Gold breakouts: stand aside after a run-up
One of our gold systems buys and sells breaks of a daily level. In our testing it behaved very differently depending on what gold had done in the previous month, so we measured that instead of tuning a number.
The state
The state is the net move of gold over the last month of trading days, compared with its own readings over the last year (backtest, gold, BlackBull, Nov 2022 to Oct 2026). It is computed from gold's own closing prices, known before the day starts, and uses no calendar and no fixed threshold: "high" and "low" mean above or below gold's own recent middle.
What we found
- After a strong run-up (drift above its own middle), level breaks lost on average in both of two non-overlapping periods, Nov 2022 to Oct 2024 and Nov 2024 to Oct 2026. The relationship had the same sign in both periods.
- After a quiet or falling month (drift below its own middle), the same breaks earned.
- The mirror version, trading only after run-ups, was worse in the second period, as the idea predicts.
- With the switch on, the system was profitable in both periods and its worst equity drawdown in the first period fell from about 22 % to about 12 %. It trades about half as often.
(Backtest, MetaTrader 5 Strategy Tester, real ticks, gold, BlackBull, two periods as stated, $10,000 start.)
How much to trust it
- The reason is plausible, not proven: a breakout needs room to run, and after a long advance much of that room has been used.
- It rests on two periods and several variants that were tried, so we treat it as a candidate. The product that uses it, Keystone, stays in the Optimising group.
- Profit without the largest five percent of trades is not positive in either period, so the result leans on a few large winners.
- It is a backtest. It is not a forecast, and it says nothing about direction.
Related: Regime cards for strategies that work in some markets.