Halfway into writing last week's post about Ricky Rubio, I thought I could make a crude but entertaining ranking of NBA players based on how high the Rubio article comes among the Google search results for '(player last name) + worst shooter ever'. (1)
Here were my results:
Kendrick Perkins: 5
Rajon Rondo: 7
Zach Randolph: 2
Ron Artest/Metta World Peace: 6 (both times, which was impressive)
So far, so good. But what if the small sample size was misleading me? I thought about how Google searches worked and tested some counter-examples.
Kevin Garnett: 1
Steph Curry: 6
Klay Thompson: 2
Reggie Miller: 1
Ray Allen: 6
My hypothesis was falling apart. I guess there was only one thing left to check.
Ricky Rubio: 1
Well, at least the baseline case went smoothly. But despite my minor success, I decided to scrap the concept.
This exercise is probably a decent enough introduction to show how Google works. Intuition suggests these results all differ by the one term I changed with each search (the player's last name). But focusing on what is changing is a common reasoning error. Google does not consider changes to a user's most recent search. Rather, it just looks at each search on its own terms (!) and brings up the best links.
Kevin Garnett's 'Rubio Score' being one is unrelated to his ability. His score is one because he is shown in the article strangling/caressing Rubio's neck. (What a great guy!)
This process reminded me of Michael Lewis's discussion of confirmation bias in The Undoing Project. Confirmation bias means shaping evidence to support a conclusion. It starts by forming an opinion or conclusion prior to gathering evidence. Once evidence collection begins, a biased judge will fit the evidence to support the conclusion. On the other hand, an unbiased judge would use the evidence to lead to a conclusion.
In the case of The Six Degrees of Rubio, it would have been very easy for me to run several hundred Google searches using bad shooters. I could have collected as many examples of a 'one' score as I wanted and presented my findings to my reader(s). If I made a pie chart or a PowerPoint slide showing these results, the findings would appear even more impressive (and thus would have to be True).
But they would have proved nothing. The way to disprove my idea was in the second example. By running examples using great shooters, I showed how the name mattered little in the search and disproved my hypothesis.
The exercise makes me wonder about the best approach to battle confirmation bias. The usual presentation of the phenomenon makes forming an initial conclusion appear like the main issue. With a conclusion in hand, the bias is described as almost inevitable.
I'm sure remaining entirely judgment free in this first stage is one way to avoid falling victim to the bias. But this strikes me as a difficult task. I almost always form a first impression when meeting someone new and I tend to rely on gut instinct to gauge if a given statement 'seems' true when I first hear it.
A more sensible approach considering my natural and nearly unavoidable tendency to have a gut reaction might be to seek out evidence unquestionably disproving the initial conclusion. It is not a big problem if I assume someone is my height because our eye levels are the same when seated. The issue is when I still think so after we both stand up and I realize the top of the other person's head lines up with my rib cage. The key is to find ways to stand up, so to speak, and look at the evidence from many angles.
In the case of my Rubio Index, I tried using some great shooters as test cases. Kevin Durant was a promising first but he scored a 10+ (he's a reliable double-digit scorer, I suppose). But this trend did not hold. When Reggie Miller scored a one on my 'Rubio Index', my hypothesis was in trouble.
I was at a crossroads, I suppose. I could continue tweaking my 'model' and find better search terms to use. Or I could move on to something more productive (though what could be more productive than this...).
I opted for the latter and finished reading The Undoing Project. I'll write a little more about this book soon.
Until then, thanks for reading.
Tim
Head Writer, TOA
Boston, MA
Rubio Score: 10+
Footnotes / imagined complaints
1. Like I said, I was drunk and it was snowing and every flake is unique...
The concept vaguely resembles the 'six degrees of separation' idea, I guess, but I really have no idea how all this got started. Is the question of how close each NBA player comes to showing up alongside the Rubio article an important one? I'm sure I thought this exercise would tell me something at the time.
