he fact of the matter is that the talent represented by compensatory picks awarded a team losing a star player to free agency very rarely match that of the player lost.

It's shoddy analysis of this type that gives Sabermetrics a bad name. It's not a question of giving up Roberts for less value, it's a question of whether the value of those compensatory picks is greater than Roberts at the cost of $50M over 5 years, or whatever his free agent value turns out to be.

And then Kerry compounds his error by treating contributing players, who have huge value, as if they are afterthoughts, essentially assuming that if you don't get a star from a compensation pick the pick was worthless. When the average MLB starter makes close to $6M a year, getting control of one average MLB starter at the MLB minimum for 3-4 years is alone worth $20M+.

Perhaps it's better to trade stars earlier, but Kerry makes no comparison of the value you can receive in trades versus the value you can get in compensation picks. So his entire point is moot, meaningless, pointless, futile, insubstantial, without meaning, or point...

Lili Von Shtupp: You're finished. Fertig! Verfallen! Verlumpt! Verblunget! Verkackt!

While your point is correct, you do have to keep in mind that you don't get the star or the contributing player. What you get is a lottery in which you have a 1.5% chance of getting the value of a star, a 7% of getting a regular contributor, a 30% chance of getting a journeyman/cuppa, and a 62.5% chance of getting nothing. Depending on what values you use for star/contributor/journeyman/cuppa, you are still looking at maybe a value of say $3-5 million as I sketch things on the back of the envelope (60 million for star, 20 million for regular, 2 million for journeyman/cuppa, and nada for the misses). You'd want to compare that, of course, to the value from retaining and the typical value for a trade option.

The first thing you need to do is throw out the 64 supplemental picks from 2006-08, because only a handful of first-rounders from those drafts have even made the big leagues yet (though one supplemental from '06 was a cat named Joba Chamberlain).

So now you're talking about 4 out of 196 "stars" (however they are classified), not 4 out of 260; plus 18 "contributing level" players. I don’t know how those are determined, either, but here are more than 18 sandwich-pick players I wouldn’t mind having spend their best and cheapest years on my team

Clay Buchholz
Adam Jones
Aaron Rowand
Kelly Johnson
Scott Hatteberg
Brad Wilkerson
Mark Teahen
Jarrod Saltalamacchia
Jeff Mathis
Aaron Heilman
Dustin McGowan
Chris Duncan
Casey Fossum
Jerome Williams
Jason Marquis
Gabe White
Jim Parque
Matt LeCroy
Trever Miller

And missing out in this analysis is that star players are usually type A free agents - and get more than a sandwich pick as compensation. Usually they fetch either a late first or an early second round pick, although there are exceptions when a team signs multiple type A free agents in one offseason.

This kind of draft analysis always misses a major point: some prospects will have trade value before it's clear that they aren't major leaguers. Right now the Giants could probably get a decent middle reliever for Nick Noonan. Will Noonan ever amount to a single win share in the majors? I have no idea, but that doesn't mean that he'll never have value to a big league club. The Indians could have used Adam Miller as the centerpiece of a blockbuster deal at one point. They chose not to, but that doesn't mean Miller was useless because he wasn't producing at the big league level.

is #5 common knowledge around here? it seems so simple, but I don't think I ever made the connection.

Regarding #5, I do not think that the consideration of trade value changes the mean value of a sandwich pick. If the Giants make a good trade, where they give up a prospect who ultimately fails and receive a solid contributing middle reliever, they get more value, but that's not value internal to the pick - it's value created by a good trade. They could just as easily make a bad trade, where the middle reliever puts up an ERA of 6.5, loses a couple close games, and heads for the DL while the sandwich pick makes a couple of All-Star teams.

Also, I'm pretty sure philly's draft studies already covered this point. (from SoSH)

IIRC, compensation picks actually don't or only barely create value commensurate to slot bonuses - bonuses are way too low at the front of the draft, but don't go down at the proper rate as you head into the 20s and 30s.

It may simply be too complicated a question to answer in the aggregate. The question "what is the average return on the Nth pick in the draft?" is probably fine to answer in the aggregate. But the question "Should team A: (1) keep player X for their final year, paying them C1 $ and offer arb and get the picks receiving the average return for said picks (which are yet to be precisely determined); (2) keep player X for their final year, paying them C1 $, and not offer arb; (3) re-sign player X for B years and C2 $; (4) trade player X before the season for players Y1, Y2, ... Yi; (5) trade player X mid-season for players Z1, Z2, ... Zj?" is just a wee bit hard to handle in the aggregate because B, C1, C2, Y1-Yi and Z1-Zj aren't ever gonna be in your model. Oh yeah, some of those trades might involve you paying money and/or taking back a bad contract so that has to be factored in (call that C3). And the signing bonuses for the picks (call that D). I'm not even gonna bring up the PR aspects (which I just did so call that P :-).

Now, on a case-by-case basis, we at least have some notion of whether the team is likely to be competitive in the coming year and the values of those missing variables. The average value of (roughly) the 16-30th pick (you can't receive a 1st round pick higher than that), the 31st-whatever compensation pick, and 2nd round picks is one variable to consider.

But if you were gonna try to answer the big question in the aggregate, you'd want to look at all cases where a player was entering their final offseason/season with a team and see what the average (baseball) value was of each of those 5 options. I still don't think that really tells you anything useful about what to do in any given case. For example, even if you knew that the average return on a mid-season trade was worse than the average return on the picks, it doesn't tell you anything about what to expect from the specific players you're being offered (Bagwell, Kazmir vs. Dewey, Crapout and Howe). But at least you're trying to cover the appropriate scenarios.

(It's pronounced Cra-poe)

They could just as easily make a bad trade, where the middle reliever puts up an ERA of 6.5, loses a couple close games, and heads for the DL while the sandwich pick makes a couple of All-Star teams.

Right, but in your scenario, the sandwich pick will show up as an overall success in a study that is designed like the linked study. By tying the sandwich pick to a binary definition of "good" or "bad" -- or an arbitrary definition of "star", "useful player", etc... -- a study misses the value of prospects whose fates are unknown. I'm not saying it's going to make sandwich picks seem 500% more valuable, but it's an overlooked point.

My favorite example is Dan Meyer, who was the key piece of the Tim Hudson trade. If the Braves don't offer arbitration to Steve Karsay, they don't get a sandwich pick. If the Braves don't get a sandwich pick and draft Meyer, maybe the A's take a better offer for Hudson. (Or, heck, maybe the Braves substitute Brian McCann, but now we're getting a little too high off the hypothetical gas...) The Braves won their division by two games in '05, and it's entirely plausible that the difference between Hudson and the alternative was the difference between making the playoffs or not. But in this study and the SOSH study, Dan Meyer is a data point confirming the theory that sandwich picks rarely help the drafting team.

Right, but in your scenario, the sandwich pick will show up as an overall success in a study that is designed like the linked study.

Yes, and it will actually be a failure for the team that made the pick. In the Dan Meyer case, a failed pick produces real baseball value for the drafting team, and in my scenario a successful pick produces zero baseball value for the drafting team. The two scenarios balance out exactly.

It is true that the value of a pick includes greater variance than these studies suggest, because it's possible to turn a good pick into nothing or turn a bad pick into something, but it should not affect the mean.

I guess it's possible that ballclubs systematically over- or under-value prospects and thus tend to make better or worse trades for prospects, but I don't know of any evidence of such.

EDIT: clarity

Right, but in your scenario, the sandwich pick will show up as an overall success in a study that is designed like the linked study. By tying the sandwich pick to a binary definition of "good" or "bad" -- or an arbitrary definition of "star", "useful player", etc... -- a study misses the value of prospects whose fates are unknown. I'm not saying it's going to make sandwich picks seem 500% more valuable, but it's an overlooked point.

I don't think that makes any sense. Surely the average trade value of the prospect, which is what we're discussing, shouldn't be any greater than the expected value of the draft pick as a player, unless you think that the market misvalues prospects.

EDIT: It could affect the size of the potential upside or downside, but it shouldn't affect the expected value.

EDIT#2: Or what MCoA just said.

But if you were gonna try to answer the big question in the aggregate, you'd want to look at all cases where a player was entering their final offseason/season with a team and see what the average (baseball) value was of each of those 5 options. I still don't think that really tells you anything useful about what to do in any given case. For example, even if you knew that the average return on a mid-season trade was worse than the average return on the picks, it doesn't tell you anything about what to expect from the specific players you're being offered (Bagwell, Kazmir vs. Dewey, Crapout and Howe). But at least you're trying to cover the appropriate scenarios.

You're right in what you write, but I think your comment misses the mark nonetheless; you're not deciding whether to accept a specific trade, but which strategy to pursue in the first place.


One thing that TFA misses, however, when it says that "Mid-season trades are always a possibility" is that mid-season (re)-signings are a possibility, too. As, for that matter, are re-signings after the player reaches free agency. The choice isn't limited to sign/trade him before the season or lose him to FA & get draft picks.

It is true that the value of a pick includes greater variance than these studies suggest, because it's possible to turn a good pick into nothing or turn a bad pick into something, but it should not affect the mean.

Ah. I get it. Assuming a bell curve distribution for the success of prospect-for-veteran trades, this makes sense. I suspect that prospect-for-veteran trades tend to provide more real baseball value to the team acquiring the known quantity, but that's just a hunch, so I'll shut up now.

Objection withdrawn. Apologies to Birds in the Belfry.

EDIT: This is why I lurk.

This should have been titled:

Sandwich Picks: Not as Delicious as They Sound

you're not deciding whether to accept a specific trade, but which strategy to pursue in the first place.

Then seems to me this is a bad way to decide which strategy to pursue since there's no reason you can't pursue several of those strategies at once. How can you decide whether the value of an offseason trade of Brian Roberts is worth more than the average value of the picks you'd get for him unless you find out what players you can get for Brian Roberts? To decide whether to pursue an offseason trade based on the idea that they rarely work out well (assuming that's true which I have no idea) seems like a very bad idea to me.

For an extreme example, trading Zambrano for Kazmir is always a good idea no matter which "strategy" is best.

Besides, all of these options occur at different times. You go to Roberts' agent in late 2008 and get an idea what he'd be looking for to extend. If it's a good deal, you probably take it. If not, you sit down with Jim Hendry and others at the winter meetings and find out what they might give you for Roberts. If you don't get good offers then you really start to debate your options but with a mid-season trade or the draft picks as a fall-back position.

Also you know that, basically, you can't do worse than the draft picks unless you screw up. OK, there's always the risk a player will accept the arb offer, you have to weigh that and it could leave you worse off. But still, in general, at the end of the day you get your picks. So it really is a situation where the question is whether you can generate a specific offer (trade or contract to the player) that provides more value than those picks are likely to provide not which is the best general approach. (You have to balance current vs. future need as well as projected revenues against that too of course.)

(Note you're not even picking until June 2010 and you're picking players of such uncertainty that the best of them are 3-4 years from the majors at least so an aggregate study of the average value is something worth relying on there.)

Sandwich Picks: Not Many Heroes on the Menu

Then seems to me this is a bad way to decide which strategy to pursue since there's no reason you can't pursue several of those strategies at once. How can you decide whether the value of an offseason trade of Brian Roberts is worth more than the average value of the picks you'd get for him unless you find out what players you can get for Brian Roberts? To decide whether to pursue an offseason trade based on the idea that they rarely work out well (assuming that's true which I have no idea) seems like a very bad idea to me.

You're getting it backwards. It's deciding whether to pursue an offseason trade based on the idea that the alternative rarely works out well.

Of course in the case of an "extreme example," you take the sensible position. But what if you're confronting a non-"extreme" situation? What if you're offered two mediocre young major leaguers for Roberts. Do you take it? Or do you shrug and say, "Nah; I'll keep him. Maybe we'll be able to trade him at midseason, but if not, I'll just take the draft picks."? You can't wait to see who the draft picks are; you need to look at the average value of those draft picks.

So it really is a situation where the question is whether you can generate a specific offer (trade or contract to the player) that provides more value than those picks are likely to provide not which is the best general approach.

But that's exactly what we're talking about!

And another way of looking at this is to compare retrospectively what the evidence shows. In other words, this analysis gives a glimpse of what history shows the value of the sandwich pick is. What have been the value of mid-season trades over the same time period? Of other alternatives? How do those values compare? Can we see whether the past decisions ended up as good ones or not, to see whether or not in the real world, GMs have been effective. Again, that may not tell you what to do in a specific situation, but it may tell you what types of moves seem to have worked out the best.

I agree with Walt Davis that context here is king. This includes having some good idea whether the free agent flying the coop will likely go to a team with a 16-30 pick (as was always almost certainly the case with, say, Mark Teixeira) ... and also having a good idea what number your sandwich pick would be in the draft. If you're the Nationals in the '06 offseason, you have an excellent idea that you'll probably get a supplementary pick in the low 30s by letting Soriano walk. If you're the Cubs, you figure on the high 40s for Juan Pierre (a steal at any price, obviously). Treating slots 31-64 as essentially the same (there were that many supplementals in the '07 draft) doesn't make a lot of sense to me.

Specifically in the Orioles case, because they are almost always a bad team, they could expect a pick in the 30s, in addition (probably) to one between 16-30. If I was wargaming that decision, I'd narrow the search (taking it back way before 1990) to:

A) picks in the second half of the first round, and
B) picks in the first third of the second round.

With "round" meaning "when the number of picks is equal to the number of teams in the league."

The results then would decidedly NOT look like a 1.5% chance of success. Though whatever you'd learn is just one piece of complex puzzle.

Sandwich Picks: Not Many Heroes on the Menu

Tasty!

I've recreated something like Kerry's database, but it was too long to post. Maybe if I remove the players who did not or have not yet reached the majors and replace them with a count by round it will be ok. I've included B-Pro's BRAR or PRAR but not FRAR because, gosh darn it, people don't like it. I have not included sandwich picks awarded for not signing the previous year's first round picks.

Jacques Jones and Troy Glaus were #37 picks.

Jason Romano was included with Gabe Kapler in a midseason trade for Todd Hollandsworth/Dennis Reyes trade. Hollandsworth and Reyes played poorly down the stretch and left as uncompensated free agents.

Joey Votto was a #44 second round pick.

Josh Donaldson was part of the Rich Harden trade. The A's desired his presence as their Midwest League catcher was called to active duty by the Navy.


* Did not sign. I don't think they should count. The Twins never got anything out of Aaron Heilman.
** Played in MLB, but had 0 RAR.

Boy are we talking past each other David. If your point is the average value of the compensation picks forms the baseline against which the value of other options is judged, then we're in agreement -- and I've said the same thing (maybe not so clearly) about half-a-dozen times now.

But what I was reacting to was this claim in the original article:

By what stretch of anyone’s imagination does it make sense (again, particularly for those teams with little hope of contending in the upcoming season) to allow a star caliber player go for a 1 in 67 chance of--some day--matching that players value? Or accepting a 1 in 10 chance of getting something of value--again, some day--in return? That, ladies and gentlemen, is a true sucker’s bet.

And that's simply dead wrong. Not only has he, for some reason, ignored the value of the other pick received and the value of the FA to the team in his final year, but more importantly, if you can't get an offer better than the draft pick compensation (plus that other stuff), then getting the compensation is not a sucker's bet or a bad "strategy" but simply the best you could do.

And then you brought in that this info tells you which "strategy" to pursue. No, it really doesn't. This is a tactical decision not a strategic one. You should always have the "strategy" of weighing the value of resigning, the value of trading offseason, the (expected) value of trading mid-season, the value of keeping and not offering arb, the value of keeping and offering arb (against the "arb risk") and money and all that other stuff (like what are the chances they'll be Type A vs. Type B, etc.). That valuation requires specific info -- not specific info about the likely value of your draft picks (and I've already said several times that aggregate info is fine there) but specific info about what trade offers exist now, guess work about what you are likely to get for this specific player in mid-season, and how much it will cost to re-sign the player. But the "strategy" of, for example, "we will trade our pending FAs during the offseason because compensation draft picks suck" is not a good way to go about things. The decision to accept/reject a specific trade offer because it does/does not suprass the average value of the compensation picks (plus the value of one year of the player, etc.) can only be made within the context of a specific trade offer not based on, say, whether such offseason trades work out on average.

Put it this way. Suppose the draft pick compensation rules changed. Suppose that the first 15 picks went to the worst 15 teams as current. Then the next picks were supplemental compensation picks to any of those worst 15 teams who lost a type A FA. Then the other 15 teams get their first pick (some of which would go to other teams in compensation for FA signings) then the rest of the supplemental picks. Under this scenario, instead of expecting, say, the 27th and 40th picks for losing Roberts, the O's might expect the 20th and 31st (assuming they are among the worst 15 in 2009).

How does that change their "strategy" or their decision-making process? I don't see that it would at all. It clearly raises the baseline against which the other options are judged but it doesn't mean you don't go out and solicit trade offers or find out what the guy is willing to sign for.

I'm not saying the average value of compensation picks isn't useful information -- it certainly is though one would certainly hope teams have had that info for years. I'm saying it doesn't tell you whether you should trade the player, resign him or let him play out the string -- that decision requires specific info.

You don't seem to be disagreeing so I have no idea what we're arguing about.

I'm saying it doesn't tell you whether you should trade the player, resign him or let him play out the string.

Then I'd want to be a GM in your division, Walt.

And then Kerry compounds his error by treating contributing players, who have huge value, as if they are afterthoughts, essentially assuming that if you don't get a star from a compensation pick the pick was worthless. When the average MLB starter makes close to $6M a year, getting control of one average MLB starter at the MLB minimum for 3-4 years is alone worth $20M+.

Nah, but he did say that according to past history, the odds of getting even one of those is around 10% (at best, even taking into account the recent draftees not getting a chance, something someone else alluded to that Kerry mentioned but the poster overlooked.) So it's not an afterthought, but still pretty Slim Pickens, wouldn't you say? Which was the whole point of the exercise...not to say that it's the end of the world, but to sit back and say 'to heck with it, we'll take the picks' might not be a very good way to go, despite how teams use that tack to pass muster with fans. As Walt points out, in some cases you might not have a lot of choices, but to sit back in self-satisfaction and say 'hey, we got these PICKS over here' usually isn't worth much more than a shrug of the shoulders...not to mention the cost of SIGNING said picks, which is going to take away some of the 'value' of that contributing player you've talking about.

Perhaps this was or wasn't 'giving sabermetrics a bad name'. but your taking things out of context doesn't seem better, does it?