obscurity. There was not a single expert who demonstrated either an accuracy or a calibration above 50%. So in aggregate, we're not very good. But when you take the average of everybody's response, that was better than any single expert. And alongside this, he also ran some simple algorithms. So over the short term, things typically will revert to the mean. Over the long term, things will typically follow the same trend that we've observed over the last many years or decades. And these simple algorithmic approaches outperformed all of the individual experts, the aggregation of all experts, like the average of their views, and had an accuracy level that was greater than 50%. And by virtue of their design, were as well calibrated as you could be, because they were able to determine. When you say I performed all of the experts, I'm assuming you mean all of the experts across a wide range of questions that they're asked, right? Correct. Yeah. So the capability of the algorithmic approaches (11/45)
You are viewing a single comment's thread from: