OptiNod Academy
Time stops — When to close a trade that is not moving
What changes if you close slow-moving trades early? This article compares 5-bar and 10-bar time stops on total profit, average result, holding period and drawdown.
You can close a position that has not moved for a long time even if it never reached the stop. A time stop is a rule that sets how long to wait and how far price must move in that time together.
Some trades are profitable right after entry, and others drift around the entry price for days. Checking whether this difference relates to the final result lets us examine "how long to wait".
In this BTC daily sample, of the 52 trades that reached +1R between the day of entry and two days later, about 87% went on to 2R before the stop. Among the 17 trades that took 11 to 20 bars to reach 1R, it was about 41%. There was a difference between fast moves and the later hit rate, but not every slow starter failed.
Based on this observation, a time condition was added to the 3×ATR trailing stop. Counting the entry bar as bar 0, if MFE is below 0.5R by the 5th or 10th bar, the trade is closed at that day's close. This assumes the decision at the close and the fill are at the same price. The two settings were computed separately, and the common entry rule and the exclusion of costs are the same as in the previous parts.
A hit time learned afterward differs from a condition you can judge now
The classification "trades that reached 1R within 2 bars" can only be known after that arrival has happened. To close trades that have not reached 1R as time passes, it has to be turned into a condition that is knowable at that moment.
So hypothetical trades still open at bar 10 after entry were examined separately. Of the 11 trades whose MFE had been below 0.5R by then, about 27% later went to 2R, and among the 125 that had moved 0.5R or more, about 70% did. This difference can be a basis for testing a time stop.
However, 11 trades is a small sample. We cannot say that 0.5R and 10 bars are appropriate criteria for other instruments. You need to check the results of all trades with this condition applied together with the results from a period that was not used to choose the condition.
With the 10-bar condition, holding period shortened and drawdown stayed the same
The base 3×ATR trailing stop had 61 trades, total +37.1R and an average of +0.61R per trade. Adding the 10-bar time stop gives 63 trades, a total of +37.1R and an average of +0.59R. Total profit is the same when shown to one decimal place, but the average per trade is slightly lower.
The average holding period fell from 15.6 bars to 14.5 bars. The maximum drawdown, computed from the R results of closed trades, is -6.0R in both cases. In this sample the 10-bar condition shortened the holding period, but it did not reduce drawdown on this basis.
Only 3 trades were actually closed by the time condition. Many of the trades that could not move for a long time reached the initial stop or the trailing stop before bar 10. Adding a time condition does not change every trade significantly.
The 5-bar condition closed more trades early
Pulling the criterion in to 5 bars, 18 of 69 total trades ended by time stop. Total profit was +39.8R and the average per trade was +0.58R. Total profit rose by 2.7R over the base rule, but the average per trade fell.
The average holding period fell from 15.6 bars to 12.4 bars, and the maximum drawdown on closed trades fell from -6.0R to -5.0R. This change has to be kept separate from the 10-bar result in the previous section. You cannot combine only the good parts of the two settings by saying "total profit was the same and holding period and drawdown fell".
Closing early lets you take later signals, so the set of trades changes too. The fact that total profit rose does not mean only that the existing trades were closed better. Check the change in the existing trades together with the contribution of the newly entered trades.
Trades that moved late and were profitable can also be cut
The trade bought at $7,330 on April 7, 2020 had an MFE of only 0.12R by bar 10 and was time-stopped at -0.27R. Held without the time condition, it would have ended at +1.17R on May 10. Excluding trades that move late has a cost as well.
If you keep applying the same entry condition after a time stop, you can receive a new signal on this instrument again. In this case it re-entered at the April 24 open of $7,484 and ended at +1.39R on May 10. Note, though, that the R of each of the two trades is based on the stop distance at the time of its entry. To compare them as cumulative R at the same risk amount, you need the assumption that the size at each entry is matched to that risk amount.
This single case does not guarantee that re-entry will succeed. Decide in advance whether to re-enter after a time stop and how long to ignore the same signal, and include those conditions in the calculation. If you decide on re-entry by picking only cases that rose after the exit, only the favorable cases remain.
Compare with time and progress fixed together
"Wait for a few days" alone can also close a trade whose price has already moved enough. Conversely, setting only the required progress leaves out when to judge. As in this comparison, set the N bars and the MFE criterion together.
- Time criterion: The entry bar is bar 0, and the judgment is made once, at the close of the 5th or 10th bar.
- Movement criterion: If MFE is below 0.5R by then, close on a time stop. If the existing stop was hit first, the trade is treated as already closed at that price.
- Keep condition: If it moved 0.5R or more, continue with the existing 3×ATR trailing stop.
- Re-entry: Check for new signals from the day after the exit, and enter at the open of the bar after the signal. Recompute the initial stop and 1R.
- Comparison items: Take total profit, average per trade, holding period and closed-trade drawdown all from the same setting.
Whether to use a time stop is judged by looking at the benefit of shortening the waiting time together with the results of the trades that were cut. In this sample the 5-bar and 10-bar settings gave different results, so do not generalize either result into a property of time stops as a whole.