Research

Why ideas are retired.

Ideas leave for reasons, and the reasons are kept. These are the kinds of reasons that have ended ideas here. None of them is a failure of the practice. Each is the practice working.

• It looked into the future without anyone meaning it.

  A measurement joined to data that was not yet known at the time. The result vanished when the join was corrected, and the rule against it was written down.

• The signal was real. The cost was larger.

  Something genuinely predicted, but trading on it paid more than it earned. A signal is not an edge until it survives what it costs to act.

• It was the market in disguise.

  A result that was mostly a tilt toward the market's direction. With the market removed, nothing was left.

• It worked on the data it was fit to, and nowhere else.

  Strong in the stretch it was searched on, absent on data it had not seen. In-sample results are a screen, not a verdict.

• A data fault made a model look stronger than it was.

  A fault in the inputs inflated a result until it was found and corrected. A small real effect survived. The finding was recorded, not hidden.

• The mechanism named never appeared.

  An idea built on a specific flow was checked against a control group that behaved the same way. If the control behaves alike, the mechanism is not the cause.

• Two related names did not move together after all.

  Economic links are not statistical links. Pairs that seemed bound by the same business did not hold together when tested.