Trade deadline post-mortem: what worked, what didn’t?



Now that the summer trade deadline has passed, we’re starting to reflect on how our model matched up to reality. We’re still catching up with all the trades, so the data is not yet complete. But in the meantime, we can share a few observations.

So far, it looks like the majority of the trades fit our model quite well, especially the ones in the small-to-medium range. Overall, the data suggests it’s following a normal distribution curve, where there’s a cluster in the middle, some outside of that range, and then a few outliers. This is normal.

Those that fit well:

  • Shane Greene to the Braves was a nice fit, value-wise, especially after the chatter of the Tigers wanting a top prospect like Carter Kieboom, where the value gap was enormous – it was good to see the Tigers come back to reality on that one.

  • Tanner Roark to the A’s was also a close fit, especially when we learned cash was included. Others, such as Roenis Elias, Sam Dyson, Jesus Aguilar, Nate Jones, etc. all fit our model.

Our benchmark is to match or exceed the correlation standard set by the major projection systems, such as PECOTA, ZiPS and Steamer. So far, we appear to be slightly above their benchmarks, according to analyses such as these. 

The misses seemed to fall into two categories: high-need/marginal-win value situations or information gaps.

High-need/marginal-win value situations:

The Braves needed relief help badly, so much so that they agreed to take on all of Mark Melancon’s money (and negative value). Interestingly, they chose that route rather than pay in prospect capital. The marginal-win value case applies here too, in that the Braves value Melancon’s expected win contributions going forward in a playoff race and postseason run more than other teams would, or in other circumstances.

The big trade of the day, Zack Greinke to the Astros, was a case of both high-need/marginal-win value and information gap. The Astros want to win another World Series, and research such as that provided by fivethirtyeight suggests that teams in their positions should indeed pull out all the stops to improve their chances of doing so. The information gap aspect was in the money: there are conflicting reports about how much money the Astros are responsible for going forward. Our pre-trade valuation on Greinke was low, based on the amount of money that has been and will be deferred, per Spotrac. By our calculations, that would mean they’re responsible for about $83M (discounting for inflation). Yet we’re now hearing that the Astros are only on the hook for $53M, which changes the picture dramatically – there’s about a $30M difference there, which explains the prospect package they gave up.

Information gap situations:

The trade of Jesus Sanchez and Ryne Stanek to the Marlins for Trevor Richards and Nick Anderson illustrates this point. Sanchez is a consensus Top 60 prospect by all major evaluators; that translates into a higher value than what he was traded for, which suggests that the Rays knew more about him than the evaluators did. (We’re not prospect evaluators ourselves; we base our valuations on their ratings, so if they miss something, so do we.) Culling the roster for an anticipated roster crunch also played into it, as well as filling a need for an MLB-level starter in a playoff race.

We consider these trades outliers. Major-league baseball is a market that is small, closed, and inefficient – compare it to, say, the stock market, which is large, open and efficient. It’s small, because there are only 30 actors; it’s closed, because they keep a lot of information private; and it’s inefficient for those reasons and because it’s largely illiquid – it’s hard to move or acquire an asset sometimes.

So in markets with these characteristics, there is going to be high variance. That is an economic truth. We therefore accept the variances, in the hopes that for the most part, the rest of the data points fall close to the mean.

Because it’s an illiquid market, it means some teams don’t match up with others very well. The Yankees, uncharacteristically, couldn’t find any moves to make. We think that’s because they didn’t have the right prospect capital to work with – they are holding on to Deivi Garcia (rightly so); below him, there is a gap and a package of assets that may or may not be attractive to anyone else. Complicating matters for them, guys like Greinke and Mike Minor have NTCs that block them from being traded there.

It’s also significant that pieces that were expected to move, like Madison Bumgarner and Zack Wheeler, did not. We believe that’s because their true value was under 11.6 (in both cases, we have them in the 7s), and 11.6 is our estimate of the value of the compensatory draft pick the Giants or Mets would receive if they issued those guys qualifying offers at the end of the year. So that was the floor for trading, and no one wanted to meet it, because it was an overpay.

So even though those were not deals we can capture, we think there’s something there that validates our model as well.

We also know that the summer deadline period is much more volatile than the offseason, because the divide between buyers and sellers is more pronounced, and the marginal-win concept takes hold. In the offseason, there is more stability and equilibrium. That’s why we back-tested the model in the offseason, because the findings are more reliable in a stable environment than in a noisy, volatile one.

That also means that when you test the model in a volatile state, there will be higher variance.

Having said all that, we’re going to review all the inputs of the model now that we have more data to work with, and fine-tune a few things to make it more accurate than it is. We’re very open to feedback, and want to be as transparent as we can, so if you have any suggestions on how we can continue to improve BTV, please let us know: contact us at [email protected]


John Bitzer

Founder and editor


About the Author

John Bitzer

John Bitzer

Founder and editor of
  1. Nate Dub

    Trading for a need will almost always require an overpay in value. The one problem I have with models like this is that it assumes that a trade HAS to be equal value in order to make sense. It doesn’t always work that way.

    The Cubs notoriously overpaid for Aroldis Chapman. It was a major need for them, and they won the World Series. In any market, baseball or otherwise, there has to be some benefit to the controller of commodities to relinquish control. This is supply-demand at its finest. It makes little sense for the Pirates, for example, to give up Vazquez for what would be considered “fair market value”, when they control him for several more seasons. They SHOULD expect teams to significantly overpay for him, or tell teams to go find a closer somewhere else.

    And trading valuable future talent for immediate roster help is just something that successful teams have to do. It’ll rarely be a perfect fit, value-for-value wise, but it’s also par for the course. I think the model/algorithm/program has to take into account team needs and other factors that often require an overpay going forward.

    I don’t know how you do that. Perhaps it requires a downgrade in the value of prospects (which I think are far too high in many cases… surplus value on prospects is a laughable exercise), perhaps it requires a better formula for desirable players available for trade, I’m not sure. But I think you have to reflect need-based overpays in some way on here.

    • John Bitzer

      Thanks for the feedback, Nate. Great points. I agree with you.

      Anticipating overpays is obviously tricky, in that some teams will pay more than others based on need. Coding that would quickly become an exercise in game theory, and require a more dynamic model that flexes with each team move. That’s something we’d love to do if we can.

      In this phase, we’ve hewed to a mean-regression line that factors in those needs in an aggregate sense. To address your Vazquez example, we know it would take an overpay for the Pirates to move him, which is why his number reflects two built-in premiums already (one for an extra month of service in October, another that reflects that he would go higher than fair value due to market pressure). The fact that he didn’t move suggests that the Pirates wanted even more than that (and by the way, they may also be influenced by trying to compensate for the disastrous overpay of the Archer trade). But the Dodgers didn’t get him because they didn’t want to pay that price, and the fact that we have Lux significantly higher validates that decision.

      That also suggests that, in general, our prospect values are not too high. Lux didn’t move. Deivi Garcia didn’t move. Tucker and Whitley didn’t move, even for Greinke. Lewis and Kiriloff didn’t move. Carter Kieboom didn’t move. Teams place a high value on these guys, and our numbers reflect that. Ask a Brewers fan if they’d want to trade Keston Hiura for, say, Matt Olson (who is ranked higher than Hiura on the Fangraphs’ trade value list). I did — and not a single one would do that deal. I’m an A’s fan, and I would do it in a heartbeat. Granted, that’s anecdotal, but I do think it passes the smell test.

      There’s been a trend in the market for a while now to hoard top prospects, because they’re so valuable. Lux for Vazquez was a non-starter for this reason. There was no equivalent this year of a Gleyber Torres for Chapman move, and I doubt we ever see that again. Teams place a high value on that young impact talent, so we think we’re okay with our valuations there.

      That said, we see that teams are not at all gun-shy about moving middle-tier and lower-tier prospects for immediate help. The Greinke package was full of those lower-impact guys. Overpays based on quantity rather than quality are less painful, so we may factor that into future numbers.

      Anyway, thanks again for the feedback. We’ll continue analyzing everything and fine-tuning it.

      • Nate Dub

        I’m more than aware of how teams value young talent, and from an economic standard, I get it.

        But teams will always value their young talent way higher than the players are probably worth. It would make no sense for the Yankees to lessen Deivi Garcia’s value. From their standpoint, they don’t want to part with him unless “the price is right”, which is more or less to say, unless we can get someone like Noah Syndergaard straight up for him. And there’s a lot of other factors… marketability being a key factor.

        For example, the Yankees have a scout who just compared Jasson Dominguez to Mike Trout and Mickey Mantle. As stupid and insane of an expectation as that is, it’s out there, and now, the Yankees HAVE to place an extremely high value on him, if for nothing other than the fans. Could you imagine if they traded Dominguez for Robbie Ray? You keyed on this by bringing up Hiura. The Brewers would be taking a huge risk to move him. And thus, his value is insanely high.

        Teams also value their guys, especially their top guys, at their 100th percentile expectations. I don’t think we should, or can, dance around that. Gavin Lux has been lights out in the minors, but his .519 BABIP is comically unsustainable. His .358 BABIP he had in AA might not be all that sustainable, either. But how much of the Dodgers’ valuation is wrapped up in the numbers? He’s probably not going to sport a 140 wRC+ in the majors long-term. But the value is wrapped up, in many ways, in that 100th percentile expectation.

        Of course teams aren’t moving those guys when they have huge expectations, and fans have huge expectations.

        I think part of my disagreement comes from the way we talk about value. Prospect value is largely, if we’re being honest, “perceived value”… because it’s largely unmeasurable in any objective sense. It’s a projection-play. Whereas MLB players are based on “field value” (your own term, which I like). It’s insane to equivocate the two.

        If a teacher gave me a 97% in math class because she thinks I’ll be really, really good at math, but I’ve never been to a single class, but you an 86% because you were good, but you showed up, worked hard, studied, and felt good about getting an 86%… you’d probably say there’s something wrong that you get docked for performance and I get aided simply by my projection.

        I think the formula is fallacious when this is how we value different players.

        • John Bitzer

          Again, fantastic points. And yet, there is a practical side to it. Prospects do get traded for veterans all the time, so we need to have some logical way of equating the two. As we mention on the site, we are basing our prospect valuation model on what appears to be very solid research by Craig Edwards of Fangraphs — research which has been validated by a similar study by some very smart guys at The Point of Pittsburgh.

          Individual cases will of course vary, based on information gaps and the wider range of outcomes naturally associated with prospects. Your point about 100% percentiles touches on the fact that it’s inherently an exercise in probabilities, which is absolutely true — and which GMs have acknowledged. Those probabilities are baked into our models already, in the sense that they’re baked into the ratings of the sources we use for them. But I agree with the essence of your point, which I interpret to be: applying a number (which represents a sense of certainty) to something that is inherently uncertain is challenging, and carries with it the same sense of uncertainty. But then again, that is what probabilistic modeling is all about. We’ll try to be clearer about that going forward.

          Anyway, thanks for the fascinating discussion.

          • Nate Dub

            Your interpretation only covers half of my point.

            Yes, applying a number to something uncertain is a challenge, BUT (and this is where I think we part ways in our opinion) it’s not a fair stance to say “Felipe Vazquez, who has been stellar in the majors and has actual numbers is worth the same as Keibert Ruiz, who hasn’t played one single game in the majors.”

            It’s no wonder the Pirates wanted a bunch of guys for Vazquez. And if the coin was flipped, the Dodgers would do the same.

            One party gets evaluated on what they’ve done, and the other gets evaluated on what we think they’ll do. And this site treats those values as similar. They aren’t. Very rarely do teams in baseball see them the same (as evidenced to the asking prices for actual MLB talent).

            The truth is, the unwillingness to move prospects like Lux, Garcia, Kieboom, Tucker, etc. has less to do with “surplus value” and everything to do with potential. As a Giants fan, I would hate to see Joey Bart (74.2) for, say, Andrew Benintendi (58.6), but I also don’t think Bart has more value than Benintendi (despite my unpopular low view of Benintendi). I just don’t want to see the potential of Joey Bart go somewhere else.

            When I see this sort of discrepancy where a prospect drafted in 2018 is worth more than a guy who has been worth almost 10 wins in not even 3 full seasons (there’s a very real and possible outcome in which Joey Bart isn’t worth 10 wins his entire career). I have to wonder, are player values depreciating simply because they become major leaguers?

            When unknown commodities are worth more than known commodities, there’s a problem in the matrix.

  2. Kevin Chen

    Hey John, love your work, a couple questions that you may have heard before.
    1. For minor leaguers, should the age, league, optionability, or 40-man roster/Rule 5 status be listed? I consider these to be situational factors that influence what these prospects are worth to their own or other teams. For example, the trade for Ramon Laureano would be viewed as highway robbery for the A’s according to this site’s valuations if the reader didn’t know that Laureano would otherwise be eligible for Rule 5, and he would definitely have been taken. The Brewers traded Domingo Santana at a discount because he had no position in their OF and was out of options while Ben Gamel, whom they received in return, was not.
    2. When listing a player’s remaining salary, why use the full value of the contract and not account for any amount that has already been eaten by a previous team? The amount of surplus value a guy like Mike Leake would have to the D-backs should account the fact that they are only paying him 6 of the remaining 18 million through 2020, which is a much better deal even for his production in a vacuum.

    • John Bitzer

      Hi Kevin,

      Thanks for the feedback. Great questions. Let me take them one at a time:

      1. It certainly wouldn’t hurt to list those factors, you’re right. Behind the scenes, those things are influencing the value. Age and league/level are definitely factoring into the source ratings we use, and Rule 5 risk and optionality are factoring into our own calculations, which we talk about it in our methodology here: Showing that data is more a matter of bandwidth for us — there are only two of us updating 2,400 players manually, so it’s very time-consuming, but we can try.
      2. Yes, we typically do that once a trade occurs with cash paid down — we just haven’t updated Mike Leake’s numbers at the time of this writing. See Zack Greinke. Once we slog through all the trades that happened at the deadline, all will be updated to reflect current salary obligations, such that it’s clear going forward what any team trading for them would have to pay. (Note, however, that sometimes it’s not clear: sometimes cash is included in a multi-player deal, and it’s not tagged to any particular player, so even though it may even out that particular transaction, it may not carry over into any one player’s data after that). Anyway, Leake will be updated shortly.

      Thanks for the feedback!

Submit a Comment