May 22, 2014
Representativeness, Valuation, and Tanaka
Back at my old stomping grounds, I used to write about the same topics that I currently write about at Baseball Prospectus, but when I wrote about a behavioral phenomenon that had particularly negative consequences I would call it a fantasy pitfall. The representativeness heuristic can be placed in the fantasy pitfall category. What follows is a discussion of the representativeness heuristic, its impact on fantasy baseball decision making, an example of that said impact regarding Masahiro Tanaka, and a look at how we can hopefully minimize the negative impact of the representativeness heuristic on fantasy baseball decision making.
Note: while this is going to be most helpful for the offseason, I always find it best to investigate and write about such topics when they are fresh in my mind. Because this topic is particularly relevant to players coming over from Japan, Cuba, and the amateur ranks, this article should still be of some help during the season.
1. The Representativeness Heuristic
The representativeness heuristic is the extent to which people judge probabilities “by the degree A is representative of B, that is, by the degree to which A resembles B.” Instead of predicting the likelihood that I am a cat owner based on the base rates of cat ownership in northeast America, cat ownership amongst age 20-30 males, etcetera; people will make their predictions based on how closely I resemble a stereotypical cat owner (whatever that looks like to them). Representativeness can be helpful when caution is beneficial and when being overly cautious has little or no cost. For example, a dog foaming at the mouth is representative of a dog carrying rabies; thus, if we assume a dog foaming at the mouth has rabies and, consequently, act more cautiously in the dog’s presence, then we are less likely to contract rabies. If the dog’s mouth was foaming for some normal reason or if it had really not been foaming at all, then our cautiousness only cost us a potentially nice encounter with a dog. This was probably more explanation than was necessary, but I did so because the oft-beneficial nature of the representativeness heuristic is what makes it very difficult to shake during those times in which it is a hindrance. It is when the representativeness of an object (in our case a baseball player) does not align with its true nature (talent level) that using representativeness to make predictions can easily lead us to incorrect judgments (valuations) about the object.
2. How the representativeness heuristic relates to fantasy baseball
Valuation is the name of the game when it comes to fantasy baseball. Once a player is a known quantity, once we know who he is as a player and the nuances of his game, we project his future performance based on past performance and the knowledge we have built up about him. For unknown quantities to the fantasy baseball community such as Japanese players, Cuban players, and unheralded prospects, we do not have that previously mentioned stockpile of information; thus, our brains are tempted to use representativeness to estimate future production. For Japanese players in particular, because they play the game in such distinct manner, a new player’s future production tends to be estimated based on his likeness to one or several of his Japanese predecessors. The issue here is that the player’s past performance (statistics) and, moreover, skills (scouting) will be a better indicator of predicting the player’s value than will the degree to which he resembles a particular set of his predecessors. Rarely does a player follow the same exact path in production as another player and even if a player does so, the chances that we are going to pick that specific player as a comparison at the time of valuation (either the beginning the of the season, the beginning of a career, or right after the amateur draft) are very slim. Even if we select a more general player type to use for predicting the future value of a player, we will tend to underestimate the most exceptional parts (both good and bad) of the player’s game because we are anchoring our valuation to a representative population. Lastly, the representativeness heuristic is indiscriminate regarding the types of valuation errors it can cause in that it can cause us to either overrate or underrate a player. As per usual, I have blabbed on, so let us take a look at an example to see what this really looks like.
3. Example: Masahiro Tanaka
You probably know how this story goes, but it is a good example of the representative heuristic misleading a valuation; more specifically, my valuation. Masahiro Tanaka is a dominant pitcher from Japan with great control, great command, and filthy stuff. While his dominance at such a young age in the NPB was certainly more “Yu Darvish” than anything else, everything else about Tanaka, from his arsenal to his physique and appearance to the team he signed with, was more “Hirkoi Kuroda” and “Hisashi Iwakuma.” Consequently, I evaluated Tanaka (at least to a significant extent) on the degree to which he resembled Kuroda and Iwakuma. I started with these two pitchers as the foundation for my valuation and worked from there. For my valuation I said that Tanaka is younger, might throw a little bit harder, and has a better splitter. Ultimately, this left me having Tanaka somewhere in the high 20s-low 30s range for starting pitchers. Obviously, this was a mistake and I was probably not the only one who made it. More importantly, this was not good process-bad results; this was undervaluing an exceptionally talented player.
There were indicators pointing to Tanaka’s excellence that I should have noticed, but instead I anchored on his likeness to less dominant pitchers. The point is not that I should have judged Tanaka’s value on his likeness to Darvish rather than Kuroda; the point is that I should have judged Tanaka’s value outside of the representativeness heuristic. Had I stuck to the facts at hand—dominance over the NPB as a 24-year-old, 93-mph fastball, great control and command, best splitter on earth, plus slider, great competitor, workhorse/durable, a lot of professional innings on the arm—I would like to think that I would have ranked him in the high teens. Instead I used a now-38-year-old pitcher for the basis of my Tanaka valuation and it cost me. We know the risk of and our brains’ preference for the representativeness heuristic, but what we need is a way to evade this heuristic when necessary.
4. Method for overcoming the representativeness heuristic
The representativeness heuristic is one of the few heuristics that, when brought to our attention, can be consciously avoided. Going forward, to identify whether or not my valuation of such a player has been affected by the representativeness heuristic I am going to ask the following questions upon completion of my valuation:
- What assumptions did I make?
- What questions did I ask?
- What comparisons did I use?