A few days ago, I heard a story on NPR.com that offered a great tip about making predictions. I strongly recommend that you listen to the story, "Want to Make Better Predictions? Researchers Explore How." In under three minutes, it provides valuable insights into predictions and acting on them.
Shankar Vedantam, host of the Hidden Brain podcast, pointed out an interesting element of our irrationality. When we try to predict things, we quickly reach a point where additional analysis becomes counterproductive. "We break predictions into their elements .... People are not very good at weighing different details in a prediction."
Once we get below the surface level, there can be a lot of moving parts affecting the prediction. We generally think breaking a decision into component parts improves our process, but it can backfire because we aren't good at intuitively understanding the relationship between the parts.
For instance, for an upcoming football game, we may think we're been thorough in picking the winner because we've analyzed two decisive advantages that team has: their defense produces lots of turnovers and their offense is high scoring. Because we skip over the possible relationship between those two things - defensive turnovers gives the offense more scoring opportunities - we could be double-counting the same advantage, or over-weighing its influence.
This problem doesn't apply solely to picking winners in sporting events. When you think of it, many of our decisions depend on predictions: at work, we have to predict how long it will take to finish a task in allocating our time; at home, we have to predict the reaction of family members in deciding whether to overlook or confront a particular situation; in our financial lives, buying a home depends on predicting the overall costs, our other expenses, and our earning ability.
How can we improve our predictions? The long answer, of course, would involve having or being an expert predictor. The reason it's possible to beat prediction markets (like the stock market or sports betting) is that the best predictors understand the interrelationships among component parts of predictions.
But for those of us without access to precise regression analyses, it might makes more sense to focus our prediction on one thing we do know: how things have turned out in the past. In the NPR story, Theresa Kelly, who studies consumer behavior at Washington University in St. Louis, explained how she applies to estimating time to complete an assignment: "Rather than try to think about all the circumstances surrounding this present attempt to complete the assignment, I'll just say, 'Well how long has it taken me to complete similar assignments in the past?' In general, that's a better way to make predictions because there's all kinds of reasons why thinking through the details of the case at hand can lead you astray."
Simplifying your decision process in that way can makes sense, especially when the component parts might be difficult to tease apart. Look at what experience tells us. And, in the future, keep track of the results of experience.