It is a longstanding cliche in sports that with a new season comes a clean slate; a chance for fans to forget about the seasons previous and convince themselves that this is the year, without any rebuttal from the league standings. Over the years as analysis has become a larger part of baseball and a larger part of the media covering baseball, this clean slate has become increasingly less clean. There are webpages and articles devoted to the myriad of projection systems publicly available as the season nears and, while they are certainly not infallible, they are the best predictor of team performance out there, and even do a better job predicting rest-of-season records a few months into the season. Before the explosion of modern baseball analytics and its presence in online publications, during this most recent offseason, a fan of the Giants could not so unreasonably wish for his team to bring in Bryce Harper and make another World Series push. Nowadays the aforementioned projection systems (which weren’t so high on the Giants postseason chances) lie right at the fan’s fingertips. On top of this, most teams now realize that there is little difference between finishing with 70 or 75 wins; both totals result in missing the playoffs and the latter results in drafting later in the following season’s amateur draft.
The projections are by no means perfect, but with the deadline approaching, a natural question which came to my mind was if the projections could be improved by attempting to account for the change in rosters throughout the season. While most projection models account for players in the high minors getting a few at-bats throughout the season, they could never make the assumption that the 2018 Dodgers were going to get 2.8 WAR from Manny Machado, who belonged to the Baltimore Orioles at the start of the season. If, however, projections could predict which teams were more likely to add value in transactions throughout the season, they could provide a small boost to such teams. As is customary for this time of year, we are currently discussing which teams are buyers, teams often willing to trade future talent for present talent in order to make a playoff push, and which are sellers, those teams out of contention for 2018 looking to eschew present value, especially value in the form of players in the last year of their contract, in exchange for prospects. As such, if the projections are doing their job, we should see teams that project to be in the playoff race adding talent throughout the season, with the other teams losing present talent.
So what does the data say? I went back through MLB’s transaction log and noted the WAR accrued from players with their new teams after a transaction. For example, after Machado’s deal to LA, he was worth 2.8 WAR and as such the Dodgers gained 2.8 from the transaction and the Orioles lost 2.8 (ignoring all other pieces in the deal in this illustration). This method is not perfect as it assumes every player moved would have seen equal playing time if the team that traded them elected to keep them around, but it will suffice. Below is a plot in which all 30 teams from the past three seasons are represented by a dot, with the x-axis denoting PECOTA’s preseason projected win totals and the y-axis denoting aggregate WAR gained through mid-season transactions.
Altogether, this plot seems to have a slightly positive trendline, but the data is so loosely clustered around that line that it would be irresponsible to draw any immediate conclusions. This data becomes more interesting however, when we split teams into three subsets. We will define the good teams to be those teams projected to win more than 85 games, the bad teams to be those projected to win at most 75 games, and our third bunch will be those teams projected between 75 and 85 wins. For the good teams we have the following.
We see almost all teams in this chart gained value over the course of the season, with most teams gaining between 0-2 WAR. Within this group of teams it seems there is little correlation between the amount of WAR added and wins, meaning a team projected to win 97 games does not typically add much more talent than a team projected to win 91. This makes sense and in fact one might expect the opposite effect (worse teams in the playoff push might feel the need to acquire more talent), but the data does not support this either. Instead we see that most teams projected to have a reasonable chance at the postseason seem to add on average 1 win. For bad teams the story is, again as you might expect.
Again, most teams lost value over the course of the season, and while a few teams added value, that primarily came from transactions involving players who posted negative WAR totals with their new team. Teams in the thick of a playoff race have to obtain value from somewhere, and generally you can expect teams projected to lose a lot of games to be eventual sellers, as teams in this set averaged a loss of roughly 1 WAR. For teams projected in the middle of the pack, however, the story is different.
Here the data suggests a team in this range seems just as likely to be buyers at the deadline as they are sellers. This leads to a necessary discussion on the ways in which random variance will always get in the way of trying to account for mid-season trades. During the 2018 preseason the Mariners, Blue Jays, and Cardinals projected to have similar seasons (projecting for 82, 85 and 80 wins respectively) and this is how things stood the morning of July 31, 2018.
The Mariners were right in the middle of a playoff push, grossly outperforming their pythagorean record (a measure of how many games a team “should have” won based on their run differential) with the other two buried beneath a slew of other teams in their respective leagues. Seattle over-performed their projections in 2018 and saw a chance at breaking the longest active playoff drought in the four major sports and made a push to acquire talent. While projections do a reasonable job predicting end of season win totals for teams, they will never be able to account for the high variability of half-season records, and teams like the 2018 Mariners will almost always be buyers, despite projecting to be an average team.
There is one more way variance throws a wrench in trying to project WAR obtained through mid-season transactions. Just as teams can over or under-perform their projections in half of a season, so too can players. For a recent example, look no further Jonathan Schoop’s 2018. In 2017, Schoop put up a 5.2 WAR season thanks to a line of .293/.338/.503 (and perhaps in part because of a career high .330 BABIP). Before his trade to Milwaukee in 2018, Schoop had posted a modest 1.3 WAR, making him an interesting target for teams looking to add offense. With Milwaukee he hit .202/.246/.331 and was worth a paltry 0.1 WAR, a far-cry from what the team hoped to get from their most notable deadline deal. Thus even if projections could improve their accuracy in predicting who the likely buyers and sellers are, a team that moves all in at the deadline might get very little from the acquisitions as a result of the limited number of plate appearances the acquired players will take for their new team after the trade.
If there is anything to take away from the 2016-18 data, it could be that generally the best teams add talent which they can usually depend on acquiring from the worst teams in the league. But seeing as the top teams might also lose more value when their starting players miss time due to injury, the value added in mid-season deals could very easily be negligent. Moreover, the projections are not reliable enough to predict which teams among those projected to win roughly half of their games will be looking to acquire talent and which teams will be selling talent. So while the projections are certainly not perfect, it seems a minuscule amount of accuracy is to be gained by attempting to account for players moving at the deadline.
Featured Image: Keith Allison, Flickr
