielding pool and the bottom of the hitting pool, and therefore, they would both be equals at SS and CF.  You are arguing that the repl level for a hitter in CF is much higher than at SS.  I’m saying not.

***

http://www.infoplease.com/ipsa/A0760194.html

That’s the NHL’s top 50 all-time.  17 C, 9 RW, 5 LW.

***

As for your question, restate, using a rate as per opp, since a SS will have double the opps of a 1B.

According to Patriot:
http://gosu02.tripod.com/id14.html

The offensive runs per out of the 2B and SS, from 1940 to 1969, has been 90% of league average.  According to the Darwinian theory being espoused by not-me, that means that the average fielding 2B = average fielding SS.  Is this even possible?

Also note that the average hitting OF, in that same 30 year period, was equal to the average hitting 1B.  Seeing that the OF includes CF as well as the corner positions, is this even possible? 

Therefore, I even reject that MLB has a pattern.  There is no “wisdom of crowd” at work at all here.

I see no reason for the positional adjustment to change year-to-year, or thirty-years-to-thirty-years, unless the context has changed (GB/FB tendencies different or grass/turf, or LHH/RHH).

The positional adjustment should remain static, otherwise.

The 1970s certainly looks like a time period where the introduction of the turf caused the leagues to accept worse hitters at SS, in return (theoretically) for better fielders.

According to the Fans’, here are the teams with clearly better fielders at 2B than SS: Orioles (Roberts, Tejada), Tigers (Polanco, Guillen), Royals (Grudz, Berroa), Rox (Carroll, Barmes).  And, I think this year has been pretty low.  (Teams have smartened up.) I would bet if I go back through the years, that I’d find alot more unbalanced situation, where the clearly better fielder was at 2B, probably one-fourth of the time.  Having Orlando Hudson at 2B while you have Russ Adams at SS makes no sense.  Kaz/Reyes.  Having Ichiro at RF, in a post-Cameron world makes no sense.  Are these minor blips that will self-correct soon?  It took a while for Ichiro to be sorted out.

And, I’m still waiting to here on how the DH adjustment makes any sense.

A lot to think about and react to, which I’ll try to do later.  On DH/1B, I’m not sure what’s so mysterious there.  In the AL the avg DL hits a little worse than average 1B, but just a little.  My guess is this is effectively one pool of players, and the younger/healthier players play the field and they tend to be slightly better hitters as a result.  It’s one place where the relationship between average and replacement is probably a bit different, as David noted earlier can be the case.  But clearly we would set replacement level offense for DH at least equal to what we believe repl 1B produce.

The Centerfielders in the 50s included Mays & Mantle, who were the 2 best players in baseball. Also, 1b was a relatively
weak position in the 50s. So, the average CF in the 50s was probably better than the average 1b.

I think one of the problems is the definition of replacement player. First of all, for next year, Atlanta is not going to
use a player at that level. There are probably about 5 types of replacement players : (1) In Season Emergency players,(2)
Start of Season, low budget team (KC); (3) Start of season, mid budget team, (4) Start of Season high budget team, and
(5) the Yankees, and anyone who might be spending with them at this time.

Teams in categories (3) & (4) are not going to base their decision by comparing the player to a marginal major leaguer. The
category (4) teams, those that want to contend, are going to compare their decision against an average starter, because
for the most part, they will have at least an average starter in most every spot. They will be trying to replace those
starters they don’t have who are not at least an average starter.

If a team wants to win, they can’t just go for the best value at each position. if you do, you will spend $2 million on
players worth $3 million, but lose to the team that spends $5 million on the $5 million player.

I think the error in going back and calculating replacement value based on whether a player holds his job is that all
of the studies I’ve seen on this using sabrmetric type calculation. The problem is that the decision maker was not using
these methods. They weren’t thinking “we need a 1b who creates 5 runs,park adjusted.” They were mostly looking at Triple
crown stats, maybe plus slugging, not park adjusted.

How are you comparing the fans assessment of players across the defensive spectrum and coming up with things like “Roberts is better than Tejada?”

Fans are clearly “rating” fielders in the context of their positions, even though they are not supposed to (I assume).  I dont think you can claim there is much validity in comparing fans’ comparison of players at different positions.  Obviously there is some.  Even the best first basemen I am sure are going to be rated a lot worse than the average or below averaqe SS or CF in these “position-neutral” ratings, but you cannot tell me (you can, but I won’t beleive it) that when a fan rates a 2B and SS or a SS and CF, that we can tell which player is better across postions.  There is just no way a fan can know how they compare on an “even keel.” For example, a “great arm” at SS and a “great arm” at CF may not be even close to the same thing even though the fans may be instructed to rate players independent of their positions.  You cannot even tell since the throws are completely different.  Same thing with instinct, first jumps, etc.  They are so qualitatively different from position to position, that there is no way you can then compare those traits on an even keel.  At least I don’t think.

And I agree with Guy (I think it was him) with regard to, “What is the positional adjustment for a SS who is +10 when he moves to 1B and -50 when a first baseman moves to SS (which is probably the case)?  You split the difference.  While you might be right technically, that there is a positional adjustment independent of these offensive values we are talking about, but it may next to impossible getting handle on what that should be, even with tons of UZR data.  And what if there was lots of independence among the positions like in football?  Would there be positional adjustments then, even if there still were a large difference in average offensive performance across positions?  I don’t know.  And what of you are 100% right in that there is a static, logical adjustment accross the defensive spectrum, but that it varied tremendously from player to player, depending on their skill sets?  Would your positional adjustments even be practical then?  And I still don’t know how you get around Guy, David, and my arguments that the only thing that counts is the value of replacement playes at each position.  Are you suggesting that those values will always correspond to your theoretical positional adjustments?

I’m getting to the point where I think we are going around in circles here and arguing about angels dancing on the heads of pins and am ready to give it up.  I don’t think we are all talking about the same thing.

I certainly don’t think you can value A Jones or any other player properly using any kind of positional adjustment unless you know the run value of a replacement players as compared to an average major league player.  And as that number changes, which it does all the time, the value of Jones and every other player MUST change, since the replacement player is the only known quantity (relative to the average ML player) that has a known value (around 500,000), and therefore you CAN’T use a static value to value players at each position.  So you (Tango) can call your static number, the number that never or rarely changes a “positional adjustment” and you can use that to tell us which players at different positions are better or worse than on another, while I will use a number I call a “value adjustment” to tell me which player should be payed more at a certain designated position (whatever position he plays or you want to move him to) and which player I would rather have.

I guess where are two “adjustment factors” become equivalent is when they are far apart, you would say that the player is at the wrong position and must be moved.  That is the only way I reconcile our two positions.  The problem with that is that you have very little evidence that we can determine what happens when a player is moved from one position to another, other than maybe in the OF and from 2b to ss or ss to 2b.

And how does catcher fit into your scheme?  If you say that it doesn’t because it is a unique position, I would have to say that it is a weak argument, as you can say that more or less about any position, and it is just a matter of degree.  Personally I feel that SS and 2B are unique positons from the standpoint of an OF (but not vice versa).

I don’t know if I posted this already (I lost a long post earlier today), but I played baseball for 30 years from LL to semi-pro in my 20’s and then amateur men’s league in my 30’s.  Any good IF could play OF with a little practice.  Any good IF with speed and a good arm could theoretically play excellent OF.  Not so for OF.  No way a career OF’er can move to SS or 2B (or probably even 3B), especially at the ML level.  Does then ever even happen?  Does any OF ever move to 2B or SS.  I can’t think of any.  In order to play ML caliber IF, fielding ground balls, turning DP’s, cutoffs and base coverages, etc., you have to practically have a lifetime of playing IF and fielding tens of thousands of ground balls.

If that is the case, how can you possibly use any kind of a “system” to quantitately compare IF to OF?  I really don’t understand that.

I think I have the differences between the two approaches figured out and it is simply semantics and what you are using positional adjustments for.  In fact, I am sure I have it figured out.

Tango’s way simply compares the “inherent” value of one player to another regardless of the “going rate” for a player at a particular position.  The other way (MGL, DG, Guy, et al.) is an adjustment which includes supply and demand and is used for salary purposes.  It is as simple as that.

An average SS is better than an average 2B, and no one will argue with that.  However, if there happen to be a lot of good SS, then their value will go down and the replacement level at SS will go up.  If the number of good SS goes up enough, an average 2B will be worth more than an average SS simply because of supply and demand.  So the correct way of valuing playes for salary IS to use replacement level offense (again, technically offense and defense combined), whatever that happens to be in any given year.

Still the average SS has more “inherent value” than the average 2B, but because there happen to be so many good SS in that one year, they get paid less, or conversely the 2B gets paid more. 

So while Tango’s method is technically the “correct” one, theoretically, it cannot be used to value players in practice.  In fact, it really has no practical value.  You have to use that (Tango’s adjustments PLUS supply and demand for that particular season.  Replacement level approximates the combination of those two things (the more the supply, the higher the replacement level since there are a fixed number of slots available at the ML level for each position).

Now, one consequence of supply and demand being out of whack, and this is important, is that more players who CAN play 2B should switch to 2B, especially from SS, since they are worth more, and there are not enough of them, at least relative to SS, and fewer players should play or want to play SS, or switch to SS, since there are too many of them and they are getting paid less than they are “worth” according to the “inherent value” formula.

Now, how much of this movement goes on (I’m sure some of it does) to re-establish equilibrium, and how long it takes, is another story.  But in any case, it is irrelevant to the salary value of a player at any given time.  Salaries are strictly determined by a player’s wins above replacement at that position.

In fact, we can forget about positional adjustments completely.  If we want to see who should get paid more, an average CF with +10 lwts or an average SS with +5 lwts, we simply see how many runs/wins each player is relative to a replacement player at that position.  The difference between an replacement SS and a replacement CF is whatever you want to call it.  IF you want to call it a positional adjustment, which I suppose Tango would not like, fine.  If you want to call it a “temporary salary positional adjustment” that is fine by me.  Whatever it is, it will vary all the time, and will NOT represent the inherent value difference among players at different positions, at least until equilibrium is reached, which may or may not occur (probably won’t).

But I just don’t see any practical application of Tango’s “pure positional adjustmnets” other than to say which player is better than which other player in a theoretical league where we don’t know that the balance is and we don’t know the replacement level at each position.  We sure can’t use Tango’s model to figure out how much Andruw is worth, which was the original discussion.  For that we merely have to know how many runs in offense and defense he is greater than a replacement CF.  That goes for all players.  And the easiest way to do that is to use the mean or median (I think mean is fine, although median is probably much better, actually a certain “x-tile” is even better) offensive level at each position and assume that replacement is 15 to 20 runs worse than that.

Fans are clearly “rating” fielders in the context of their positions, even though they are not supposed to (I assume).  I dont think you can claim there ... (rest truncated)

Just another thing for me to prove.  I believe for the most part that they do adhere to this constraint, with the possible notable exception of the catcher.

I will tell you one thing.  There was no bigger shift in positions than Erstad going from CF to 1B back to CF.  If you look at his Fans’ Scouting over those 4 years, you will not be able to figure out which one was for which position.

So, it’s up to me to find more such position switches, and see to the extent this is true.  I’ll add this to my todo list.

***

If that is the case, how can you possibly use any kind of a “system” to quantitately compare IF to OF?  I really don’t understand that.

I’ll just have to use your own UZR data against you!  It happens enough that we can say the impact.  To SS doesn’t happen much, but to 2B does, and certainly to 3B.  And, when you look at 2B - 3B switches, they are very neutral. 

Your argument then is you can accept a career OF going to 3B, you can accept a career 2B going to 3B, but, that you’d find it very hard to accept a career OF going to 2B.

I’ll add this to my todo list.

***

As for your other point, the “replacement” level for a position is not easily determined, unless you also look at the other positions in the pool.  You can’t look at the CF replacement level, unless you also look at the repl level for LF and RF.  My point is that you have yet to show how they are in balance, while I have given 4 examples so far as to how they are not in balance:
- Fan data
- UZR data
- historical data from 1940-1969
- HS, College, Minors, Majors progression does not guarantee Darwinian order

***

I don’t think my position on catcher is weak, any more than I’m excluding pitchers as well, for the same reason I don’t compare QB to LB.  I am simply looking at the pool of players that is a candidate for that position.  For the most part, catchers are their own pool.  You could put Adam Everett in CF tomorrow.  You couldn’t put him at catcher tomorrow.

***

I’m with you on the supply/demand, in general, but…
Still the average SS has more “inherent value” than the average 2B, but because there happen to be so many good SS in that one year, they get paid less, or conversely the 2B gets paid more. 

... that wouldn’t happen.  MLB will in fact pay the average SS more than the average 2B in this case.  MLB teams will lean far more to the inherent value than to the supply/demand value for that given year. That is, if we suddenly had a HS issue, where we had tons of great SS and few good 2B, MLB will respond by giving the SS much more money than the 2B.

The larger point that I’m making is that the replacement level should be based on the “fundamentals”, while you are arguing it should be based on the “technicals” (to use stock-valuation speak).  Because there is alot of rebalancing going on, and I showed how in a 30-yr period, MLB could be completely out of whack with respect to the 2B/SS, as well as the 1B-LF/RF issue, I don’t believe the “technicals” model works. 

Teams pay based on the fundamentals.  Just look at the average pay for a 2B.

Tango, disregard everything I have written so far except my last post. I think that speaks for itself.  I DID NOT say that the major league are in equilibrium.  In fact, I said or implied that they probably weren’t and there is no reason to expect that they are.  They should tend towards it, but the constant flux of players in and out, players changing and sharing positions, as well as inherent biases by teams, suggest that they rarely ARE in balance. 

Why are you saying that, “I have yet to prove they are in balance?”

You are almost ignoring my last post and creating straw man arguments and red herrings.

My final position is explained in my previous post, which I think is correct.  Before that, I was thinking out loud.  There is nothing left in disagreement I don’t think.

There are two definitions of “positional adjustments,” one relating to inherent value and one relating to how much a player is “worth” at any given time as compared to a similar replacement player.

If player A is at ANY poistion and his defense is 5 runs below average and his offense is 20 runs above average, how much is he worth?  If you can find plenty of freely available players who are 25 runs below averwge in offense and defense combined, then this player is 40 runs above that replacement player or 4 wins above him.  Since replacement players get paid around .5 mil, by definition, then this player, REGARDLESS OF POSITION (we don’t need a “positional adjustment") should get paid in a perfect world 4 times whatever the going rate for a marginal win is plus 500,000.  I don’t see it any other way.

And BTW, since NO players go from the OF to SS, you might as well throw SS in the same category as C.  And since hardly anyone goes from OF to 2B, throw 2B in there too.  You don’t have the “transitive effect” in switching positions.  Any player can move around the OF and we should be able to predict around how many runs he gains or loses.  Players who go from OF to 3B (are there more than a handful?) have some unique skills that enable them to play 3B effectively or at least adequately.  My guess is that most OF do not have that skill and if we forced them to play 3B they would lose 20 runs!

Please read this:

So if 1% of all OF go to 3B and say lose 3 runs and 99% don’t go there presumably because they would lose a lot more than 3 runs (which I think is reasonable, how in the world can you say that the adjustment from OF to 3B is 3 runs (or 5 runs)?  You can’t!  Those adjustments have to apply to every player in the OF, not just the ones who are chosen to move.  This is selective sampling at its worst (or finest)!  Selective sampling times 10!

I said I was with you on the understanding of the supply/demand thing, so our disagreement is whether to value them based on fundamentals (inherent) or technicals (supply/demand).

***

I understand the selective sampling issue, and it’s up to me to show that this is not the case.  It’s not just the “1%” of best OF candidates that are asked to play 3B, which is why they don’t look so bad at 3B.

This is why the Fans’ data is so important.  I can construct a profile of exactly the kind of RF are asked to go to 3B or the kind of LF that are asked to go to 2B, etc.  With this data, I’ll be able to ascertain the kind of guys that do get moved from RF to 3B.  Ichiro, for example, would make a sensational 3B.  Or SS for that matter.  But, in general, you don’t move stars around.

And, jsut because you don’t have many guys going to SS doesn’t mean they are unique!  It just means that only an excellent fielder should go there.  If you have an excellent OF, you put him in CF!  Why move him to SS?  I already said the SS/CF are on par.

2B/ss/3B move around alot to each other.  That’s one pool.

LF/CF/RF is obvious, and that’s another pool.

LF/2B, and RF/3B have alot of movement.  (CF and SS aren’t really involved, for obvious reasons.) However, that doesn’t remove them from the discussion.

You have 3 pools, the IF, OF, and C.  In terms of “distance” IF and OF are probably one-tenth the distance that either is to catcher.

1B is a “fill-in” position for the most part, a discard.

***

As for your “average”, like I said, if the average SS is hitting better than the average 2B, your system will ensure that both will be paid exactly the same, because you are forcing the average off+def at zero for each position.  In some years in 1940-1959, when the LF and RF were much better hitters than the 1B, your system would have paid all three guys the same, even though the 1B is a dumping ground for the corner OF.

And, like I said, show me where MLB actually pays the way you are saying.  I have no doubt thatthe average 1B gets paid far more than the average 2B.

And BTW, since NO players go from the OF to SS, you might as well throw SS in the same category as C

Off the top of my head, active guys who have been used at both SS/CF on at least a semi-regular fashion in the majors;

Willie Bloomquist, Tony Womack, and Chone Figgins.  I’m sure there are a few others.  It happens more than never, though it is admittedly not common.

At the minor league level, the SS to CF move happens all the time.  Justin Upton and Adam Jones are two of the big name examples.

As well, would you make the avg 1B paid the same as the average DH?  The avg 1B is a much better hitter than the average DH.  And, MLB does not pay the average DH the same as the average 1B.

MLB teams follow the fundamentals (inherent) to a much larger degree than the technicals (supply/demand).

I agree with the large majority of MGL has said.  But I don’t think the concept of “inherent value” is meaningful.  You say that “An average SS is better than an average 2B, and no one will argue with that,” and then concede that the “average SS has more ‘inherent value’ than the average 2B.” The first statement is true in the sense that an avg. SS would perform better at 2B than the avg 2B would perform at SS.  But, I can get average defense at SS and avg defense at 2B from replacement players, so the value of both is zero.  Or rather it’s zero UNLESS I have to pay a price for that defense in the form of lost offensive production.  And I do: let’s say it’s 5 runs for a repl SS vs a repl 2B.  So I now know exactly how much more valuable the SS’s defense is: 5 runs—it can’t be more valuable than what it costs.  There is no ‘inherent value’ separate from value over replacement.

Now, if we learn that the avg SS would be +15 on defense at 2B, then we have a problem. It means a lot of repl SSs should be starters at 2B somewhere.  But, I still see no reason to think such large, systemic disparities currently exist.

To go back to where this started, Tango said:  “My research clearly shows that the fielding contributions of the average CF is on par with those at SS,” and that “To put CF *below* 2B/3B is in my opinion, and to use your [MGL’s] words, preposterous!” But it’s not preposterous at all.  In fact, we know that teams are willing to pay a higher price—in reduced offense—to get avg defense at 2B than in CF, and a much higher price to get avg defense at SS.  That tells us that those IF skills are more scarce, and that makes them more valuable.  (It does appear that avg def CFs are more scarce than avg 3B these days.)

If an average CF really is “better” than the avg 2B, then Tango shouldn’t want Endy to play LF. Move him to 2B!  He’ll be +25 to +35 on defense, and well above replacement on offense, and be worth $10-15 mil.  I can’t prove this wouldn’t work, but the fact that this shift virtually never happens (as MGL pointed out) is pretty strong evidence that these skills are not so fungible.

*****

Patriot’s chart is very interesting and helpful.  Maybe this is a you-say-toMAYto-I-say-toMAHto-thing, but to me the chart is perfect confirmation of my claim that the defensive “market” has been remarkably stable for the past three decades.  Except the drop in offense at SS in the 70s and, maybe, increased offense at DH in the 90s, all other changes are in the rounding error category. 

My point is not that we have reached perfect equilibrium, only that the game does evolve in the direction of greater efficiency over time.  So teams apparently valued defense at 1B as highly as OF defense from the 40s to the 60s.  Were they wrong to do so?  Perhaps.  But then again, we’re talking about different athletes:  those OFs were overwhelmingly white, while the OFs in the next three decades were mostly black.  It’s certainly possible that the SD of defensive talent in the OF was much smaller then, and putting a decent-fielding but not especially fast player at 1B was as useful as putting him in the OF.  But even if teams were undervaluing OF defense then, infficiences thirty to fifty years in the past is completely consistent with a “Darwinian” analysis.

From 2000-2005: Melvin Mora is the biggest one. Donnie Sadler, Jolbert Cabrera, Denny Hocking.  Mora was also in LF and 3B.  His UZR, per 162 GP:

+15, CF
+5, 3B
+2, LF
+2, SS

(All with 100 to 300 games.)

Just looking at Mora, and it’s not clear.  Otherwise, you’d guess the average LF = average SS.

Anyway, I’ll present the data in due time.

Endy is LH.  Otherwise, I would put him at 2B.  And the handedness issue may be the largest sticking point.

***

At the very least then, you must use long-term positional adjustments (or replacement levels), and not the year-to-year.  Otherwise, in some years, even recent years, the average hitting SS is better than the 2B.

"Endy is LH”

OK, but why isn’t Taguchi playing 2B?  Why didn’t Pettis move to 2B?  Or pick your good-fielding, light-hitting (for an OF), RH OF.  Your analysis suggests that a plus defender/weak hitter in CF would basically be a Gold Glove, avg hitting 2B.  That move should happen all the time if your analysis is correct.

I don’t understand why Pettis would need to move to 2B.  The average CF is about +7 runs relative to the Chipper Lane line, fielding-wise.  The average 2B is about +2 or +3 runs relative to that line.

The average CF is a league-average hitter, while the 2B is probably about -5 runs relative to league average.  So, the average CF is +7 off+def, and the average 2B is -2 off+def.

Therefore, there should be more CF that move to 2B.  But, it shouldn’t be Pettis, who would be +20 or +30 runs in CF.  It should be the CF who is about a -10 runs as off+def (remember the avg CF is +7), a guy who will be riding the bench as a CF, but would have a decent chance at a 2B position for a weak team.

Or, move more guys from CF to LF, and move LF to 2B.

Yes, a league average SS should get paid exactly the same as a league average 1B and any other position in a perfect world, where the supply of players at each position is the same.  But that is not the case because of supply and demand and maybe some biases among the people who set the salaries.  And is there such a discrepancy among league average players in salary?  I don’t know that there is.  If there is, I would think that there would be because as I always say, 1B and corner OF sluggers are overvalued by teams in general and get paid more than they are “worth.” We are trying to determine what a player is worth in a fair world or on a team that we control, are we not?  I don’t care what other teams or the league as a whole chooses to pay players.  Obviously it affects what I end up paying my players, but…

If 1B are way overvalued, then some other position or positons will be undervalued, so I will simply have a replacement 1B and a great player at the position or positions that are undervalued.  I still need to know the dollar value of every player at any given time and that, for the umpteenth time, is determined (as far as I can tell) by the number of runs below average, offense and defense combined, that a replacement player is.  Sure, it is not easy to determine that, but it can be done and must be done.  I DON’T need to know what your inherent “positional adjustments” are if I run a team.  I really don’t.  It doesn’t matter a hoot!  Of course I want to know what would happen if I am thinking about moving players around in their position, but heck, I don’t think you can answer that anyway!  As I said, you may have the perfect answer “on the average,” but there are some things that are so variable, that the perfect answer on average won’t help me with any one individual player.

And to use our favorite word, to think that Ichiro would be an excellent third baseman because he shares the same chracteristics in the Fan report as a third baseman, is preposterous!

As I also said, if a third baseman has his arm rated as a “1” and so does a RF’er, that does NOT mean that those “1’s” are equivalent, no matter how much you instruct the fans to assume as such.  And until you at least have a category for the OF’s like “hands, if they moved to the IF” there is simply no WAY that you can say that Ichiro even probably would be an excellent third baseman. Probably most of the players who play OF and IF (like Womack, Figgins, Luna) have played both positions many times in their careers.  There are probably many players, and Ichiro may be one of them, who have never played infield (maybe in youth ball).  If that be the case, I think they would have no shot at playing IF at the major league level.

And BTW, whomever listed the names of plays who play both positions (SS and OF), I specifically said “go from the OF to SS” (not the other away round)!  Most, if not all of those players you mentioned were already IF’ers or were probably both for a long time.  I was talking about career OF’ers moving to the IF, if I did not make that clear, which I though it was.  Is there ANYONE who went that route in the history of baseball?  Going from career OF to SS?  I undestnd that you have to be a good OF’er, but if CF and SS are equivalent, which Tango claims they are, SURELY there have to be plent of players going from CF to SS, almost as many or just as many as go from SS to CF (Yount, et al.).

I think I understand where both of you are coming from.  Ideally, I side with Tango, but practically, I’m probably on MGL’s side for now, because I don’t know how to usefully quantify the idea without using position specific baselines for comparison. 

So, let me offer up an analogy that I think sums up what you guys are saying. 

Tango and MGL own a Restaurant and need to hire a kitchen staff.  They need a head chef, a sous chef, a preparer, and a runner.  They put up ads for all four jobs and get 100 responses. 

Tango wants to interview all 100 guys and then hire the best four, figuring out where to stick them once you’ve determined that you’ve selected the four best chefs.  He’s not interviewing for specific jobs.  He only cares about who has the most ability in the kitchen overall. 

MGL wants to hire the best head chef, the best sous chef, the best preparer, and the best runner.  He’ll interview, say, 25 each, and pick the best one of the four groups.  He doesn’t care that the tenth best head chef candidate is still a better chef than the best sous chef candidate, because the sous chef isn’t going to be doing the majority of the cooking. 

He simply wants the guy who is the best at performing those specific tasks, and if he ends up hiring the candidates who would rank #1, #13, #32, and #49 on the overall ranking of the 100 candidates, he doesn’t care, as long as he gets the guy who will give him the best performance in that specific role for the wage he’s willing to pay. 

If we’re trying to figure out who the best cooks are for some kind of iron chef competition, we should interview Tango’s way.  If we’re hiring for a restaurant, though, we should probably do it MGL’s way, or else we’ll end up paying a head chef a lot of money to stick potatos on a plate.

I don’t have anything substantial to add on the main issue of contention here (but a very good discussion with great ideas from everyone), but I would like to echo Tango’s statement that if you do decide to go the offensive positional adjustment route, you should use long-term.  It’s just like park factors.

There are some other problems too, like the fact that IIRC, in most recent seasons, you will find that RF are better hitters then LF.  But conventional wisdom is that LF are just RF who can’t throw, and that RF is a tougher position.  It’s similar to the 1B/DH thing that’s been alluded to--DHs are outhit by 1B, but it would be proposterous to adjust 1B down below DH.  What I do on my site(admittedly not the best solution) is to lump all corner outfielders in one category and 1B/DH in another category.

As an aside, a couple months ago I fiddled around with stats for the early days of the NL (1876-1883 to be precise).  While it was a different game in so many ways, and there is a much bigger margin for error in the stats, I still think it could be of interest.  I would assume that this early game, without a farm system, with the game still very much evolving, without fielding gloves, would have a level of competetive balance much lower then the modern big leagues.  Conditions closer to college ball or something today (again, leaving aside the huge differences in the game itself).  Anyway, here are the r/o for each position v. the league average:
C=93,1B=121,2B=96,3B=99,SS=94,LF=115,CF=112,RF=97

I don’t want to jump to unwarranted conclusions, but you’ll notice that the spread is much less than it is today, although the spectrum is still fairly intact, with the glaring exception of RF.  And this is in a game where many more balls are put in play then today.  So even though the offensive numbers tell me there is a 2 1/2 run difference/400 outs between RF and a SS, I’m not buying that these positions are “balanced” in off+def (unless the shortstop position, which was still evolving at this time, was a completely different animal then it is today--and while I am not a historian, I do believe it was already considered the most important defensive position).  Anyway, I digress.

Mackowiak started his minor league career as an OF, and then bounced between the two.  His UZR much better in the IF than OF.

Melvin Mora principally an OF in the minors. Then becomes regular SS in the majors!  UZR in IF and OF comparable.

Jose Macias started career as OF, then moved to 2B in minors.  UZR slightly better in OF than IF.

Joe McEwing OF in the minors, mostly IF in majors. UZR better in OF than IF.

Overall, with these 4, pretty much a wash.  All of them pretty much average OF fielders to begin with.  And, they were average IF.  They weren’t some “elite” guys who had something special.  If Mora can make the transition to SS of all positions, how hard would it be for a LF to make it at 2B or RF to 3B?

Honestly, it doesn’t matter if it is .01% of all OF moving to the IF or 1% (I imagine it is somwehere in between).  My point still stands.  That it is useless to try and generalize from such a small and selective sample.  But that is not the point anyway.

Another reason the idea of moving players around defensively to find the best combination (like Patriot’s restaurant analogy) is not practicable is that Tango specifically found that there is generally and probably a learning curve when you change positions.  It is not really worth doing unless you have to (because of redundancy - in which case, you can usually find an equivalent player at another position to trade for), a player is completely miscast for some reason, or the usual moving of a player to the left with age and injury.

Sorry, that is David with the restaurant analogy.  A good analogy, BTW.  Makes me hungry.

Great discussion.  A little heated at times, but never uncivil. 

I think we’re all agreed on using long-term differences to assess value, not short term imbalances in supply/demand that may develop.  But that still leaves the question of how you measure it:  A) offensive differences, or B) ‘inherent’ measures of fielding difficulty. To pursue David’s excellent anology, Tango isn’t just looking for the best 4 fielders, he thinks he can calculate that a “6” preparer = an “8” sous chef.  The difficulties of measuring this are large, as Tango acknowledges, and like MGL I’m inclined to think they’re insurmountable.  But perhaps not. 

But even if B gave us a different answer than A, would it really change our assessment of value?  I don’t think so.  Suppose we learn that Endy and many other 4th OFs “should” be playing LF.  So we would set our offensive replacement level for LF lower, and Manny has become a bit more valuable.  But wait, his UZR relative to average has now declined with the influx of great fielders to LF.  Surely this would be close to a wash. 

* *

On Pettis, I was thinking he should move at the end of his career when he could no longer hit enough to survive in CF.  Shouldn’t that be a normal career path, given your analysis?

* *

I’m not sure I see why the average salary at every position should be the same.  Certainly at the end of the spectrum, immensely talented hitters will “stack up” at 1B and make the total talent there above average.  The same thing could theoretically happen with overwhelming defensive talent at SS, but I doubt that ever happens.  Since the variance in defensive talent is smaller, it seems possible, even likely, that total value will be higher at the least-skilled defensive positions.

**

Patriot’s data is interesting.  Perhaps there was effectively a single pool of OF/1B (LHs and guys who couldn’t handle the IF), and then RF was where the worst athlete in that poll went (as still happens in Little League today).  Within such a pool, a positive correlation btwn off and def skill seems plausible.  I think the same thing happens today at 1B/DH and LF/RF, although defense in right is now valued more highly.

MGL said this:

I don’t know if I posted this already (I lost a long post earlier today), but I played baseball for 30 years from LL to semi-pro in my 20’s and then amateur men’s league in my 30’s.  ... In order to play ML caliber IF, fielding ground balls, turning DP’s, cutoffs and base coverages, etc., you have to practically have a lifetime of playing IF and fielding tens of thousands of ground balls.

First off, don’t you hate it when a pro ballplayer tells us statheads the “I played this game, you didn’t” line!

***

I don’t agree at all with the “lifetime of playing IF”.  There’s 4 cases right there.  I’m sure I’d be able to find several every year, especially among the scrubs. It is not “selective sampling”.  There’s nothing special about any of those guys, that would preclude us from considering how Ichiro would do at 3B. 

Let’s consider something else.  1B is an IF position, and they get about two-thirds the plays of a 3B.  So, quite substantial.  The largest difference between 1B and 3B is the throw.  However, they are both fielding groundballs, from about the same distance and angle. 

So, what if I look at “pure 3B” who then are moved to 1B, and “pure OF” who are then moved to 1B?  Surely the 3B will have a much easier time of it.  If you needed a “lifetime of GB”, then any OF to IF move (including to 1B), should be almost as foreign as moving to catcher.  Right?

Larry Walker for example was a pure OF, and he played emergency 1B for the Expos in 1991 when Galaragga went down (and again in 1994 when Floyd went down).  In 1991, he had 30 A and 313 PO, and Galaragga was 80/887.  (Gala was a brilliant 1B.) Their E and DP rates were also similar.  In 1994, Walker was arguably better than Floyd (though Flyod was not a career 1B).

Ichiro is a career OF, and Frank Thomas is a career IF.  I am quite certain that Ichiro, at his worst, would be a far better 1B than Thomas at his best.  Was there any doubt with Darin Erstad?  Ichiro would also be an above average 3B, in under a month. 

***

Finally, I would still like to hear how you would evaluate the production of a HS or College 2B and SS.  Are you going to hold to the “avg 2B = avg SS” line?

How about in A?  AA?  At what point then?

***

Let me just say that this has been very fun!

The reason you stick your worst fielder in RF in LL is that most hitters are RHH, and they pull.  I would bet that in the early days of MLB, this was the case, that the emergence of LHH and LHP (and switch hitters) is a Darwinian evolution.

This is from studes:

http://www.hardballtimes.com/main/article/intro_to_salaries_and_win_shares/

Here’s a table of salary for each WSAR paid to free agents by position this past year:
Position Salary per WSAR
P $1,050,662
SS $1,019,427
3B $999,730
OF $801,902
1B $747,436
C $698,912
DH $639,304
2B $565,166

Doesn’t this chart go completely against what is being discussed here?  It’s exactly like I said, that players are paid on the fundamentals, not technicals.  And look, SS are paid more than 1B, on a marginal win basis! 

The average of 2B+SS+3B is a bit higher than the OF average, though if I included 1B, they’d be pretty much the same.

Among the 2B,SS,3B, like I said, the fielding levels of 3B and 2B are similar.  We know that 3B are better hitters than 2B.  Therefore, under my “fundamentals”, the avg 3B is better than the avg 2B, and, that’s what we are seeing here.  The “real” replacement level line for a 2B is much closer to the avg 2B than the replacement level line for a 3B is to the avg 3B.

Fundamentals says so, and MLB are paying as such.

I think Studes’ data makes exactly the opposite point.  Here are his WSAR by position (by the way, this is all 2003 data):

POS WSAR
--- ----
1B 296
2B 234
C 226
SS 206
3B 195

If you are right, the average 3B should produce more real value than the 2B.  But they don’t. This shows what I was suggesting:  that offensive talent accumulates at 1B at such a rate that these players have more total value than average.

If teams are paying more for each marginal win at ANY position, that suggests a major market inefficiency to be sure.  But I don’t see how it supports your point.  And I don’t personally have enough faith in fielding WS as a system to rely on these results in any case.  (I think WS assumes the avg 2B has more defensive value than a 3B—seems odd to use that data to prove this isn’t the case.) Now, if 3B and 2B are paid the same for each offensive WS, over a common replacement level, that would suggest to me that MLB thinks their defense is equally valuable.  And that would be interesting (though you would be the last to say that proves their defense is in fact equal, right Tango?).

I’ll backpedal a little about “needing a lifetime of GB to play the IF"…

Hmmm, you are right that it would be odd for me to use the Win Shares data, which has a built-in fudge to establish what I’m trying to say.  Even then replacement level is questionable here.

In any case, I can multiply the WSAR x salary per WSAR to give me salary by position, from Studes table:
1B $221,241,056
SS $210,001,962
3B $194,947,350
C $157,954,112
2B $132,248,844

So, 2B are paid far less than the other positions.
MLB pays 3B and SS about the same, so, they have no problem balancing offense and fielding here.  So, why are 2B so low?

***

I’m beginning to think that the “pool” of players goes:
Catcher pool
IF + Catcher pool
IF + OF + catcher pool
IF + OF + 1B + catcher pool

So Beltran and Ichiro belong to the third and fourth pool, even though they would certainly qualify as the second pool.  Everett belongs to the 2nd, 3rd, 4th pool.  Mauer belongs to all of them.

In order to get a replacement level, this would be the starting point.

***

As for Pettis, his offense was about -20 runs relative to a league average hitter.  Unless he stopped playing because of injuries, he should have continued to be a CF.  Otherwise, yes, he would have been a fine 2B.

2B does seem to be an interesting situation. 

Theory A:  teams are overestimating the difficulty of playing there and accepting less offense than they should.  Lots of better-hitting 3B (and CF) could play there and be an upgrade in terms of total O/D.  This isn’t happening perhaps because teams have an image of what a 2B should look like (small, quick), and position is ripe for an offensive upgrade.

Theory B:  playing 2B requires a particular combination of skills that limit the talent pool more than 3B (or even CF).  More range than 3B, and while throw is shorter on routine plays, ability to turn DP requires both agility and good arm.  Plus handles more FBs to shallow OF, and has to cover for limited mobility of 1B. 

Both seem plausible to me.  I’m skeptical of the CG-to-2B scenario, but if there’s lots of evidence that 3B can move to 2B with limited falloff in defense (not just a couple of cherry-picked examples), I’d take that as serious evidence for theory A.

* * *

Leaving Pettis aside, I think your theory does suggest that movement from CF to 2B should be common (as would 4th OFs converting to 2B early in their career).  It would seem that players might lose some footspeed before their “hands” skills, making this an especially common transition. Yet the Bernie Williamses of the world move to CR OF, they move to 1B, and they move to DH, but basically NEVER move to 2b.  I think that is a problem for your theory.

* * *

Back to your DH vs. 1B concern:  didn’t someone determine that playing DH significantly reduces offensive production, vs. playing the field?  Oh, yeah, it was you!  So some of what you’re seeing is just that effect.  If older players are also better able to handle hitting off the bench, as seems possible, would contribute further to the relatively better offensive production of the 1B.

Re: 3b to 2b moves

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

At the above link, I said:
There are 64 2B/3B pairs.  This one is the most interesting because of the different skillsets in play at those positions.  Of the guys that played at both positions, they were +0.0 as 2B and +2.5 as 3B.

So, the dropoff is pretty low.  How about offensively?  From 1950 to 1989, Patriot is showing a 12-run gap in hitting. More recently, there’s only a 6-run gap.  There has been a steady increase in hitting at 2B, and a steady decrease in hitting at 3B.  This shift may in fact be happening.  The average 3B (off+def) is likely still better than a 2B.

RE: 1B and DH, teams are reluctant to use a pure DH.  One reason is that it is more efficient to simply use a bench player who can play other positions when not DH’ing.  You see lots of teams simpy throw out whomever is not in the starting lineup or a starter at another position whom they want to rest.  Another reason, is that teams will use washed up older players who are overvalued.  Also, there could be a whole pool of very good hit, no field players in the minors, but there are not.  Teams do not like to “groom” young DH’s in the minors, so the pool is drastically reduced.  Finally, good hit no field players simply do not make it up through the ranks. The generally gete weeded out long before they reach the minors or the the majors.  If someone is a pretty good hitter but terrible fielder in high school, you won’t hear a scout say, “Yeah, he can’t field, but someday he may make a good DH; l