Gold does not behave the way trading manuals say. After more than a year of testing automated systems on XAUUSD, we found five real edges. Here we share four of them.
The data comes from the MetaTrader 5 Strategy Tester with real ticks, 2025 to 2026, across more than 2,000 trades. Every idea was tested over several different periods and with doubled spread, to discard the ones that only worked by luck.
1. Gold is not the same market all day
The same strategy wins in some hours and loses in others. In our tests, a single window of about 3 hours in the afternoon held a large share of the losses. Once we stopped trading in that window, the system improved without changing anything else.
The Asian session is also different from London and New York: candle size, spread and the way price respects levels all change.
Lesson: before changing indicators, check what time you lose. Often the problem is not the strategy, it is the clock.
2. A perfect simulation can be a trap
We tested an idea that looked spectacular in our own simulation: adding a second entry with a limit order. In the MT5 tester with real ticks, those entries won only 7% of the time.
The cause was one detail: a buy limit order fills at the Ask price, not at the price you see on the candle. On gold, with its spread, that difference turns an edge into a loss. The simulation filled orders that, in reality, only filled when price was moving against us.
Lesson: if an idea looks too good, test it with real ticks and doubled spread. If it does not survive, do not use it. We dropped it.
3. The big chart decides what happens on the small one
We trade short timeframes, but the best improvements did not come from fine-tuning the entry. They came from looking at a much higher timeframe and never trading against it.
A perfect signal on a minutes chart fails far more often when the hourly chart points the other way. This filter improved results in every period we tested, not just one.
Lesson: before you enter, move up a timeframe and ask which side the bigger market is on.
4. Two different systems beat one perfect system
We had two gold systems with completely different logic. We measured how their daily results moved together and the correlation was −0.02: practically zero. The bad days of one almost never matched the bad days of the other.
Running both on the same account, the tester showed almost double the profit with the same maximum drawdown (around 42% in both cases). A drawdown like that is high and you must be ready for it, but the point is that profit went up without risk going up.
Lesson: instead of searching for the perfect system, look for a good one that thinks differently from the one you already have.
What we tested and did not work
Sharing what fails saves months. We tested these on gold and dropped them:
- Fast moving-average crossovers as entry signals. They arrive late and repeat in every range.
- Trendy “zero-lag” or adaptive indicators. They improved some weeks and failed others; fragile results.
- More than 20 new strategies built on well-known indicators. None beat what we already had once tested across different periods.
- Over-optimizing parameters. If one value works and its neighbors do not, it is luck.
5. The fifth discovery
There is one more edge, and it is the one that changed our results the most. We will not share it here: it took months of testing to find, and it is the heart of what we built.
Just one hint: gold does not move alone.
If this topic interests you, leave your question in the comments and we will keep sharing what we learn about XAUUSD in future posts.
Risk notice: Strategy Tester results do not guarantee future results. Leveraged trading can make you lose all your capital.
Flavio Benitez — MQL5 developer focused on gold (XAUUSD).
