If someone claims to have predictive power for _something_, he/she should provide only two things: a point estimate and a confidence interval for t+h (h is given). No words, no jargon, no "explanation," no stories, etc. Just those two things. Then we can start.
-
Not even suggesting people do this maliciously (though I’m sure some do). Self deception and confirmation bias are incredibly powerful.
I even see this with professional trading teams. Them: our backtest has a Sharpe of 2 but our realized track is only about 1.2 Me: looking at their realized track which clearly has a Sharpe of 0.6
