Betting on Horses

Imagine the total processing capacity of your brain is roughly equivalent to the size of the US economy, which today is around $27 trillion. What would you say is the total processing power, in dollar terms, of your working memory? Apparently, it's about 3 dollars.

This is why, when making decisions in complex domains, we usually feel more information is better. We yearn to compensate for our natural limits.

But there is clearly a point of diminishing returns, though it tends to be invisible to us because of our deeply-rooted susceptibility to analysis paralysis.

In 1973, University of Chicago professor Paul Slovic set out to quantify those limits, with the help of eight expert horse-racing gamblers.

Slovic took the statistics for 40 races over the preceding years, then removed the dates and names of the horses involved. All you saw were the raw numbers.

His protocols for the experiment were simple. The experts would be asked to predict 40 races over 4 rounds, or 10 races per round. In each round, they would be given no more than a fixed number of prediction variables out of a potential 88. The gamblers could ask for the weight of the jockey, the average speed over the last five races, and so on. It was up to them to choose these variables.

For round 1, they were given 5 pieces of information. For the second, they were allowed up to 10. Then it was 20 in the third round and 40 in the last. The gamblers would be evaluated on their accuracy, and on their confidence in each prediction.

In the first round, they did much better than a 1/10 random chance of being correct. The experts were 17% accurate with only five pieces of information. They were also 19% confident in their predictions.

By the fourth round, their accuracy remained stable at 17%, but their confidence had almost doubled to 31%. There was some variation in individual performance across the group of gamblers, but all of them were about twice as confident, based on having eight times as much information.

In other words, they were overconfident. All the extra information did was feed their confirmation bias. They had already come to a decision based on the five essential variables, and the additional analysis was simply re-affirming their initial judgment.

Whether you're betting on growth accounts or job candidates or investments, it might be worth setting some limits. What are the 'Big Five' criteria?

It could save you time and potentially the unseemly spectacle of your unwarranted chest-thumping.