I think my issue with Yearrgh's analysis is that he makes two arguments against Swisher's projection (that I have seen):

1) Swisher had a 92 OPS+ last year, so he might be toast.

2) Kenny Williams traded him, so he must know something we don't.

On the first point, ZiPS also "knows" what Swisher's performance was last year, and probably "knows" a lot more about what typically happens to players like that.

On the second point, well, if people said that about Billy Beane, they'd be immediately jumped on as some kind of Beane fanboy or apologist.

I'm very supportive of the idea that projection systems have limitations. But it's not like we don't know what those limitations are. If you tell me that you know that Swisher has been partying much harder, or started taking steroids or got LASIK or something, or changed his workout routine, that is definitely stuff not captured in ZiPS (Presumably, anyway... Dan is a pretty smart guy though.) doesn't catch.

Unstick?

Looking at his peripherals Swisher's problem last year was entirely caused by a lower than expected BABIP. He even had the highest line drive % of his career and all his other peripherals were exactly what you would expect. Seems like he was just extremely unlucky.
Check out the first chart here;
http://www.replacementlevel.com/index.php/RLYW/comments/breaking_down_the_nick_swisher_trade

If you tell me that you know that Swisher has been partying much harder, or started taking steroids or got LASIK or something, or changed his workout routine, that is definitely stuff not captured in ZiPS (Presumably, anyway...

Smart? If I tried implementing that kind of stuff, I'd be smarter to open my veins in a warm bath!

The Senators hit fewer HRs at DC than they did on the road every year but two that they played there; the typical home team hits more at home, and in a good HR park will hit significantly more. DC Stadium wasn't Griffith Stadium, but it was not HR-friendly.

Yes, but the Senators' pitchers allowed more HRs at DC than they did on the road during the years Lock played there (1962-66).

Senators hitters HR, H-A, 1962-66: 323-334

Senators pitchers HRA, H-A, 1962-66: 426-387

Overall H-A, 1962-66: 759-721

It's incorrect to describe DC Stadium in that period as a difficult HR park. It slightly favored HRs; was overall rather neutral.

Looking at his peripherals Swisher's problem last year was entirely caused by a lower than expected BABIP. He even had the highest line drive % of his career and all his other peripherals were exactly what you would expect. Seems like he was just extremely unlucky.

Sounds like an '09 fantasy draft bargain waiting to happen...

Smart? If I tried implementing that kind of stuff, I'd be smarter to open my veins in a warm bath!

Haha, well I meant that you would be quite smart if you could find an effective way of implementing it.

Gaelan's absolutely right in #94. The GM needs to make finer-grained distinctions between players than the performance models we have will yield.

The Yankees are obviously gambling on Swisher coming back to something like his pre-2008 performance level, and if that happens they'll look good. But there is a significant chance (about 1 in 4 for OPS+ below 100, slightly worse chance of SLG below .400) that Swisher will make the deal look bad, and I don't think that should be discounted. It's a good trade in as much as the Yankees didn't pay for Swisher's pre-2008, but it's not without risk.

-- MWE

CP, I know you follow the Yanks closely. Do you really think that a Damon, Melky/Gardner, Swisher OF (L to R) is better than a Nady, Damon, Swisher OF?

I don't know, but I am under the opinion that Damon can't physically withstand the punishment of playing in CF. I honestly believe that if you put him in center, you're going to get another 2007 out of him and I don't think it's a coincidence that his bat bounced back when he played a less demanding position. I also think that Damon's defense clearly wears down as the season goes on, so while he might be a good CFer on opening day, by August he'd be awful With that in mind, I'm still not sure which is better, the only guys who can play center can't hit, and then you have 3 semi-tweeners who can don't have the bats to be really good in the corners and don't have the range to play center. Add in the up in the air defensive numbers with some of these guys, and I don't know which one is best, and because of that, I think if the Yanks smartly platooned them (without looking at the splits, I'm not sure how they might go about it, but some have suggested certain things already), they could probably eek some extra runs out of a somewhat limited group.

I'd guess that defensively Nady is average in LF, Damon maybe -5 runs in CF, and Swisher +5 in RF. That seems worth it given the offensive upgrade.

I think Damon, arm included, will be much worse than that over the course of a year.

I think Damon, arm included, will be much worse than that over the course of a year.

Even if you limited him to 135 G in CF, with another 15 at DH?

I think Damon would be fine defensively and could hold up if the Yankees put Gardner or Melky in as defensive replacements somewhat often and DH him another 20-25 games.

I'd also prefer Swisher to be the LF at home rather than Nady since there's more ground to cover.

Gaelan's absolutely right in #94. The GM needs to make finer-grained distinctions between players than the performance models we have will yield.

It would be cool if we had a poll among GM's before the 2008 season on whether Jones or Bay was a better bet to bounce back. Bay was already under contract, and we only know for a fact that 1 GM of 30 thought Jones was worth 18 million.

Count me as skeptical though, that as a group they would have picked the right player.

I think Damon, arm included, will be much worse than that over the course of a year.

Even if you limited him to 135 G in CF, with another 15 at DH?

I essentially agree with CP's assessment in 109. I think the Yankees could probably safely wring about 50 games in CF out of Damon, assuming they take care resting and rotating him into LF, but I don't think his body would hold up for 135 games in CF. If they push him or he pushes himself into playing that many games, I'd expect him to be more like -10ish runs, including arm.

I wouldn't be surprised to see a rotating 2009 CF of Damon, Swisher, and whichever of Melky/Gardner sucks less. This is assuming the Yankees sign Teixeira... if Swisher is the full time 1B, well, God help us all.

Even if you limited him to 135 G in CF, with another 15 at DH?

I think so. Even his defense in left was noticeably worse by the end of the year. Yes, that was due to the fact that he slammed himself into a wall for no particular reason, but it seems that he's going to do that every year. I could very well be wrong, that's just been my read on Damon, maybe he can push through it because it's a contract year.

I agree with Gaelen, pretty much.

A projection is only a singular data point as is any other information you might find reported and the fact that a competent GM all but dumped him.

You have to take all the information and then make a value judgement.

I'm very supportive of the idea that projection systems have limitations. But it's not like we don't know what those limitations are. If you tell me that you know that Swisher has been partying much harder, or started taking steroids or got LASIK or something, or changed his workout routine, that is definitely stuff not captured in ZiPS (Presumably, anyway... Dan is a pretty smart guy though.) doesn't catch.

Well none of us *know* things like this, but Kenny Williams *might* know things like this, that us outsiders are simply not privy to. Again, given that Kenny Williams is a competent GM, especially at making trades, shouldn't you at least give pause when he dumps a like Swisher for seemingly little in return that there could be something else going on?

Agreeing with Gaelan, BFFB, etc. here, I think, but no one is saying "Kenny Williams traded Swisher for what looks like 40 cents on the dollar; therefore, there *must* be something going on with him that we don't know about", simply that it makes it worthwhile to consider that possibility.

if people said that about Billy Beane, they'd be immediately jumped on as some kind of Beane fanboy or apologist.


Fair point. That crap sure does get old.

Well none of us *know* things like this, but Kenny Williams *might* know things like this, that us outsiders are simply not privy to. Again, given that Kenny Williams is a competent GM, especially at making trades, shouldn't you at least give pause when he dumps a like Swisher for seemingly little in return that there could be something else going on?

I could put more stock in this POV if anyone could say what, in fact, this something would be. If Swisher is truly damaged goods (injured, has a drug problem, etc.) then Kenny is taking quite a risk to his own reputation if he's trading Nick away. Alternatively, if it's just a freakish drop in talent level then I haven't heard much from people who agree with this summation, not even a theory.

Yikes, what a silly thread. :-)

I'm sympathizing with Yeargh. Not sure I agree with him yet but he never said Williams had more information or ZiPS was wrong or many of the other things he's been accused of saying.

What he said was that Swisher's poor performance last year and the uncertainty about his future that it introduced hurt his trade value. And all of the empirical evidence we have (i.e. this trade) agrees with that assessment.

We all carry some sort of model around in our head for evaluating trades. This one is an outlier. A good researcher when they find an outlier doesn't say "wow, this is a steal for the Yanks"; rather he asks "is there something I don't know?" Now after giving it some thought, he might still say "wow, this is a steal for the Yanks" but you should take more into consideration than ZiPS lines and money.

Now we'll likely never know what we don't know about Swisher, Betemit, Williams and Cashman. But there are a few things we know for sure. Billy Beane, a GM we generally think of as being somewhere between good and excellent, traded a cost-controlled, power-hitting young OF for two pitching prospects -- an odd but not necessarily bad trade on the surface (for the record, I didn't like the trade). (We can add the additional information that he interestingly just traded for a better version of Swisher in Holliday.) Swisher followed that trade with a bad year. Kenny Williams, another good GM who we generally don't think of that highly, just traded the still young and cost-controlled Swisher for what appears to be even less than what he gave up even though there's no reason to think he's under substantial payroll constraints.

So a good GM trades the young, cost-controlled Swisher for less than what he seems worth, Swisher has a bad season, another good GM trades the young, cost-controlled Swisher for less than what he seems worth.

Hmmm....

Of course GMs screw up all the time -- I've posted such plenty of times. And they're human and act on emotion. And they have tons of uncertainty in their valuations of players too. I'm not saying that's conclusive evidence that what we're missing here outweighs the information that goes into our projections, but it does give us some reason to think that the projection may be a little rosy.

Also ... it is safe for us to assume that Williams shopped Swisher to more teams than the Yankees (or the Yankees called him just as likely). It is safe to assume that this set of players isn't what Williams asked for initially but is the result of a process of negotiation. From this, we can conclude (with rather limited certainty) that both Williams and Cashman think this is "fair" for Swisher -- i.e. Cashman wasn't willing to give up much more than this for Swisher. The big variable here being that it's quite likely Williams and Cashman value the players in this deal differently to begin with.

Still, I have no problem calling this a win for the Yankees as they gave up nothing of substantial value _to them_. Even if Betemit morphs into a league-average 3B, he was going to be nothing but a backup on the Yanks (in 2009 or beyond) so even if Swisher tanks, the Yanks would lose very little (some money and 300 PA of probably superior backup IF play). Meanwhile, they fill one of their needs -- whether Swisher ends up in CF, RF, 1B, or super-sub. This also makes them younger. And in the Delgado option thread, I noted that the Mets seemed to be putting off their 1B problem for a year (i.e. they were not going for the long-term solution of Teixeira) and that I thought that could be a mistake and that it was reminiscent of the Yanks sticking with Bernie rather than signing Beltran. This is at least a longish-term solution to one of the Yanks' holes -- or at least it has the potential to be unless Swisher tanks.

The White Sox could win the trade as well but I think that requires Betemit turning into a league-average 3B and/or the pitchers really paying off. I think the Sox did need to move one of Swisher/Dye/Konerko/Thome but, of those, Swisher would have been the one I'd be most reluctant to trade. Now, the trade market for Konerko (older, less athletic, more expensive, also coming off a bad year) has to be worse than the one for Swisher so Williams probably couldn't do anything there. And Dye and Thome will likely be better hitters than Swisher for 2009 so I can get that too. Still I'd rather have dumped Konerko for a bag of balls and picked up a chunk of his salary than make this trade.

Still, the Sox gave up very little of value _to them_, at least for 2009. But it's not clear they've filled one of their holes and they probably have hurt the 2010-2012 Sox.

He started to smell a bit like Jeremy Giambi last year, and that is a potent aroma

It would be cool if we had a poll among GM's before the 2008 season on whether Jones or Bay was a better bet to bounce back. Bay was already under contract, and we only know for a fact that 1 GM of 30 thought Jones was worth 18 million.

FWIW CBS and perhaps Yahoo probably have data WRT fantasy drafts and auctions...

He started to smell a bit like Jeremy Giambi last year, and that is a potent aroma

ouch. Yanks must hope he starts to smell like Carlos Peña did when he was getting shuttled through all those teams.

But what I really want to draw people's attention to is how much variation there is around these projections. This isn't a criticism of ZiPS and, in fact, Dan is to be praised for making these distributions available.

But in statistics, we generally (due to tradition) think in terms of 95% confidence intervals. And what is the 95% confidence interval on Swisher's OPS+ for 2009?

Roughly somewhere between 65 and 155.

I could have told you that. Chuck LaMar could have told you that if he'd ever heard of OPS+.

Virtually every player in MLB falls into that interval -- some only because they'll never get a chance to show us how much worse than a 65 OPS+ they can hit.

And the 95% CI on Betemit? Roughly 55 to 140.

And the difference between Swisher and Betemit? Well, the difference in mean OPS+ projections is 16 but the standard deviation on that difference is, roughly, +/- 30. In short, we don't have the slightest clue which one of these is the better hitter and we'd expect Betemit (as a 3B) to outhit Swisher about 40-45% of the time. (I'll caveat that I didn't control for differences in sample size at all)

That's the scary thing about baseball -- small differences are hugely important but the data is generally insufficient to give us adequate power to detect those small differences. Noise outweighs signal by a lot.

And frankly, maybe I shouldn't give this trade so easily to the Yanks. ZiPS projects Betemit to a 99 OPS+ which is a little below median for a 3B (about 105 last year for those with 400+ PA). Swisher is tough to tie to a position, but the 1B/LF/RF/DH combined is 124 which puts him just below-average as well -- interestingly, a 115 OPS+ is smack-dab between Mike Jacobs and Josh Willingham and I think it's fair to say that Williams got much better return for Swisher than the Marlins got for those two guys. (Note, if you limit it to LF/RF, the median is still 123)

That last bit about Jacobs and Willingham is interesting. Three inexpensive corner players, aged 28-30, projected to OPS+s ranging from 106 (Jacobs) to 116 (Willingham), have been traded already this offseason and none have brought a substantial return. Jacobs (especially) and Willingham generate their offensive value in very different ways than Swisher and Swisher has substantially more defensive value (again especially relative to Jacobs) so it makes sense that he'd bring the highest return but ...

yes, this appears to be fair market value for Swisher.

Maybe these types of players are undervalued on the market right now or maybe enough teams have caught onto the value of defense and the availability of decent-hitting corner players that it's just not that big a deal to acquire one or let one go.

I'll reiterate what I said last year when Swisher was traded -- if he's an average-ish defensive CF, then obviously he's (potentially) quite valuable and corner hitters are the wrong comparison group.

And continuing the theme in #123 ...

There are a lot of defensively-challenged corner players on the FA market this year -- Manny, Dunn, Burrell, Ibanez, Giambi -- plus Teixeira. If you are a team looking to trade such a player (whether for money or to open a spot for another player), the market is already pretty saturated and maybe that means you should move quickly. Someone earlier made the point that maybe one thing Williams liked about this deal is that it could get done quickly and that may indeed be the case.

So I think I've talked myself around on this trade.

First, the Sox seem to have gotten "fair market value" for Swisher. Note that in itself doesn't make it an even trade, it just means that I don't think it's all that likely that Williams could have gotten more in return.

Second, if you're going to believe ZiPS on Swisher, you have to believe it on Betemit and there's simply not much difference between a 115 OPS+ corner player and a 99 OPS+ 3B. So the difference here is one of defensive value with one potentially huge advantage for Swisher (can he play an average-ish CF) and one probably small open question (the difference in defensive runs between Betemit as a 3B and Swisher as a corner player).

The extra body the Sox picked up compensates at least in part for those small advantages Swisher likely has.

Note, there may be a question as to whether ZiPS (or any system) has difficulty projecting part-time players into full-time roles perhaps especially when they have gotten the platoon advantage more often than a starter would.

Third, both teams have (potentially) filled holes by trading "surplus".

If I had to bet, I'd still put money on the Yanks "winning" this trade. But, upon reflection, the imbalance here is less than I initially thought and it's definitely not so lopsided to consider it a _bad_ trade by the Sox -- i.e. it's in the gray area. The only way it becomes lopsided in the Yanks' favor is if Swisher hits (roughly) as projected AND plays an averageish-CF ... the chances of that can't be very high (25% maybe). But the chances of it becoming lopsided in the Sox favor seem very, very low (5-10% maybe?).

It would be cool if we had a poll among GM's before the 2008 season on whether Jones or Bay was a better bet to bounce back.

Well, it would be cool but not sure what it would tell us. One assumes that whatever extra information a GM has about a player, especially one they are trading for, comes from a fair amount of research by the GM and his staff. For a poll, GMs aren't going to put in that work of course. Now a poll of GMs who considered trading for Jones or Bay and who therefore did some research on them might show us something. But without that, I doubt their opinions are going to be based on too much more than stats, personal impressions and rumors (maybe they hear more accurate rumors than we do).

As to the Yanks plan -- were I them, I would still be on the lookout for a CF and a 1B. I would want an OF of Damon/CF/Swisher, somebody good at 1B (doesn't have to be Teixeira depending on how else they can spend their money) with Nady getting a ton of playing time backing up/semi-platooning at LF/RF/DH/1B. If I don't find what I want at both 1B and CF (at least one spot needs to be filled), then Nady moves into the starting lineup and Swisher is my main option at either 1B/CF -- with Gardner/Melky getting a reasonable amount of CF time if Swisher's my main guy there (with Swisher getting a fair amount of 1B/RF time when Gardner/Melky start).

It is interesting that most likely the Mets and probably the Yankees are out of the Teixeira market. I don't think the Red Sox are that interested in him either, especially if the Yanks aren't. This trade really improves the Angels' chances of resigning Teixeira.

Mister High Standards: I asked for 3:2 on what I thought was a 50/50 bet, basing that interpretation on ZiPS. You were the one saying that ZiPs was being too friendly to Swisher, in which case you would not view it as a 50/50 bet, but rather one in your favor. If your own analysis led you to believe it was at least 61/39 in your favor, then, assuming you were right, you'd come out ahead by offering me 3:2.

Yeaarrgghhhh: What might Kenny Williams know that we don't? I can see three possibilities: 1. He's hurt, in which case he won't pass the physical. 2. His core skills (bat speed etc.) have deteriorated, which any opposing scout could pick up just as easily. 3. His approach has gotten messed up, which a new coaching staff might be able to correct. In cases 1 and 2, Cashman doesn't make the trade. In case 3, well, hitters make adjustments all the time.

Gaelan: I am saying that there was no way to tell between Andruw and Bay last offseason. All you can do is take an average of the possible paths. Unless you can show me that the standard deviation among GM's in terms of the performance of the players they acquire and unload relative to their projections is greater than the variance you'd expect from random chance, I have no reason to believe a real skill exists there. (That was a mouthful, but the idea is no different than clutch hitting).

Mike Emeigh: who is discounting the risk that Swisher will suck? That is baked into his projection.

AROM--I actually would have picked Andruw, on the grounds that he was far more athletic than Bay.

The only way it becomes lopsided in the Yanks' favor is if Swisher hits (roughly) as projected AND plays an averageish-CF

Or if 2008 turns out to be a fluke?
Willingham/Jacobs are known quantities. Swisher is more of a high risk/reward acquisition. I don't think lumping them together is particularly appropriate, nor do I think they are thought of the same way by most people.

Mike Emeigh: who is discounting the risk that Swisher will suck?

Well, someone said way back at the beginning of the thread:

I'm sorry, but this is an outright fleecing.

which certainly seems as though it's discounting the not-insignificant risk that Swisher IS toast.

-- MWE

No, it's not discounting that risk at all. It's just saying that the gap between what Swisher will be paid and our best guess of what he will be worth--which includes both the (say) 30% chance he sucks and 30% chance he really kicks ass--is far greater than the gap between what Betemit + Márquez + Núñez will be paid and our best guess of what they will be worth.

Willingham/Jacobs are known quantities. Swisher is more of a high risk/reward acquisition.

Swisher's been around just as long as Willingham/Jacobs. Just because his most recent year was a turd sandwich does not make him a lessor known quantity than those guys. I like Swisher much better because unlike those two, he can play average or slightly better defense at the corner positions.

which certainly seems as though it's discounting the not-insignificant risk that Swisher IS toast.

There's a risk like that for everybody. My guess is if you looked at a bunch of players over the years who Marcel-project to hit 250/370/450 or whatever, and group them into those whose last year was crappy, whose last year was average, and who are coming off career years, you won't find any difference in what they do going forward.

In fact, I'm almost positive of it, and if it were not so then Marcel wouldn't work, and we would adjust by using different weights than the 5/4/3/2, weighting the recent years more heavily.

It seems to me that this conversation boils down to Dan R. saying "This is a steal--he's projected to be a 115 OPS+." And then Yeargh saying "Hey, there's no guarantee he'll reach 115." But what's left out of that is that there's just as good of a chance that he'll be over 115. 115 is the baseline based on what we, as outsiders know. If I say, "hey, this model shows that the sun will come out tomorrow," someone else pointing out "You're forgetting that there's a 30 percent chance it will rain!" is not adding anything. I already know that there's a chance it will rain. I'm just noting that, if I'm betting my bottom dollar, it's going to be on the sun.

It seems to me that this conversation boils down to Dan R. saying "This is a steal--he's projected to be a 115 OPS+."

Then be sure to read #122 and #123. It's a trade of a 115 OPS+ 1B/LF/RF vs a 99 OPS+ 3B -- roughly equal right there. A 115 OPS+ is only a "steal" if Swisher can play an adequate CF (and the Yanks play him there).

Willingham/Jacobs are known quantities. Swisher is more of a high risk/reward acquisition.

Did you read #122? Nobody is a "known quantity." Swisher will hit somewhere between a 65 and 155 OPS+. Roughly speaking, the 95% CIs on Willingham and Jacobs are going to be nearly identical. There's a 40-45% chance that Betemit outhits Swisher -- in raw terms.

And c'mon. Jacobs has 1600 career PAs, a 110 career OPS+ and is a high-K/low-BB power hitter; Willingham has 1650 PA, a 117 career OPS+ and is a high-K/medium walk (and high HBP!) power hitter; Swisher has 2114 PA, a 112 career OPS+ and is a high-K/high-BB power hitter. The ISOs for the three are 236/206/207. Willingham has a higher career OBP than Swisher thanks to the higher BA and the HBP. As hitters, these guys are really quite similar -- not identical obviously, but if we can't compare 244/354/451 to 266/361/472 then our conversations are gonna get mighty limited around here. And the best known quantity of the bunch is Swisher.

And true, Swisher's 600 PA in 2008 might be a fluke. Even more likely is that Betemit's 200 PA in 2008 (86 OPS+) were a fluke.

But, you know, I already said Swisher's the better player due to his defensive value -- which is why he brought a much better return than Jacobs or Willingham did.

Yeaarrgghhhh: What might Kenny Williams know that we don't? I can see three possibilities: 1. He's hurt, in which case he won't pass the physical. 2. His core skills (bat speed etc.) have deteriorated, which any opposing scout could pick up just as easily. 3. His approach has gotten messed up, which a new coaching staff might be able to correct. In cases 1 and 2, Cashman doesn't make the trade. In case 3, well, hitters make adjustments all the time.

Careful. The argument here is essentially that Cashman and Williams have the same information. Then how was Cashman able to rip off Williams?

If in fact they have the same information then the simplest (though not necessarily correct) conclusion is that they both agree that this is a "fair" trade. Williams asked for more than Cashman was willing to give up, Cashman offered less than Williams was willing to accept and they ended up here. A fleecing should only result when one GM has better information or knows how to better use that information.

So either Cashman (and you!) are way smarter than Williams or Cashman and Williams roughly agree on Swisher's value (and you're way smarter than both of them! :-)

This is not to deny that GMs get ripped off by other GMs all the time nor to deny that there's probably a 75% chance I will be here saying a GM was an idiot when the next trade happens. :-)

The argument here is essentially that Cashman and Williams have the same information. Then how was Cashman able to rip off Williams?

The same way he always does--it's some sort of voodoo.

So either Cashman (and you!) are way smarter than Williams or Cashman and Williams roughly agree on Swisher's value (and you're way smarter than both of them! :-)

No, not necessarily. They could be equally smart but Cashman could just be right in this case and Williams wrong.


Personally, I don't want to live a world where we've come so full circle from the analysis side of things that nobody ever gets to call their favorite GM an idiot.

And the best known quantity of the bunch is Swisher.

Just because adding up all star, all star, scrub over the last 3 years comes out to a line similar to someone who was consistently slightly above average doesn't mean those players are essentially similar. Yes I read 122, yes I'm aware that no one is really a "known quantity" but I think you are being purposely difficult and glossing over an obvious point with another obvious point that isn't particularly relevant.
Yes, there may be a 40-45% chance Betemit outhits Swisher but as is clearly stated in the projection above it's almost 3 times more likely that Swisher hits like a "STAR." I'm sure the difference is less stark when looking at Willingham/Jacobs but still significant and is a key to properly evaluating the players as well as the motivations for and returns received from their trades.

"Count me as skeptical though, that as a group they would have picked the right player."

AROM you god damned apostate, don't you know the way forward in the stat community is top be 100% deferential to all inside baseball decisions and realize that they're depth of knowledge and intelligence is beyond the capabilities of our puny comprehension?

Being critical of GMs is so 1998. Get with the program, only suck ups get ahead.

AROM--I actually would have picked Andruw, on the grounds that he was far more athletic than Bay.

Was that before or after he became a whale?

No offense intended to whales, of course.

Just because adding up all star, all star, scrub over the last 3 years comes out to a line similar to someone who was consistently slightly above average doesn't mean those players are essentially similar.

Then you don't believe any of the projection systems. Every projection system has, to my knowledge, agreed that personal trends are pretty much meaningless. The closest they come are some minor tweaks based on similarity comps. But yes, they are all based on the basic notion that "all-star", "all-star", scrub adds up to the same thing as "average", "average", "average."

Maybe Dan will be nice enough to post Willingham's percentiles.

And Swisher's two best seasons are a 125 and a 127 OPS+. As I noted above, median for starting 1B/LF/RF/DH last year was 124, 123 if yo go with just LF/RF. In those two seasons, Swisher played 1 and 59 games of CF. He was average as a hitter relative to position.

it's almost 3 times more likely that Swisher hits like a "STAR."

Of course he is, he's a corner OF. Betemit is a 3B. Those "star" ratings aren't relative to position you know, it's basically the chance that he puts up an OPS+ over 125-130.

The range of their projected performances is nearly identical -- for Swisher it's about 90 points of OPS+, for Betemit about 85 points. The shapes seem fairly similar -- Betemit has about a 16% chance of exceeding his projection by 20%, Swisher looks to be about 12%, Betemit appears to have a slightly better chance of being worse than his projection. The difference between them _as hitters_, in raw terms, is almost entirely in the difference in the mean projections ... and the difference in the mean projections is almost precisely the difference between "mean" corner OFs and "mean" 3B.

Based on their ZiPS projections, Betemit is basically every bit as good a hitter as a 3B as Swisher would be as a LF/RF. Swisher is a "star" only if he can play an adequate CF. This is a bad trade for the Sox only if Betemit is a crappy defensive 3B (and/or ZiPS is badly wrong about how good a hitter he is).

On Swisher vs. Betemit:

1. A 115 OPS+ LF is NOT equivalent to a 99 OPS+ 3B. The positional gap is (in my system) 9 points of OPS+, so it would be equivalent to a 106 OPS+ 3B.

2. Not only is Swisher a plus defensive LF, but Betemit is a weak defensive 3B (-6 per Sean Smith).

3. Betemit has been somewhat shielded from LHP over his career, compiling just 272 PA against them (and posting a ghastly .636 OPS) compared to 953 vs. righties (a ratio of 22%). By contrast, Swisher is at 28%. If ZiPS "knew" that Betemit has been underexposed to LHP, and that he'd face them significantly more often in a starting role, its 99 OPS+ forecast would decline, probably by about 3 points.

4. Betemit's career high in PA in a season is 412. We have no idea how his body would hold up to the rigors of playing every day, whether he'd get hurt or lose effectiveness.

So: A 96 OPS+ is 0.2 batting wins below average per 162 (in the AL). He's a further -6 with the glove, so that's -0.8 wins per 162. And say -.2 wins per 162 for baserunning (he's -.23 baserunning wins over his last 482 PA per Dan Fox's EqBRR, so that's -.33 per 685 PA, and then regress to the mean by 40%), makes 1 win below average per 162. Replacement 3B in the AL are 2.4 wins below average per 162, so that makes him 1.4 wins above replacement per 162. Since he's never been a regular, we just have no idea how durable he is, so I'll just pull a number out of my head and say 130 games, which is now 1.1 wins above replacement per year. Finally, I'll knock off 0.1 for the possibility of declining effectiveness as he tires due to greater playing time, and wind up at exactly 1.0 wins above replacement per season. So at $4.8M/win, his market value is $4.8M. He's in his second year of arbitration, which means he should earn 60% of his market value, or $2.9M, and then 80% of it or $3.8M in 2010, after which point he is a free agent. That means he has $1.9M of surplus value in 2009, and $1.0M of surplus value in 2010. Total surplus value of $2.9M--compared to $34M for Swisher.

Cashman was able to rip off Williams because he interpreted the available information better than Williams did.

Walt, as I mention in the explanation, STAR is equivalent to being in the top 20% of the starting players at the position listed for the player offensively, so it is based on position. Now, as a RF, Swisher's STAR % goes down.

No one's mentioning that when we discuss Swisher's and Betemit's OPS+ we should really consider that Swisher's is way more tilted toward OBP.

Upon further reflection, I think the variablity of Swisher is a positive for the Yankees, though it proobably was a negative for the ChiSox. When you think of it this way, the trade makes more sense.

Given their budget, the financial committment to Swisher is negligible to NY.

If they were given the choice of a guaranteed 115 OPS+ player or a guy with a 30% chance of busting (90 OPS+ or less, or injured), a 30% chance of a 115 OPS+, and a 40% chance of 125 OPS+ or more, they'd take the risky guy.

If Swisher rebounds he's a cost controlled near all-star, if he busts, they get rid of him, if he's averagish they use him as a 350-400 AB bench guy, where he has real value due to versatility.

Now, Chicago, given their apparent budget constraints, wants more certainty. They'd choose the guaranteed 115 OPS+. So, it makes sense they want out from Swisher's contract, b/c he's not that guy.

An issue I have with the proposition that the sequence star-star-scrub is equal to scrub-star-star is that the projection systems all use as their base players who actually had a decent amount of playing time in the year being projected. In the historical database used, guys who lost their skills and didn't play the year after get tossed out of the equation because there is no data. So there is a survivorship bias to the projection method. It's like doing a study to project the health prospects of 50 year old smokers in their next five years and ignoring the one who died.

Keith Law says Swisher's bat speed was down last year. I don't know if he's right, but it's something to consider.

The "bad year" phenomenon is more of a problem with older players. Like Frank Thomas - all the systems are probably going to project a decent year. All that means is that if Thomas stays healthy and hasn't had his skills erode to where he can't hold down a job anymore, then he'll have a decent year. Gee, thanks.

If somebody can correct me on the methodology, please do so.

Being critical of GMs is so 1998. Get with the program, only suck ups get ahead.

Cashman was able to rip off Williams because he interpreted the available information better than Williams did.

Yeah, having a nuanced point of view about a trade really is exacly like being a suck up. Dismissing the fact that there might be something going on beyond the numbers when a trade seems lopsided at first glance is a much better way to approach the issue - is that what you're saying?

It's like doing a study to project the health prospects of 50 year old smokers in their next five years and ignoring the one who died.

Not quite--it's more like leaving out the people who were euthanized. It could well be that many of the guys who stopped playing still had skills but were not given a chance precisely because of that horrible year.

The "bad year" phenomenon is more of a problem with older players.

Nate Silver said a while back that PECOTA puts more weight on the most recent year when projecting extremely old and extremely young players.

What does this trade look like if Swisher never returns to form, while Betemit becomes a cheaper Uribe and Marquez turns into Gavin Floyd?

Yeah, having a nuanced point of view

There's nothing "nuanced" about an appeal to authority. It's just a cheap shortcut. No less of an authority than Alan Greenspan assured everyone that there was no housing bubble. Authorities blow it all the time.

If a move appears stupid on its face, it's not up to those saying so to prove without any shadow of a doubt that it actually was stupid. It's up to somebody to come up with actual reasons why it might not be. So far all I've seen is "Kenny must know something we don't" and that's pretty thin gruel. Another thing Kenny knew that we didn't was that center field was the easiest outfield position to play (though apparently not easy enough for Swisher to play there anymore).

"This is a bad trade for the Sox only if Betemit is a crappy defensive 3B..."

Fortunately for the Yankees, he is.

Another thing Kenny knew that we didn't was that center field was the easiest outfield position to play

That's not EXACTLY what Williams said. If you're going to snark at him him, at least get it correct.

This is what he said,

Center field is actually the easiest of the outfield positions to play from a health perspective.

That's a pretty big difference. And not as crazy as it might seem. Centerfield might give a player a longer amount of time to slow down from running, a longer "amortisation" time, as opposed to a sudden stop. There is actual hard physical basis for Williams' claim.

Gaelan: I am saying that there was no way to tell between Andruw and Bay last offseason.

What if you, as a member of the Pirates FO, knew that Bay had been working very hard in the offseason to return to his prior to injury fitness levels? Nay, what if you, as a member of the Pirates FO, had managed to convince Bay that it was in his interest since he's approaching FA, to work closely with the team's conditioning staff, and also hitting coach / instructor, to get back into prior to injury shape? What if the conditioning coach working one on one with Bay throughout the offseason had submitted weekly updates to you on his progress, to the point that you knew that in terms of strength and flexibility he was approaching or had reached the levels he was in 2006?

What if based on daily experience you knew that Andruw has a fondness for Ho Hos, donuts, various other sweet confections and tasty baked goods, and a disdain for physical conditioning, and concluded, reasonably, that as he got older, his laissez faire attitude towards conditioning would start to take a toll?

I'm not saying that ZiPs should take this stuff into account. As Dan said, it's pointless. The necessary information is not, and is never going to be, available.

What does this trade look like if Swisher never returns to form, while Betemit becomes a cheaper Uribe and Marquez turns into Gavin Floyd?

If Marquez turns out to be an above-average major-league pitcher, this won't be a terrible trade for the White Sox regardless of what Swisher or Betemit do the rest of the way. I'm just not holding my breath waiting for this to happen. Williams's ability to project pitching prospects is pretty uneven - sure, he went out after Floyd, Danks, and Jenks, but he also went after Jon Adkins and Nick Masset, and he traded away Jon Rauch for the corpse of Carl Everett. I'm not holding my breath that Marquez will be a star.

Gaelan: I am saying that there was no way to tell between Andruw and Bay last offseason. All you can do is take an average of the possible paths. Unless you can show me that the standard deviation among GM's in terms of the performance of the players they acquire and unload relative to their projections is greater than the variance you'd expect from random chance, I have no reason to believe a real skill exists there. (That was a mouthful, but the idea is no different than clutch hitting).

It is the job of a GM to beat the projection systems. Any GM who cannot should be immediately fired. Now, note I am not saying that Kenny Williams is that guy, nor am I saying that, per AROm's suggestion, that more than half would pick the right guy. I'm saying that a knowledgeable human being should be able to beat the projection system over the long haul. Moreover, I think the suggestion that it is impossible is as ludicrous as Voros' suggestion that a mother's love for a child can be calcutated quantitatively and explained biochemically.

No edit, so I'll add on. Take a look at Andruw Jones' ZIPS for next year. Anyone who takes the over on that is insane.

That said it is true that my claim is not falsifiable because the root of my clain is that there is no such thing as "true talent" because talent changes over time. So it is possible for Mr. Jones to rediscover his stroke and do better than his projection. If this were to happen, however, it would be apparent to the eye and a good GM should be able to notice.

Which points to another problem with projections. It takes too long to notice if someone's ability has changed. By the time you notice it is too late to do anything about it.

I should add the point that the fact that my claim is not falsifiable is not a weakness but a strength. The falsifiability of science is bad epistemology.

Finally, I don't mean any of this as a criticism of ZIPS. If I was a GM the first thign I would do is hire Dan or ARom or someone as competent as them. I'd start every decision making process with their analysis. I just wouldn't stop there and when I was forced to choose between two players with the same projections at the same cost I wouldn't close my eyes and pick blindly. I'd look for reasons to prefer one over the other.

I think this calls for a study, Gaelan, because I really don't agree with you, but the truth of the matter is that neither one of us have any clue. Are some GM's really better than others at acquiring major league players who overperform their projections, and unloading those who underperform their projections? We should be able to address this question the same way we address clutch hitting, or pitchers' control over BABIP, etc...no?

Just to make sure we're on the same page, though, if in fact there is no statistically significant ability by GM's to beat the publicly available projection systems, will you admit that I'm right? (If there is such an ability, I would gladly admit that you are right).

It is the job of a GM to beat the projection systems.

No, it really isn't. Talent evaluation is only part of a GM's job, and at this point as far as evaluation of major league players is concerned, the variance among GMs for that skill isn't all that wide.

Take a look at Andruw Jones' ZIPS for next year. Anyone who takes the over on that is insane.

What does insane mean? Will you give me 10-1 odds? 20-1? 100-1?

I think this calls for a study, Gaelan, because I really don't agree with you, but the truth of the matter is that neither one of us have any clue. Are some GM's really better than others at acquiring major league players who overperform their projections, and unloading those who underperform their projections? We should be able to address this question the same way we address clutch hitting, or pitchers' control over BABIP, etc...no?

Just to make sure we're on the same page, though, if in fact there is no statistically significant ability by GM's to beat the publicly available projection systems, will you admit that I'm right? (If there is such an ability, I would gladly admit that you are right).

I guess I need to clarify. I'm not claiming that GM's actually have this ability (I have no idea). I'm saying that human beings have the ability anad your GM should be one of those human beings that can beat the projection system.

For instance I have no privately available information, no access to scouting reports, and no training whatsoever in evaluating baseball players and I think I could pick out a handful of players likely to do better or worse than their projections. GM's should be better than me.

Moreover I would suggest that everyone here thinks the same as I do since every ZIPS threads is full of people taking the over/unders on players. As a group my bet is we suck and aren't much better than the projections. However I also think that there are some people who could consistently beat the projections. The real trick would be to do so for a statistically significant sample, granting all of Walt's points above.

Doesn't the risk profile of a team depend on where their expected wins are? To take an extreme example, let's say you have two players A and B, where A is certain to be a 3 WARP player and B has a 50% chance of being 6 WARP and 50% chance of being worth 0. Then it seems to me that for a team that expects to win 88 games, adding player A should increase their probability of making the playoffs more than adding player B, while for an 85 win team the reverse is true. So if you accept that Swisher is a higher variance player than a typical player with his expected performance (it's not obvious to me that that is true, but I'll grant it for the sake of argument) whether his variance is a good or bad thing depends on how far, on either side, you think the Yankees are from being in expectation a playoff team next year.

I agree with Gaelen up to a point. Take injuries, the general approach of a projection system is to take all the data a player produced, whether they did it healthy or not, and assume that the player has similar chances of being healthy/hurt in the future, producing a weighted average from the data.

I don't think GM's could beat this for all players, and probably not for most. But certainly for a subset of players, they should have enough additional information to beat the projections. One example is John Smoltz, I think I've got a blind projection for him of about 130 innings and a 3.50 ERA - I think everyone will take the under on that. Adam Everett's defensive projection should have given him some bids last year even despite his lack of hitting, but he didn't sign for much with the Twins.

Maybe the GM's weren't aware of his defensive stats, but it's also likely that they thought recovering from a broken leg would diminish his one skill.

These are the examples I can think of, with publicly available information. GM's have access to more than we do, but let's not go too far here. They make plenty of mistakes, and those of us who crtiticize moves around here are right an awful lot too.

On Andruw Jones, we can look at him now and say his 2007 signaled a real collapse because he's a lazy fatass. But he's been a fatass for years now. As for what GMs thought of him, all I know is one GM didn't think he was worth offering arbitration to, and another thought he was worth 18 million.

One of the arguments I have had with Szym over hte years is the shape of production. Andruw Jones was ugly in his PAs. Unless he makes a significant physical change, he's not going to be good. *Watching* the players, rather than merely studying the career and year-to-year stat lines would tell you which was going to be good and which one was not.

Maybe the Dodger hitting coach said he could fix Andruw in 10 minutes. But it was a bad selection at the time, and plenty knew it.

Chris - did you get my email thanking you for the link you sent me?

Again, Chris: unless we actually have projections informed by such subjective observations, it's impossible to know whether they really add anything or not. Care to give it a whirl?

Gaelan: I am saying that there was no way to tell between Andruw and Bay last offseason. All you can do is take an average of the possible paths.

Welll...

1: Most of what I read said that Bay was hurt in 2007- and that effected his production- Jpnes was either out pf shape or prematurely aging- or both
2: I'm was in 3 different fantasy leagues- in all 3 EVERYONE thought Jones was toast- Bay was drafted as if 2007 didn't happen.

I think there was a way to tell the difference between the 2- but you are not even willing to look at the possibility.

Hm, in my league, which is highly competitive, Bay went for $14 and Jones for $15. In ESPN live drafts Bay went at #94 overall and Jones at #105; again, not much difference. It was, in my view, just as reasonable to think that Bay was suffering from the early "old players' skills" deterioration that has been discussed at length on this thread (particularly given his late debut), while Jones' fielding ability and general athleticism bode well for his rebound and longevity. Your fellow owners were either very lucky or very good!

To be clear, I'm not saying ALL available information is included in projections. As AROM notes, we obviously won't be penciling in Tim Hudson for many innings next year, regardless of what PECOTA thinks. But I do think that the burden of proof here is on the prosecution: that until someone can demonstrate conclusively what subjective information from what sources enables them to beat PECOTA/ZiPS/CHONE systematically and consistently, I'll stick with the null hypothesis that they can't. (Certainly, I think the projection systems could be substantially improved by including dummy variables for specific injuries--any pitcher coming off Tommy John surgery should of course have a comp list chock full of TJS survivors!)

To be clear, I'm not saying ALL available information is included in projections.

Of course not, can't be done.

while Jones' fielding ability and general athleticism bode well for his rebound and longevity

The big problem is that athleticism has eroded pretty quickly

Dewan's Plus/Minus had Andruw at 15 outs/22 bases (so 13 runs) above average in CF in 2007, which was a pretty strong indication that his athleticism was still intact. If his body were clearly falling apart, he shouldn't have been able to cover that much ground.

Is there anyway to prove which defensive ratings are more accurate? For example, Zone Rating shows Andruw Jones as being 16th among qualified CF in 2007.

Is there anyway to prove which defensive ratings are more accurate? For example, Zone Rating shows Andruw Jones as being 16th among qualified CF in 2007.

Which ratings, used as the basis for projections, best predict next year's team-level defense? For team level defense I would use park adjusted DER, also adjusted for the pitching staff's mix of groundballs, flies, pops, and liners.

No, I haven't done it. Don't have the time, but I'd sure be interested in the results.

What are we using to compare the metrics now? Mostly y-t-y correlation (how well does it predict itself), which ones pass the smell test (have the least number of shocking ratings), or the assumption that because one uses more detailed input data, it must be a better metric.

All assumptions that could use a good objective test.

Walt Davis,

I don't agree with your suggestion that a third baseman (Betemit) who puts up a .248/.318/.457 line in 316 PAs has nearly the same value as a right-fielder (Swisher) who puts up a .253/.366/.454 line in 581 PAs. OPS+ adjusted for position is a very crude tool, and one that really does not help in analyzing the situation. Once you take into account playing time and the appropriate weighting of OBP, the right-fielder's projection is far superior.

There have been many players with old player's skills who have faded early. Phil Plantier. Randy Milligan. Swisher doesn't quite fit the bill because he started out as a centerfielder. In this, he is more like Rick Monday or Chili Davis.

Swisher doesn't quite fit the bill because he started out as a centerfielder.

Actually, he didn't. He was primarily a right fielder in the minors, and didn't put in any significant amount of time in CF in the majors until 2007.

-- MWE

until someone can demonstrate conclusively what subjective information from what sources enables them to beat PECOTA/ZiPS/CHONE systematically and consistently, I'll stick with the null hypothesis that they can't.

I think this method artificially limits the space of analysis and badly hurts our ability to understand baseball.

Why must we always return to "true talent"? Why must the "true talent" to "systematically" beat projection systems be demonstrated before we can ask specific questions about specific situations? This is problematic for two reasons.

First is a simple fact. True talent doesn't exist. People and organizations are always in flux. True talent is a chimera, a theoretical construct built from heterogeneous things to help us understand certain other things better. We need to recognize that true talent is something we created, and ask for which situations is it useful to ask about true talent and which uses of the concept of true talent will help us better understand baseball?

Second, following from that, it is again false that something which can't be demonstrated systematically doesn't exist. A team can make a good decision about a particular player that correctly weights their information, but also make incorrect decisions elsewhere. Certain talent evaluators can be better and worse at evaluating different players based on different information, certain organizations can develop temporary structures that are more effective at doing certain kinds of jobs but not keep these structures together long enough to be shown "systematically" effective or they may be deployed in different ways toward different ends.

(This is my take on "clutch" as well - the world is full of entirely real things that are also in flux and impossible to demonstrate via statistical methods.)

When we take incredibly complex topics such as talent evaluation and decision-making within baseball front offices, and the likely production of ballplayers, and we limit ourselves to only asking questions about things that have already been demonstrated to be true via certain statistical measures, we will end up with a profoundly impoverished understanding of baseball.

We have to be modest in our claims when we're talking about things without this type of statistical support, and we need to be constantly self-critical that we're not just reproducing problematic forms of conventional wisdom or our own fanboy and fangirl interests (my GM is awesome!), but we need to be asking all these questions. That's how we're going to have the most fun with baseball, that's how we're going to understand it best.

Primey for 173?