Chapter 10 of Daniel Kahneman’s book Thinking Fast and Slow covers familiar ground to anyone who follows sports statistics. The chapter addresses two of the major mistakes humans make when dealing with statistical informtion: ignoring small sample sizes, and seeking out patterns in outcomes that are actually random chance.
At one point, he discusses the “hot hand theory”, where it is shown that basketball players do not actually have streaks where they get hot; the streaks we observe are pretty much what we would statistically expect from a random distribution of shots, given a player’s overall shooting percentage.
That, too, is familiar territory to people who follow sports statistics. But recently, I came across something new (to me, anyway) about this topic. What has been missing from the discussion is an explanation about why the outcomes of human performance is distributed randomly.
Suppose you had a very precise pitching machine. It throws the exact same pitch to the exact same location every single time. And yet, a human batter facing that machine and those pitches would not hit the ball with the exact same result each time. Why not?
You might just chalk it up to, “we’re humans, not robots.” But that’s not really an adequate explanation, is it? Which part of human mechanics introduces the randomness in our performance?
A good explanation for the randomness in our performance can be found in this Ted Talk by Daniel Wolpert. At about five minutes in, he explains that the chemical transmission of nerve signals from our brains to our muscles and back is extremely noisy.
When we use a machine metaphor to describe our bodies, we make assumptions about our systems that aren’t accurate. The signals that our bodies send aren’t nearly as clean as the electronic signals our computers send to its peripheral devices.
Think about someone you know who is hard of hearing. Or about having a conversation in a noisy restaurant. What happens in those conversations? Often you can’t hear every word, so you have to make a guess based on context as to what was actually said. You are forced to fill in the gaps with guesses based on past experience. And sometimes those guesses are wrong.
And because the noise disrupts the signals randomly, our guesses are wrong randomly. If we could somehow replace our nerves with copper wires or fiber optic cables, though, our performance failures would be reduced significantly.