In today’s MLB, it feels like superstar players are being traded every year. Teams are looking to jumpstart a rebuild, and the best way to do that is often by trading their best player. However, how often does a team actually get a return value worthy of the superstar himself? In this study, we looked to identify potential superstars who could be traded in the future and what their ideal return package should look like.
The idea behind this study came from the rumors swirling around the Cleveland Indians and Francisco Lindor. Cleveland has the opportunity to trade one of the best players in the league if they feel they are unable to achieve the ultimate goal of winning a World Series while Lindor is under contract. A consistently low-spending team like Cleveland will unlikely be able to afford Lindor once his contract expires after the 2021 season. With this in mind, they should likely prepare to look for trade packages that will give them a significant return of prospects who can eventually total or exceed what Lindor’s WAR (Wins Above Replacement) has been for the Indians.
We also believe that many times the discussion around star players conflates whether a player should be traded with what the return was; this could lead to faulty decision making. Instead, we believe that these should be separated out into three distinct and equally important sub-questions:
- How should a team decide whether or not to consider trading their star?
- How can they objectively value the potential returning prospects?
- If they decide to trade their star and accept this value structure, what should an expected return be?
To start, we utilized a similar process to the 2013 Bleacher Report article “Who Wins Blockbuster ‘Superstar for Top Prospects’ Trades Most in MLB?”, but with a slightly higher bar for entry. While author Jason Catania sets the threshold for “Superstar” as a +4 WAR season, we only looked at players who had had a +7 WAR season in the past, or whose trade had generated significant buzz to merit inclusion (an example of the second would be Houston’s trade for Gerrit Cole. Cole had been a highly touted pitching prospect at the time, but had only put up 2.2 WAR on average each of his years in Pittsburgh). This gave us a list of 78 trades.
Bleacher Report then used 20 trades from this list as a sample size to determine stars vs. prospect value. Similarly, we selected 26 trades (⅓ of our initial set) for our training set. From there, we collected the WAR of the superstar from the previous season, the player’s highest WAR in his career, average season WAR, and cumulative WAR in his career. We also included the WAR of other players going in the same direction as the superstar (i.e. if Superstar X and extra player A are going to the team, we included the WAR of extra player A). We also included the names and positions of all players returning in the trade. All of this data was taken from Baseball Reference. From here we began our analysis of our three questions.
Question 1: Should we consider trading?
For this question, we proposed a series of four sub-questions which captured the majority of the trades on our list:
- Is this team projected to finish more than 5 games behind the 2nd Wild Card?
- Since the inception of the 2nd Wild Card, the lowest winning rate of a 2nd Wild Card team was the 2017 Minnesota Twins (85-77). If a team is projected to finish 5 games out of the 2nd spot, they are likely going to be a .500 (or near .500 team) – however, given the nature of seasonal variance, a projected 5-game deficit was more appropriate than simply saying that a team should be projected to have a losing record.
- Does this player’s salary take up (or will take up next year if the player is in arbitration) 12% or more of the total budget?
- If all 25 players on the team are paid evenly, then each player would earn 4% of the total budget. If this player is earning 12%, then they are earning 3x an equally distributed budget. For teams with lower payrolls, expensive players (like Lindor) will likely well exceed this threshold.
- Is this player set to become a free agent at the end of the year?
- A team may decide to trade a valuable piece, given their impending free agency, rather than attempt to offer a qualifying offer or go empty handed (if the player is good but not worthy of a qualifying offer contract).
- Has this player publicly feuded with local media or internal staff (manager, GM, etc.)?
- In a discussion at a Sloan MIT conference, Houston Rockets GM Daryl Morey and Golden State Warriors GM Bob Myers debated whether or not a team should trade its star player if he doesn’t have the right attitude. Morey highlighted that star players are scarce, while Myers pointed out that unmotivated players drain those around them. In baseball, this could be considered the “Manny Ramirez Question”: in 2008, the Red Sox were fresh off a World Series victory and on track for another playoff appearance. However, their superstar Manny Ramirez got into a public fight and shoved a traveling secretary. This led to him being traded to the LA Dodgers and being replaced with Jason Bay — after the trade, Bay put up 1.9 WAR, but Ramirez put up 3.5 for LA. There was a clear decrease in individual productivity in favor of perceived improved chemistry.
If any of these four answers are “yes”, then we propose that this star player should be considered “on the block”. The team has no requirement to trade him, and may decide (after performing other internal analytics) that their team and future are better by retaining this player. We make no judgements or statements about that – instead, we are merely proposing that this four step process can get an executive in the right frame of mind to determine if they should consider trading their star. From there, they can then follow questions 2 and 3 to determine what a projected package would be and if that return would be sufficient.
Question 2: How should we evaluate prospects?
In 2013, Bleacher Report attempted to answer the same question we did with a slightly different spin: who wins superstar trades? They followed a similar methodology of excluding good-but-not-great players, and ran into a problem: they could not fully evaluate the return package of prospects. Since minor league WAR does not officially exist, most of the prospects had a 0.0 or N/A value rating, leading them to throw up their hands and say “we need to have some way of evaluating the success or failure of the prospect(s) involved relative to the major league player(s)—and trying to do so without a legitimate sample of big-league performance by the prospect(s) to this point wouldn’t be fair. In other words, sometimes it’s just too soon to tell.”
Bleacher Report exclusively utilized Baseball America top 100 prospect ranking to determine the value of prospects. However, this is a wildly subjective and flawed metric. For example, future superstar Mookie Betts (on our list) appears at #59 on Baseball America’s top 100 ranking for 2014, behind players like Mark Appel (#26), who never threw a pitch in the big leagues, or – to stay within Boston’s farm system – Garin Cecchini (#51), who accumulated 35 total at-bats over his major league career. Clearly, this system can be improved upon.
At the 2014 SABR analytics conference, a panel of Jim Callis, Jonathan Mayo, Bernie Pleskoff, and Barry Bloom discussed the difficulties of evaluating minor leaguers, and they all shook their heads at the value of statistical data for evaluation. In fact, Mayo said, “I can look up minor league stats on my smartphone here if I wanted. But they’re not the complete picture… I’m a lot more excited about a 19-year-old or 20-year-old kid performing well in High-A than I am about a 22-year-old kid who’s tearing up the Midwest League or the South Atlantic League when they’re three years older than the optimal age for that league.” Pleskoff cited his skills as a scout and suggested that “as a scout, I’m taught not to look at statistics at all, but mechanics. Not only with my eyes but with my ears. So it’s not all stats, it’s how he goes about his business.” The Mayo/Pleskoff model falls into the same pit that the Baseball America model falls into: subjectivity interrupts player evaluation. It is near impossible to quantify the difference in how a player “sounds” in a way that would make sense to another person.
In comparison, at almost the same time that Baseball America released their top 100 prospects list and Mayo and Pleskoff were speaking at the SABR conference, Carson Cistulli at Fangraphs released an alternate way of measuring minor league batters: attempting to recreate WAR. Cistulli’s method successfully removes the black-box-ness of using prospect rankings and allows for a true apples-to-apples comparison of hitting prospects.
In a very simplistic format, WAR = BR + RR + DR, or WAR is equal to batted runs plus baserunning runs plus defensive runs. Cistulli proposed the following adjustments to utilize available data for minor league hitters:
|Major Leagues||Minor Leagues|
|Baserunning Runs||(1.6489*SpdScore) – 6.4862|
Pitching WAR is much harder to calculate for minor leaguers, given how inconsistent the defense is behind them and the general lack of minor league pitching data. As Wayne Winston pointed out in Mathletics when talking about the problems with using win-loss record and ERA to judge a pitcher, “A pitcher with good fielders behind him will clearly give up fewer earned runs than a pitcher with leaky defense.” Saves are similarly problematic, if not more so, given the very small sample size (closers pitch exponentially fewer innings than starters). Winston proposes using WINDIFF, and indeed both his DICE and WINDIFF metrics would be incredibly useful for evaluating minor league pitchers, but the data needed to generate those stats was not readily available for this study in Baseball-Reference or in Fangraphs. Similarly, Winston’s DICE metric and ERA equation are missing two key factors we thought were important:
- There is no analysis of minor league vs. major league talent, nor any proposed way to handle players who pitched at multiple levels (do home runs in AAA count the same as A? Would this be used to predict an ERA at AA or at AAA or MLB? Etc.)
- There is no way to compare position players and pitchers using DICE/ERA. WINDIFF could be useful here, but that requires a level of play-by-play information that was not readily available in Fangraphs or baseball reference. A professional team might have their scouts input every play into a database and then perform WINDIFF, but for our purposes that granularity of data was considered out of reach.
Therefore, we needed a different solution. We needed to be able to generate minor league WAR for pitchers, much like we did for batters. In order to do this, we took every pitcher who pitched in 2019 and performed linear regression based off of FIP, IP, GS, and G. Our R regression captured 59% of the variance between these four explanatory variables and a pitcher’s WAR:
Clearly our WAR model strongly values innings pitched and FIP (FIP is a metric like ERA where lower is better), and attempts to focus only on things that a pitcher can control: walks, strikeouts, home runs, and contact.
We then utilized both of these equations on all of the minor leaguers who were involved in any of the trades, and the WAR distribution is as follows:
This graph is relatively normally distributed with a mean of 1.79 WAR. This bodes well when we compare it against a WAR histogram of batters with over 100 AB in 2019 and pitchers who pitched over 25 innings:
In both cases, there is relatively normal distribution with a right skew to cover the superstar pitchers/hitters (those above 7 WAR).
A final note on the methodology for Question 2: some players played in multiple minor league levels during a season. In order to fairly way all metrics across all levels, we performed a weighted average of all metrics by PA for batters and IP for pitchers.
Question 3: What is an expected return?
In order to solve this, we took the 28 trades we curated and performed linear regression on them to generate both # of returning players and total expected returning WAR. We dummy coded each position to see if positions tended to get more or fewer players simply based off of position. In this case, when we forced our linear regression intercept through 0, the R-squared jumped to 0.96. The results were:
|Position||+/- Returning Players|
|WAR of Previous Season||0.19|
|Additional Player WAR||-0.69|
While we’ll get into test cases in a moment, for clarification here is how this formula would apply. Let’s say we were trying to project the return for Johnny Cueto of the SF Giants (assuming no other player involved):
P = 2.34 + 0.19*(WAR of Previous Season) + 0.34*(Avg WAR) – 0.03*(Cumulative WAR) – 0.69*(APWAR)
P=2.34 + 0.19(0) + 0.34(2.6) – 0.03(34.3) – 0.69(0)
Therefore, we would expect Cueto to garner a return of ~2 returning players.
We then performed the same process for expected total WAR return. Using the same table:
|Position||+/- Returning WAR|
|WAR of Previous Season||0.69|
|Additional Player WAR||-1.29|
If we fit Cueto as an example again, we would expect the return WAR package to be:
P = 1.96 + 0.69*(WAR of Previous Season) + 1.09*(Avg WAR) – 0.11*(Cumulative WAR) – 0.69*(APWAR)
P=1.96 + 0.69(0) + 1.09(2.6) – 0.11(34.3) – 0.69(0)
Therefore, we would expect the total WAR return to be ~1.
To put these pieces together, we expect that if the Giants trade Cueto, they should expect to receive 2 minor leaguers whose WAR totals to 1. Interestingly, this is about 75% less than the actual 4.6 WAR return that Cincinnati received for Cueto from the Royals in 2015:
We would expect that if the Royals of 2015 were to trade for 2020 Cueto, they would only need to part with Finnegan and Reed.
To test this, we used 10 trades from our 78 trade complete set and compared actual results vs. our predicted results:
|Player||Predicted # of Players||Predicted Return WAR||Actual # of Players||Actual Return WAR|
|Yoenis Cespedes (to DET)||2||6.1||1||3.8|
|David Price (to TOR)||4||6.0||3||3.98|
|Craig Kimbrel (to BOS)||4||5.4||4||9.2|
|Aroldis Chapman (to NYY)||4||6.2||4||7.06|
|Justin Upton (to SD)||1||2.4||4||7.53|
As expected by the regression analysis R-squared values, our # of players returning did quite well: it correctly pegged 6 of the 10 trades spot on, and it was within one player of the rest (with the exception that the Justin Upton trade was an extreme overpay).
The return WAR part is less accurate (again, in line with our R-squared). In several cases we have extreme over/underpays: Detroit clearly underpaid on Cespedes and San Diego clearly overpaid on Upton.
There are many reasons a team may overpay: an injury to a star player where there is little minor league depth at the position, for example, may cause a playoff-bound team to overly value immediate security over their long term future. This current model does not — and indeed cannot — factor in that level of subjective evaluation. This model is, instead, meant to be an unbiased yardstick for a team to determine the value of assets. If a rival team offers to overpay for your asset… well… never interrupt your enemy when they’re making a mistake.
Before moving on to hypothetical trades at the 2020 trade deadline, let us examine the unique nature of this year’s deadline. The 2020 trade deadline was thoroughly different in comparison to any other year’s deadline, given the peculiar nature of this 60 game season. In an average season, the trade deadline is the final day of July and the season lasts until the end of September – a player traded at the deadline will be with his new team for 2/7ths of the season. It is unclear if players on contracts that expire at the end of the year will be valued more (they will now play in 50% of their new team’s games vs. the 29% of a normal deadline acquisition) or if they will be valued less (they will now play in 30 games vs. 60).
If we look at the recent trade of Brandon Workman, we can predict that the latter will be true. On August 21st, 2020, Brandon Workman and Heath Hembree were traded to the Phillies for Nick Pivetta and Connor Seabold. If we fit this information to our model:
|Player||Additional player WAR (Hembree)||Expected Returning Player #||Expected Return WAR||Actual Returning Player #||Actual Returning WAR|
A sample size of one is no sample size at all, but closers have typically been overpaid for in previous samples we looked at. It is possible that pitchers will see a dramatically deflated evaluation and that positions players will not — given the nature of baseball, in a 30 day stretch there will be fewer opportunities for a closer (or a starter) to pitch than in a 60 day stretch.
However, we must make a stand for our hypothetical trades. Therefore, we will expect a 50% returning WAR package from what was predicted (which is exactly what we see in the Workman trade) and predict that teams are valuing days on the club vs. a normal season rather than percent of total days on the club (50% of the season vs. 28% of the season).
Question 4: What are some hypothetical trades for 2020?*
*Please note that since the deadline has already passed, these are just that – hypotheticals for the purpose of this study.
For this, we took the following sample of players as potentially on the block:
|Player||Expected Player #||Expected Return WAR||Expected 2020 Return WAR|
|Jackie Bradley Jr.||2||3.9||2.0|
Clearly, 2020 is a sub-optimal time to trade stars. Therefore, if we were advising Mike Chernoff, we would advise him not to trade Lindor (our initial question), if this 50% of expected WAR continues to hold true with other trades leading up to the 2020 deadline.
However, players who were projected to return 4 or less WAR in return should still be on the block as the drop off is less steep. For example, the drop from 0.9 to 0.5 WAR for Cueto is much smaller than the 7.9 to 3.9 drop for someone like Kris Bryant.
Therefore, let us fit 3 possible trade scenarios (scenarios created prior to 2020 MLB Trade Deadline):
Trade #1: Johnny Cueto to the Houston Astros
|Expected Return # of Players||2|
|Expected Return WAR||0.9|
|Expected 2020 Return WAR||0.5|
|Proposed Return Players||1|
|Proposed Return WAR||1.85|
Houston’s starting pitching took a major hit when Justin Verlander went down with an injury. If we look at all pitchers who started for Houston this year, only 3 have a WAR above 0.1, whereas Cueto has 0.5 so far this year. The most recent rotation plug, Brandon Bielak, has a FIP of 6.82 over 22 innings, whereas Cueto has a FIP of 4.03 over 31 innings.
San Francisco, on the other hand, has a need for high level minor league SP depth. Our proposal for a first asking point would be AA SP Chad Donato. Donato is not on any of Houston’s top prospect lists and is Rule 5 eligible this offseason, which means that Houston may be more willing to trade him than in other circumstances.
For San Francisco’s side, Donato has a minor league WAR of 1.85, which would represent an overpay, meaning it would be a high positive for San Francisco as well. Finally, by trading Cueto, the Giants will have an open 40-man roster spot and will likely be able to protect Donato in the 2021 Rule 5 draft.
Trade #2: J.D. Martinez to the San Diego Padres
|Expected Return # of Players||2|
|Expected Return WAR||3.7|
|Expected 2020 Return WAR||1.9|
|Proposed Return # of Players||3|
|Proposed Return WAR||2.97|
San Diego currently has two 1B for over $20MM on their roster through 2022 (Wil Myers and Eric Hosmer). This deal would help San Diego resolve that long term logjam. In the immediate future, Martinez’s addition would solidify their DH position: all of the players who have rotated through the Padres’s DH slot have barely cracked 0.0 WAR this year. While Martinez’s current WAR is low, this is likely due to a small sample size as his walk to strikeout ratio is in line with career norms, but his BABIP (batting average on balls in play) is more than 50 points below average (this is a metric that helps calculate luck and he is due for regression to the mean).
Boston has a need for starting pitching depth and lower level OF depth, and have the financial flexibility to take on Myers’s contract which has an AAV of $13.8MM. Therefore, the Red Sox should ask for MLB IF Wil Myers, SP Cal Quantrill, and A+ OF Jeisson Rosario. This combined WAR package is 2.97, which is above the expected 2020 return because Boston would be taking on the extra contract of Myers as part of the deal.
Trade #3: Mike Minor to the NY Mets
|Expected Return # of Players||3|
|Expected Return WAR||5.2|
|Expected 2020 Return WAR||2.6|
|Proposed Return # of Players||3|
|Proposed Return WAR||7.09|
New York should trade for Minor to join their rotation as no SP in their rotation, besides, of course, Jacob DeGrom, has a positive WAR. On top of that, the underlying fundamentals are not kind to the rest of the Mets rotation: the Mets are 7th in baseball in terms of FIP (4.12), but Robert Gsellman’s FIP is over 6 and Steven Matz’s over 7. Minor’s FIP is currently 4.4, which is slightly above his average — since his counting stats like ERA are higher than average, this implies that he’s due for regression to the mean in the 2nd half of this season.
Because of that, the projected first ask is A 3B Mark Vientos, AAA SP Franklyn Kilome, and AA 1B Jeremy Vasquez. It is possible that the Mets will balk at including Kilome, but the prize for Texas here is Vasquez (3.3 of the 7.1 WAR package): Texas has weakness at the 1B level in AAA and AA, the NY Mets have two capable 1B already on their roster (Pete Alonso and Dom Smith), which will make Vasquez expendable, and he does not appear on MLB’s top Mets prospects list. Therefore, if the Mets say “Either Vientos or Kilome plus Vasquez”, Texas can already consider this a successful deal.
There are many parts of an MLB trade that are difficult to quantify – including general desperation – which made this year’s trade deadline is going to be more confusing than most.
Beyond the uniqueness of this year, a major league team that is interested in implementing a process like this for their team would likely want to devote extra resources towards developing a better way to calculate WAR for pitchers, especially when those pitchers compete on multiple minor league levels. For our purposes, we did weighted averages based on innings pitched; in the introduction, we posed questions to Winston’s DICE model: “Does a home run in AA count the same as a home run in AAA? And are these projections for one level? Multiple levels? All levels?” Yet we ourselves struggle with those answers.
Another next step for this project would be to run further regression between proposed minor league WAR evaluations and major league performance. Does WAR at the A level better correlate to major league success than WAR at AA? Are they all equally valid or flawed?
One additional consideration would be contract length. Our methodology did not include this, as we wanted to focus on 10 years of superstar trades, and remaining contract length was not as widely discussed in 2010 regarding trades as it is now.
Ultimately, we believe that our tool is not meant to be a set-in-stone yardstick, but instead a first step towards the proper evaluation of minor league players when considering trades. By utilizing this method, a major league team should be able to:
- Predict the inherent value of their star player in terms of prospects.
- Objectively evaluate the value of minor leaguers presented as trade chips.
- Discover undervalued assets in opposing teams’ minor league systems.
- Ensure they do not overpay when trading for another team’s star player.
by Justin Halpern and John Smith, Northwestern University