Time stops: the exit nobody talks about

[Shadow Trader]

Stops and targets get all the attention. A time stop, closing a trade after a fixed number of seconds or minutes regardless of price, rarely gets a mention, yet it can transform a scalping system.

The logic: many entry signals have a short shelf life. If price hasn't moved in the expected direction within, say, ninety seconds, the reason for the trade has probably expired. Staying in means holding a position with no remaining edge, exposed to random drift and the next news spike.

To find the right interval, plot average trade result against holding time. Often profits peak somewhere in the first few minutes and then fade toward zero or turn negative. The time stop is placed near the peak.

It also frees capital and attention, and reduces the chance of a stale trade turning into a large loss.

Time stops also pair neatly with the MAE and MFE analysis from earlier. If winners typically hit their target within 90 seconds and trades that go beyond 3 minutes are mostly losers, a 3-minute cut-off removes the dead weight with surprisingly little downside.

A caution: time stops that trigger during a quiet moment can lock in a small loss right before the move arrives. Testing on tick data, not bars, avoids fooling oneself about that.

A practical way to find the right cut-off is to build a small table. Rows are holding-time buckets: 0 to 30 seconds, 30 to 60, 1 to 2 minutes, 2 to 3, 3 to 5, and over 5. Columns are the number of trades, average net result, and win rate. If the strategy has an edge concentrated in the first two minutes, the table will show it clearly: strong results in the early rows, fading to zero or worse in the later ones.

The cut-off belongs somewhere near where the average result crosses zero. Cutting earlier sacrifices profitable trades. Cutting later keeps dead weight.

Variation by market condition matters. In active sessions the edge may fade after ninety seconds, and in quiet ones it may last five minutes. A dynamic time stop, scaled to recent volatility, can capture that, though it adds complexity and one more parameter to overfit.

Before adopting any time stop, test it on a separate period. A pattern in holding times that's strong in one quarter and gone in the next was probably noise, and the table should be rebuilt periodically to check whether it still holds.

Has a time stop been tested, and what interval worked best for the strategy?