Nearly two years ago I made a simple observation. Stephen Hill, a rookie receiver recently selected in the 2nd round by the New York Jets, had just had a very good debut game, catching 5 passes for 89 yards and 2 touchdowns. I wanted to put that game into some sort of context. I had access to a game log database dating back to 1995, so I ran a quick query and discovered that in that span only four rookie receivers had topped either 20 fantasy points in standard scoring or 25 fantasy points in PPR scoring in a week 1 debut. The four receivers were Anquan Boldin, Randy Moss, Eddie Royal, and Stephen Hill. (Additionally, Marvin Harrison joined them as the only rookie receiver to rank in the top 5 in week 1 of his rookie season.)
Now, this observation was simple fact. The numbers represent an objective truth which is easily verifiable. Taken at face value, there was nothing at all problematic with it. But today, as Stephen Hill is unemployed after having been cut by the receiver-starved New York Jets just two years into his career, it seems an opportune time to state that while my initial observation was undoubtedly true, it was also wrong.
You see, I did not make that observation and tweet it out as part of some larger objective data dump. That was not a random fact that I happened to stumble across and felt was worthy of sharing. That observation was the end result of a process called motivated reasoning, which is itself one of the primary tools of confirmation bias, which is one of the most pervasive and devastating flaws in our mental software. I had been a proponent of Stephen Hill’s dynasty value since the day he was drafted; he was, in all aspects, one of “my guys”. He came from my favorite part of the NFL draft, (the early 2nd round, which is high enough to signal strong talent but not so high that it carries the “first round” cachet and resulting premium). He was an athletic marvel on a receiver-needy team. He even had a good story, hailing from a Georgia Tech team that had just given us Calvin Johnson and Demaryius Thomas, two other raw-but-athletically-gifted phenoms. The purpose of my observation was not to place Hill’s performance in an objective context, but to paint it, (and, by extension, my support of Hill), in the most favorable light possible. The data may have been correct, but they were being used in an unfair and incomplete manner.
For instance, I just as easily could have noted that the fifth-best rookie debut came from perhaps the greatest draft bust of all time at the receiver position, former Detroit Lions’ receiver Charles Rodgers. A list of comparables that includes Rodgers, Royal, Moss, and Boldin looks markedly different from one that includes just Royal, Moss, and Boldin. The former group implies a 50/50 chance that a receiver will still flame out, whereas the latter suggests a two-in-three chance that the receiver will at least make a very strong argument for the Hall of Fame, (and, in fact, insofar as Eddie Royal can be rationalized away as a lone outlier, the implied odds might seem even more favorable). In reality, no possible list of comparisons would have the slightest impact on how Hill's career turned out, but it's clear we allow associations with greatness, (or associations with true awfulness) to color our perceptions of a player. I could have noted that other players in the top 10 in terms of rookie debuts included Louis Murphy, Deion Branch, Doug Baldwin, and JuJuan Dawson. I could have noted that fellow rookie Alshon Jeffery actually had a fairly strong debut of his own, checking in at 11th on the list. I didn’t note that, of course, because Alshon Jeffery was not one of “my guys” and I therefore did not have a vested interest in interpreting his results in the most favorable light possible.
This is the danger that confirmation bias poses: it causes us to price in the most optimistic data while disregarding the pessimistic data entirely. It causes us to lose the ability to treat a player with balance and fairness. It skews our valuations and diminishes our accuracy. Worst of all, it is a sneaky, pernicious force that creeps in even when we are on guard against it. I did not set out two years ago to make a puffed-up case for Stephen Hill. At the time, the “20 point” cutoff seemed perfectly reasonable and natural. There seemed to me to be a fundamental difference between Hill’s 5/89/2 and Rodgers’ 4/38/2 or Murphy’s 4/87/1, a difference that put Hill’s debut in a completely different category. Hill had the 5th most receiving yards of any receiver in the sample! Hill was one of just four receivers with two touchdowns! These things clearly set him apart from the “good” debuts and raised him into the class of “great” debuts.
With the full benefit of hindsight, the biases in my reasoning seem obvious, but that’s how biases work. Everything feels great until you’re forced to cut that promising 2nd-round receiver you spent so much to acquire because he’s not currently playing in the NFL. The flaws in your reasoning remain hidden until you’re conducting a postmortem on just how exactly everything went so wrong. In fact, this is even a popular method for combating confirmation bias: imagine a future where everything has gone wrong and conduct a fictitious postmortem on the most likely causes.
I’m sure I’ll be writing more about confirmation bias in the future, but for now I simply wanted to call it out as a natural enemy of any good fantasy football owner. Never will it be more powerful than after the first week of the season; when we have so little good information on which to base our decisions, we are left with plenty of room to fill in the gaps with our own imaginings and machinations. Surely some players that you like will have a big game in the coming days, and some players that you have bet against will come up small. Just as surely, some players that you like will come up small, and some players that you bet against will have big games of their own. If our goal is to assemble the best team we can, it is imperative that we treat each new development as even-handedly as we are able.
