Trading System vs Its Own Entry Signal: The 200 EMA Filter Made 8 of 12 Runs Worse, and the Exits Beat a Plain Hold 13 Times in 20

A trading system is usually sold as a whole, with the entry signal at the centre. I wanted to know which part of a system actually changes the result. So I ran the five systems from my forex trading systems test three ways. The trend filter, the part most courses insist on, made results worse more often than better. The exits mattered more than the filter, in both directions.

How did I separate a system into its parts?

• Full system: filter, entry, stop, target and time limit, one trade at a time, as described in the systems test.
• No filter: the same entry and exits, but trades in both directions regardless of the 200 EMA or SMA. Only three systems have a trend filter: moving average trend (EMA 20/50 cross), MACD and 200 EMA, and RSI-2.
• Entry only: the same filtered entries, closed after 12 bars of the chart (1 hour on M5, 3 hours on M15), with no stop or target. All five systems.
• Data and cost: EURUSD M5 June 3 to August 26, 2026 (16,926 bars), gold M5 June 16 to August 26, 2026 (13,736 bars), M15 built from them. 1 pip per EURUSD trade, $0.30 per ounce of gold.

Because trades are taken one at a time, a different exit changes how many signals get skipped, so the trade counts differ a little between versions. I show them in every table.

Did the 200 EMA trend filter help?

EURUSD in pips, gold in dollars an ounce. The MACD system without its filter also drops the “below zero for a buy” condition, since that is part of the same trend idea.

The filter was better in 4 of 12 runs, one of them by only 0.04 pips (MACD on EURUSD M5), and worse in 8. On EURUSD M5 it never turned a losing system into a winning one. The one place it did was the moving average system on gold M15, from -$1.72 to +$0.51, and that run had only 38 trades and lost money in the first half.

A 200 EMA filter keeps you on the side of the longer trend. On these charts, over 12 weeks, that did not pay in most runs. My higher-timeframe filter vs slow moving average test looks at the same filter question from another angle.

Did the stop and target do better than a plain hold?

Ten of the 20 runs. The exits were better in 13 of 20: all four Bollinger and RSI runs but gold M5, all four London breakout runs, and 6 of the 12 others.

The exits changed results more than the filter did, and in both directions. The two systems with exits built around the market, Bollinger and RSI (exit at the middle band) and the London breakout (stop at the range, close at 16:00), did better with their own exits in 7 of 8 runs. On gold M15 the Bollinger entry lost $1.48 an ounce held for 12 bars and made +$1.69 with its own exit. The entry finds a stretched price; the exit takes the snap back and does not wait for it to drift away.

The fixed ATR brackets were mixed. The moving average system’s 1.5 ATR stop with a 3 ATR target turned +0.95 pips a trade into -1.22 on EURUSD M15, and -0.14 into -1.07 on EURUSD M5. A 1:2 target looks good on paper, but here about two trades in three lost. My fixed vs ATR stop test and the stop loss distance test show how much stop placement changes a 5-minute result.

Is the 12-bar hold result real, or luck?

Some of the better 12-bar numbers come from few trades. The MACD system on gold M15 made +$3.80 an ounce a trade with the hold, on 57 trades over 10 weeks while gold rose $323. Without more data I would not trust it, and I would not trust the MA trend’s +0.95 on 58 EURUSD M15 trades either. The useful finding here is the direction of the effect across 20 runs, not any single cell.

What does this mean if you build your own system?

• Test the filter, do not assume it. Run the system with and without it. Here the 200 EMA filter hurt two times in three. My indicator combos vs their own parts test found that adding confirmation indicators did not lift the result either.
• Spend more time on the exit than on the entry. The same Bollinger entry went from a loss to the only passing system in my test by changing how it closed.
• Match the exit to the idea. A reversion entry wants a reversion target, such as the middle band. A trend entry with a fixed 1:2 bracket on M5 mostly paid the stop.
• Read results by half and against random entries. The algorithmic rule vs random entry test shows how to do that.

The Bollinger and RSI rules, with the exit that made them work, are set up step by step in my Bollinger Bands and RSI system guide for MT4. If you want to try exits on your own entry, the exit signals test compares 26 exit rules on one entry.

Frequently asked questions

Does a 200 EMA filter improve a trading system?

Not in this test. Across three systems on EURUSD and gold, M5 and M15, the 200 EMA or SMA trend filter made the result worse in 8 of 12 runs. On EURUSD M5 it never turned a losing system into a winning one.

What matters more in a trading system, the entry or the exit?

Here, the exit. The system’s own stop and target beat a plain 12-bar hold in 13 of 20 runs, and the same Bollinger and RSI entry on gold M15 went from -$1.48 to +$1.69 an ounce a trade just by changing the exit.

Is a 1:2 risk-reward ratio a good exit?

Not for the moving average system on EURUSD: a 1.5 ATR stop with a 3 ATR target lost 1.07 pips a trade on M5, against 0.14 for a plain one-hour hold, with about two trades in three losing.

Should I trade only in the direction of the trend?

The trend filter alone did not add an edge here. Test it on your own system with and without the filter, using the same exits and the spread, before keeping it.