2020 Prediction Contest Results
The 2020 prediction contest events have been scored, and the results are in. We had nine participants this year, and the results are as follows:
- In last place with a score of -67, our own guest correspondent @pdxsag. He had put 99% chances for a number of events that didn't happen, and 1% chances for some that did. With our logarithmic scoring system, being off by 99% in either direction subtracts ln(0.01)=-4.61 from your final score.
- @barbariancap with a score of -50; he was second place in 2018
- @unionsquareresearch, -48.6
- @sonimantennis (aka 2019 second place winner Bjdubbs), -42.9
- @TwainsMustache, -42.3
- @inner_scorecard, -39.8
- "Yours truly," @thecreditbubble, -39.6
- Last year's winner in second place this year, @irbezek, -39.3
- Winner: @walt373, -38.6; he was third place in 2019
You can tell which predictions made the contest results by which ones had the highest standard deviation of probability percentage going into the contest: whether TSLA would go below $200 (pre-split), whether the Boeing 737 would return to service (it did), whether 2020 vehicle sales would decline, and so forth.
This may be the last prediction contest, or at least we may take a break from having a 2021 contest.
For the contest itself was really a multi-year experiment in the folly of prediction. The year 2020 illustrated that for us perfectly. No one ever thought to include a pandemic, or lockdowns, or riots and curfews in hundreds of cities, or burned down police stations as potential events.
And what probability would anyone have given, one year ago, to this event: "The S&P 500 will hit an all time high during a week when passenger air travel (as measured by TSA checkpoint throughput) is down 75%"? Or how about this event, what probability: "Elon Musk will become the second richest person in the world, without demonstrating a new battery technology, establishing level 5 autonomous driving, or even selling more than half the units of Ford's F-series of trucks"?
The winners of the contest are never the same from year to year. The one constant is that the overconfident do poorly. Having extremely confident predictions (e.g. 90% or 10%) on any event that is contentious among participants (has a high standard deviation) is associated with poor performance in the contest. The better performing contestants seem to have an easier time intuitively feeling the range of outcomes that a trip through life's quincunx can deliver.
What seems to happen specifically is that a very rigid set of expectations, a very definite view, a simple narrative, crashes on the rocks of reality. In fact, I am learning to avoid any kind of simple narrative for thinking about the future. Consider predictions such as, "there's going to be hyperinflation / a crash / a civil war" next year. Reality is a lot more complicated. Bitcoin hit an all time high and oil hit an all time low in 2020. What probability would anyone have given that? And does it fit a simple label like "hyperinflation"?
Since we can't predict, what can we do? The only investing strategy that I can think of for coping with the utter impossibility of prediction is diversification among that which is cheap.
Previously, the 2015, 2016, and 2017, 2018, and 2019 prediction contest results.