Thanks, I figured it was more subtle than what I was reading into it. The Buchholz case is a little ironic because, while I agree with you that Buchholz's perceived value probably didn't change that much between late April and late June (which is I think what you were arguing), in retrospect his value really has plummeted from late June to late July. And that kind of gets back to the point you were making, which is whether the collective analytic systems of all major league teams appropriately calculated the risk of a Buchholz collapse, or of the people who follow the team on a daily basis for years were more able to see it coming.
I'll put it another way, because this is really what I've been thinking about since you made this argument. I think there are two models of how trades are made. The first is probably the more popular conception, one that is at least more entertaining, which let's call "the steal." In this system, teams build excess value by having better analytic systems than the teams with which they're trading. They trade away players who end up being worse than expected, and they acquire players who end up being better than expected. The goal is to develop a model, through some combination of quantitative stats and qualitative scouting, that maximizes predictive ability. Example: the Lackey trade.
The other model is what you're describing, the time-based value model. Call it the "futures market." Here, teams have a mutual goal of trading approximately equal value, but one where value is placed differently by different teams either because of positional or temporal need. I think this is what journalists are talking about when they talk about "fit." Example: the Peavey-Iglesias trade.
"The steal" model is really the basis for Moneyball. If the market were perfect, the A's should be dead last every year. Somehow they are able to capitalize on inefficiencies that other teams can't recognize. As you say, this is increasingly less likely as teams generate better analytical models. Further, there's a game theory component to it. If you become known as the GM who only makes "steal" type trades, nobody's going to trade with you. Under this system, as teams become more advanced analytically, the game becomes about figuring out weaknesses in individual team's models. Rather than exposing a player to a competitive bidding process which would utilize some kind of wisdom of crowds thing to arrive at actual value, you try to figure out which single team would most overvalue the player based on assumptions in their model. The way to get at this is have lots and lots of trade talks with lots and lots of teams to try to reverse engineer their models. Or save the passwords when your scout moves on and hack into their server.
The "futures market" system is a gentler, kinder MLB. Less purely capitalistic. Trade partners are identified by mutual need and everybody wins. In this system, you might be more likely to see GMs who develop good working relationships with a handful of other GMs who they trust to make reasonable trades for both sides.
Under the "steal" model, each team's estimates of each player's value should be constantly in flux. After every good start, confidence in Buchholz's value should increase (even a tiny amount) and every bad start should decrease his value. And I'm using good and bad in the broadest sense to encapsulate whatever the model deems good or bad. A start where Clay gives up 10 runs but strikes out 27 might be a good start in some models and a bad start in others. But ultimately two months of sustained good pitching should increase his value. I actually think this is where you were disagreeing with posters about the increase in Buchholz's value because it was your position that Clay had been having "good" starts the entire season but he was merely unlucky. In fantasy baseball, most of the board would have basically given you Clay for a bag of Puntos in early May (including me). None of us know how much individual major league teams incorporate FIP vs. ERA vs. scouting (physical, mental, spiritual...) and so we don't really know how much his value rose or fell during the first half of this season. We also can't even say how much his trade value fell after the arm injury because we don't know how much each team expected him to get injured beforehand. My guess is that Boston (if their fans are any indication) was on the high end of estimating his probability of getting injured, although the effect was probably small. Under the "steal" model, the team should have tried to find a team that they felt was underestimating Clay's probability of an injury and traded him a few weeks ago, maximizing their return. Under the "futures market" model, a trade like that would have ruined their relationship with the other GM because it was based on a premise that they knew he was going to be injured more than the other guy did. Who would trade with you again after that?
I don't know, maybe this is all really obvious to anybody who's taken a negotiating class and I'm reinventing the wheel, but I'm curious what other people think. In some ways the Red Sox would probably be better off relying on the futures model because they can afford to pay the market rate for a win. The problem comes when they think they're in a futures negotiation and it turns out the other GM is trying to steal. Witness Lackey. The real problem, though, is that they can't seem to decide which model they're operating under. A real futures model would target a season when they're going to compete and make trades for that time. Unfortunately they can't seem to decide what year that is.