Holding-Period Discipline

Every trading idea is built around some horizon: a statistical arbitrage signal might be designed to mean-revert over hours, a value thesis might need a year or two to play out, an earnings trade might have a shelf life of exactly one trading session. Holding-period discipline means matching how long you actually stay in a position to the timescale the idea was designed for — not exiting a multi-month thesis on day two because it hasn't moved yet, and not sitting in a one-day statistical signal for three weeks because it's now "just a position."

The reason this is harder than it sounds is that price movement doesn't respect your intended horizon. A short-term signal can sit flat for days and then work all at once; a long-term thesis can move sharply in your favor in the first week for reasons that have nothing to do with the original logic, tempting an early exit that locks in a smaller win than the thesis was actually worth. Both situations create pressure to override the original holding period, and both kinds of override tend to hurt performance on average, even though any single instance might look lucky in hindsight.

Worked example

A trader takes a position based on a signal backtested to have an average holding period of five trading days and a win rate that only becomes reliable when trades are allowed to run their full course. Three days in, the position is up nicely, and the trader is tempted to take profit early — "why give it back." Historically, though, this exact signal continues working for the remaining two days about as often as it reverses, and cutting early systematically caps the upside on winners while doing nothing to protect against the losers, which get held to the same five days regardless. If the trader takes early profits on winners but holds losers to term, the realized average outcome drifts worse than what the backtest promised, purely from breaking the holding-period rule asymmetrically.

The practical fix is deciding the intended holding period before entry, writing it down alongside the thesis, and treating an early exit as something that needs its own separate justification — new information that changes the thesis, not just discomfort with an unrealized gain or loss moving around.