One of the things I find both interesting and funny about baseball is the extent to which interpersonal relationships seem to dictate personnel decisions. Examples abound, ranging from what agent Joshua Kusnick writes about in his columns at Baseball Prospectus to teams avoiding Scott Boras’s clients to reporters discussing how Billy Beane has good relationships with various GMs around baseball to the idea that Walt Jocketty still resents Jeff Luhnow enough that he wouldn’t trade with him. It makes this industry with billions of dollars in revenue each year seem rather more like high school than one might hope, and it complicates some of the ways people write about front-office moves: as though each GM has the same motivations, and those motivations can be modeled in a simple, straightforward manner.
The articles that do touch on the interpersonal dynamics are frequently written from a very narrow perspective—trying to explain one draft pick, grade one free agent signing, or speculate about one possible move before or after a trade deadline. In this piece, I’ll look a bit more broadly at the aggregate impact of one aspect of all of this: who trades with whom? How much truth is there to the idea that teams don’t trade players to divisional or geographic rivals? Do teams that got robbed in a trade refuse to trade with the team that robbed them later? This analysis can necessarily cover only the tip of the iceberg, given our absence of knowledge about actual communications between the teams (barring some more leaks in the vein of the Astros’ Ground Control records), but that doesn’t remove all validity from this exercise.
To start, let’s look at a diagram that overlays each of the major league trading relationships, using Retrosheet transactions data and including only trades that ocurred between the expansion drafts in November 1997 and realignment at the end of the 2012 season. Each pair of teams is connected by an arc that is colored orange if the teams are in the same division, blue if they are in different divisions within the same league, and pink if they are in different leagues. The thickness of the arc corresponds to the number of trades between the two franchises, and the width of a team’s wedge reflects the number of trades it made during that time period.
Unfortunately, while that chart has a certain artistic appeal to it, it’s too general to allow for any conclusions. Two things do stand out, though. The first is that there’s a real paucity of orange, meaning teams are less prone to trading within their division, and the second is that there’s a pretty substantial variation in the amount of trading done (the width of the sectors), ranging from San Diego’s 137 trades to Anaheim’s 48. The chart below sheds a bit more light on the disparities between intradivisional trading and interdivisional trading. It shows the distribution of the number of trades over all pairs of teams, grouped by the type of pairing, with the vertical bars marking the average number of trades by group.
At first glance, there is a pretty pronounced difference between intradivision trades and the other two categories, but intraleague and interleague trades seem marginally but not substantially different. (For the technically inclined, we can model this with a Poisson regression model, which yields a significant difference (p < 0.05) in means between the intradivision pairings and each of the other two; the difference of means of the interleague and same league, different division pairings is not statistically different.)
However, when one considers that this chart includes 15 years of data, the obvious conclusion is that the difference just isn’t large enough to be important—a typical pair of teams in the same division has traded two or three times over that time period, and a typical pair of teams not in the same division has traded three or four times in that period. Even in aggregate, it’s hard to care about just a handful of trades.
In a certain sense, the fact that there’s a difference at all is a bit surprising. The reason to avoid intradivision trades is that they can come back to haunt you. As White Sox GM Rick Hahn said earlier this year in reference to the White Sox’s participation in the trade that sent Jose Iglesias to Detroit, “We did a three-way that helped Detroit, in our division, and they’re still getting the benefit of that deal with (Jose) Iglesias playing short for them every day… [We’ve been] willing to help the teams that we compete with 18 times a year and we compete for a division title against.” But this doesn’t stand up to much scrutiny. The impact of any trade is zero-sum, so if one team gets totally fleeced then the other team will get the benefit of the fleecing. Thus, unless a team thinks it is likely to lose the trade—in which case, why trade?—it shouldn’t expect to be worse off making an intradivision trade than any other trade.
A different, simple explanation is that GMs are risk-averse and would rather make a slightly worse trade overall in exchange for a better outcome in the worst case scenario. Even this doesn’t make much sense, though. Russell Carleton wrote an article this summer at Baseball Prospectus in which he estimated the negative impact of losing one win’s worth of talent trading within the division to be about 0.1 percent of playoff probability—basically negligible. If that estimate is at all accurate—and granting that for the trades that lead to a multi-win impact it likely won’t be—then there’s not really measurable on-field risk for GMs to be averse to. I won’t go so far as to say it’s irrational for GMs to be averse to trading within the division, given that there are issues of fan and owner perception that may affect both revenue and the GM’s likelihood of keeping his job, but the behavior doesn’t seem to be empirically well-justified.
Given the apparent lack of justification, it makes sense to ask: Are there teams that buck the trend? To answer this, let’s first look at a chart of intradivision trade percentage by team, similar to the version shown here:
This fails to account for something important, though: AL West teams had only three other teams in their division to trade with during this period, while NL Central teams had five other teams in their division to trade with. We can see that play out above — the Brewers are the runaway leader, and the Pirates are second, while the Angels are second-to-last. To simply adjust for that discrepancy, the next chart shows each team’s percentage after subtracting (other teams in division) / 29, the percentage that we would expect if there were no difference between intradivision trading and other trading:
That’s better. You can see Milwaukee, Pittsburgh and Anaheim congeal into the pack. It’s hard to see a pattern there, though there is a clustering of teams whose front offices have not-great reputations toward the top: Miami, Arizona, the Mets, the pre-2013 Pirates. The real question is one analysts love to ask: How much of the discrepancies are just random variation?
To test this, I measured random variation by assuming that (other teams in division) / 29 – intradivision trade percentage was the same for each team, using the average value of 5.2 percent. I then simulated a set of trades for each team using that assumption, calculating the simulated intradivision trade percentage for each and computing the standard deviation of those 30 percentages. After doing 10,000 simulations, we obtain an estimate of how much variation to expect in the percentages if all the teams were equal.
As it turns out, the variability observed (a standard deviation of 3.4 percentage points) is about exactly what we would expect if all the teams had an equal propensity for trading within the division (3.5 percentage points). Accordingly, it doesn’t seem as though there’s strong reason to believe that there’s real, non-random variation between teams in their trading inclinations; even if there is, it’s dwarfed by randomness in this small a sample.
What about rivalries? People have posited that teams are less likely to trade with their interleague geographical rivals—the Yankees and Mets, Cubs and White Sox, Cardinals and Royals, etc. One reason for this might be a fear of embarrassment or losing fans to the rival; another, more practical reason that Jed Hoyer suggests is that the overlap in the media covering the two teams makes it harder to execute the trade. There were 11 interleague “natural rivalries” that were active during the entire 1998–2012 time period covered, and those pairs averaged 2.4 trades, as compared to 3.5 for other interleague pairings and 2.1 for intradivision pairings; the average drops to 1.8 when excluding the least natural pairing of the natural rivals, Seattle and San Diego. (For the technically inclined, this difference is statistically significant (p = 0.044) when tested using the same Poisson regression framework mentioned earlier.) Again, it’s hard to argue that this difference is of much practical importance, as it’s only one or two trades over a period of 15 years, but it’s still interesting that the effect exists.
Finally, what about trades between two teams where one burned the other in the (recent) past? For instance, after the Scott Kazmir–Victor Zambrano trade, did the Mets refuse to trade with the Rays? Given how vague this proposed phenomenon is—How long after a trade occurs does the losing team start to regret it? Does the enmity expire after one of the teams changes GMs? How lopsided does a trade have to be for one team to react?—a coarse analysis of trade counts is a weak tool to shed light on this, but it’s still worth seeing if there’s an obvious trend.
Using a set of 13 articles listing particularly lopsided trades obtained from Google searches, I found a sample of 34 uneven trades—the ones mentioned in at least two of the articles, under the theory that the wisdom of crowds will outweigh the stupidity of listicles. I further restricted the 34 to the 25 occurring during or after 1996, a natural breaking point in the data. (The list of articles and trades can be found in this spreadsheet.)
These 25 trades involved 23 pairs of teams, and over the 1998–2012 period those pairs traded an average of 4.2 times. This is smaller than the 4.9 we would expect if we chose a random trade, but well within the expected amount of variability. (For the technically inclined, I used resampling to estimate variability and found that about 16 percent of 25-trade samples diverged at least as much from average as the lopsided trade sample did.) Thus, repeating the caveat that this is an imprecise way of measuring any such effect, it doesn’t appear as though having lost a lopsided trade makes the loser less likely to trade with the winner in the future. Score one for the theoretically rational GMs.
So, what can be gleaned from all of this? It’s a bit too glib to call the patterns of refusing to trade with divisional or geographic rivals irrational, given the complex incentives facing GMs and franchises, but it is clear that there are such patterns. If they are visible when looking at completed trades, it’s likely that there are others that would be observed with more complete information about front office interactions.
While there are other interesting ways to look at interpersonal ties and how they affect transactions—analyzing free agent rumors or interviewing baseball insiders, to name just two—it seems vanishingly unlikely that we’ll ever get the sort of information that would allow for analysts to easily explain or predict transactions in a rigorous manner. If anything, this whole exercise highlights that no matter how much easier it may get to measure and predict what happens on the field, there are always going to be important factors fans are left guessing at.
References & Resources
- Retrosheet transactions data
- Baseball-Reference Trade Partner Tool
- Joshua Kusnick, Baseball Prospectus, “The Agent’s Take” articles
- AJ Cassavell, Sports On Earth, “The Scott Boras Effect in D.C.”
- James Wagner, The Washington Post, “Nationals-Athletics trade history speaks to comfort between GMs Mike Rizzo, Billy Beane”
- Ken Rosenthal, FOX Sports, “No surprise Luhnow’s unique brilliance drives rivals nuts”
- Russell Carleton, Baseball Prospectus, “Baseball Therapy: So You’ve Decided To Trade Within Your Division”
- Julien Assouline, BP Milwaukee, “The Brewers and Trading Within Their Division”
- Dave Cameron, JABO, “Why Don’t Certain Teams Trade With Each Other?”
- Bill Evans, NJ.com, “It’s rivalry week in Major League Baseball. How “natural” is your natural rivalry?”