When to Close a Trade With an Indicator: 3 of 26 Exit Rules Beat Closing at a Random Time, and the Opposite Cross Was Among the Worst
Key takeaways
- I kept the entry fixed and changed only the exit: 13 exit rules on the same EMA 9/21 crosses and the same Stochastic 14/3/3 crosses, 16,926 EURUSD M5 bars, 1-pip spread.
- Then I gave every rule an honest opponent: the same trades closed after a random number of bars, drawn from that rule’s own holding times, 200 times over.
- 3 of 26 exit rules beat the random exit. All three were ATR trailing stops. 10 of 26 did worse than 95% of the random runs.
- The classic “close on the opposite cross” lost 2.25 pips a trade on the EMA entries and won only 19.7% of trades. The trades had been 5.82 pips in profit on average at their best.
- Win rate said nothing about the result: exiting when RSI 14 reached 60/40 won 57.4% of trades and still lost 1.39 pips each.
Most indicator articles are about the entry. The exit gets one line: “close when the opposite signal appears”. Yet the exit decides how much of a move you keep, and every trader has watched a trade go 5 pips into profit and come back to the stop.
So I asked a narrow question. If the entry stays the same, does the choice of exit indicator change the result, and is any exit better than simply closing at a random moment?
How did I test exit signals?
- Data: EURUSD M5, June 3 to August 26, 2026, 16,926 bars. Median ATR 14 was 1.96 pips, which I measured in my ATR settings test.
- Entries: EMA 9/21 cross (891 signals) and Stochastic 14/3/3 %K/%D cross from below 20 or above 80 (1,852 signals), entered at the close of the signal bar. Both are rules I tested before, in the moving average crossover test and the Stochastic buy and sell signal test.
- Exits: opposite EMA 9/21 cross; close back through EMA 21; RSI 14 reaching 70/30 or 60/40; Parabolic SAR flip; ATR chandelier trail at 1.5, 2 and 3 ATR; fixed stop and target of 1:1, 1:2, 1:3 and 2:2 ATR; close at 21:00 UTC.
- Safety net: every rule had the same emergency stop at 3 ATR and a maximum hold of one day. One trade at a time.
- Cost: 1 pip per trade, the spread I measured in my spread and swap test. When one bar touched both stop and target, the stop counted.
The random exit is the key part. For each rule I took its own list of trades and its own list of holding times, then closed every trade after a holding time picked at random from that list, with the same emergency stop. I repeated this 200 times. If an exit indicator really knows when a move is over, it should beat its random twin.
Which exit worked best on the EMA 9/21 cross?
| Exit rule | Trades | Win rate | Net pips per trade | Median hold | Random exit, 5th to 95th pct | Verdict |
|---|---|---|---|---|---|---|
| ATR chandelier 1.5 | 736 | 24.7% | -1.22 | 20 min | -1.52 to -1.24 | better |
| RSI 14 reaches 70/30 | 310 | 39.4% | -1.30 | 70 min | -2.02 to -1.12 | same |
| RSI 14 reaches 60/40 | 521 | 57.4% | -1.39 | 20 min | -1.68 to -1.13 | same |
| ATR chandelier 2 | 684 | 26.5% | -1.40 | 30 min | -1.62 to -1.25 | same |
| Fixed 1:3 ATR | 716 | 22.8% | -1.49 | 15 min | -1.54 to -1.16 | same |
| Fixed 1:1 ATR | 757 | 41.1% | -1.53 | 10 min | -1.45 to -1.18 | worse |
| Close back through EMA 21 | 760 | 19.7% | -1.57 | 30 min | -1.52 to -1.08 | worse |
| Fixed 1:2 ATR | 737 | 28.1% | -1.63 | 15 min | -1.52 to -1.22 | worse |
| Close at 21:00 UTC | 216 | 20.8% | -1.89 | 135 min | -2.38 to -1.17 | same |
| ATR chandelier 3 | 566 | 27.9% | -1.92 | 50 min | -1.79 to -1.28 | worse |
| Fixed 2:2 ATR | 546 | 42.1% | -1.96 | 25 min | -1.76 to -1.25 | worse |
| Parabolic SAR flip | 582 | 24.7% | -2.17 | 50 min | -1.77 to -1.25 | worse |
| Opposite EMA 9/21 cross | 466 | 19.7% | -2.25 | 50 min | -1.88 to -1.25 | worse |
Net of a 1-pip spread. “Better” means above the 95th percentile of 200 random-exit runs on the same trades, “worse” means below the 5th.
No exit made the EMA cross profitable. That part is not news: the cross itself lost in my earlier test. The news is the order. The two exits most traders pair with a crossover, the opposite cross and the Parabolic SAR flip, were the two worst of 13. Both did worse than 95% of random exits on the very same trades.
How much profit did the trades give back?
I also recorded the best open profit of every trade, the furthest price went in its favour before the exit. With the opposite-cross exit, the EMA trades were on average 5.82 pips in profit at their best and finished at -1.25 pips before the spread. They gave back more than all of it.
| EMA 9/21 entry, exit rule | Average best open profit | Average result before spread |
|---|---|---|
| Opposite EMA 9/21 cross | 5.82 pips | -1.25 pips |
| RSI 14 reaches 70/30 | 5.94 pips | -0.30 pips |
| Parabolic SAR flip | 4.51 pips | -1.17 pips |
| ATR chandelier 1.5 | 2.96 pips | -0.22 pips |
| Fixed 1:1 ATR | 1.54 pips | -0.53 pips |
The reason is lag. An EMA 9/21 cross back the other way needs price to retrace far enough to pull a 9-bar average under a 21-bar one. By then a good part of the move is gone. A trailing stop 1.5 ATR behind the best price reacts within a bar or two. It keeps less of the big winners but cuts the trades that never go anywhere, and on M5 those are most trades.
Did the exits work better on the Stochastic entries?
| Stochastic entry, exit rule | Trades | Win rate | Net pips per trade | Random exit, 5th to 95th pct | Verdict |
|---|---|---|---|---|---|
| Close at 21:00 UTC | 224 | 21.4% | -0.46 | -1.05 to +0.79 | same |
| ATR chandelier 2 | 1,086 | 32.8% | -0.64 | -1.00 to -0.65 | better |
| RSI 14 reaches 60/40 | 745 | 50.5% | -0.67 | -1.01 to -0.45 | same |
| ATR chandelier 1.5 | 1,326 | 29.6% | -0.76 | -1.04 to -0.80 | better |
| RSI 14 reaches 70/30 | 369 | 29.8% | -0.94 | -1.64 to -0.36 | same |
| Fixed 1:3 ATR | 1,338 | 27.4% | -0.95 | -1.04 to -0.70 | same |
| ATR chandelier 3 | 788 | 31.5% | -0.96 | -1.01 to -0.53 | same |
| Fixed 1:2 ATR | 1,545 | 35.9% | -1.00 | -1.07 to -0.81 | same |
| Opposite EMA 9/21 cross | 1,598 | 31.0% | -1.05 | -1.06 to -0.86 | same |
| Close back through EMA 21 | 1,592 | 25.6% | -1.09 | -1.09 to -0.91 | same |
| Fixed 1:1 ATR | 1,664 | 48.7% | -1.13 | -1.08 to -0.89 | worse |
| Parabolic SAR flip | 1,654 | 29.4% | -1.20 | -1.09 to -0.94 | worse |
| Fixed 2:2 ATR | 1,187 | 50.0% | -1.20 | -1.13 to -0.77 | worse |
The Stochastic entry lost less than the EMA entry with every one of the 13 exits. The ranking barely changed: the ATR trails near the top, Parabolic SAR and the fixed 1:1 and 2:2 targets at the bottom. The 21:00 UTC close came first but traded only 224 times, and its random band runs from -1.05 to +0.79, so it proves nothing either way. Across both entries the count is 3 better than random, 13 the same, 10 worse.
Does a high win rate mean a good exit?
No, and this test shows it cleanly because the entries are identical. With the EMA entry, closing when RSI 14 reached 60/40 won 57.4% of trades and lost 1.39 pips a trade. The 1.5 ATR chandelier won 24.7% and lost 1.22. The fixed 1:1 target won 41.1%, but its average winner was 1.19 pips after the spread and its average loser 3.42.
An exit that takes small wins quickly always looks good on the win-rate line. What pays is the size of the average win against the average loss, which is the same lesson as in my stop loss distance test.
Can a good exit rescue a bad entry?
It can make it lose less, not make it pay. I ran the same exits on 319 random entries as a third control. With the 2 ATR chandelier, random entries lost 0.71 pips a trade, less than the EMA cross lost with any of its 13 exits. The same thing showed up in my rule vs random entry test: on M5 EURUSD the entry signal adds less than traders expect, and the exit and the spread decide most of the result.
Holding longer did not help either. With fixed time exits on the EMA entries, 30 minutes lost 1.68 pips, 1 hour 1.90, 2 hours 2.04 and 4 hours 2.19.
So when should you close a trade?
- Use a trailing stop measured in ATR, about 1.5 to 2 ATR behind the best price. It was the only exit family that beat a random exit, on both entries.
- Do not wait for the opposite signal of a slow indicator. The opposite EMA cross and the SAR flip gave back the most.
- Judge an exit by average win against average loss, not by win rate.
- Test your exit against a random exit with the same holding times before you trust it. If it cannot beat that, it is not reading the market.
My fixed vs ATR stop test covers the stop side of the same question, and the Parabolic SAR test shows why the SAR works better as an entry filter than as an exit.
Frequently asked questions
What is the best indicator to know when to close a trade?
In this test, an ATR trailing stop at 1.5 to 2 ATR. It was the only exit of 13 that beat a random exit with the same holding times, on both EMA and Stochastic entries over 16,926 EURUSD M5 bars.
Should I exit on the opposite crossover?
On M5 EURUSD it was among the worst exits. On EMA 9/21 entries it lost 2.25 pips a trade, won 19.7% of trades and did worse than 95% of random exits.
Is Parabolic SAR a good exit?
Not here. The SAR flip was worse than random on both entries, losing 2.17 pips a trade on EMA crosses and 1.20 on Stochastic crosses.
Does RSI work as an exit signal?
RSI 14 reaching 70/30 or 60/40 performed the same as a random exit. It gave the highest win rates, up to 57.4%, but not the best results.
Can a good exit make a losing strategy profitable?
Not in this test. All 26 entry and exit pairs lost after a 1-pip spread. The best exit reduced the loss by about a pip compared with the worst.
Want a clean indicator to install right now?
It is my own enhanced DeMARK Trend Line indicator for MetaTrader 4 and 5. Non repaint, clean, and free.
Forex educator and indicator developer. I build and trade my own MetaTrader tools, and share the ones that genuinely help.