Is 30 trades enough to test a gold strategy?

No. Thirty trades is enough to spot obvious execution and discipline problems, but the confidence interval around the expectancy is still wider than the edge you are trying to measure.

Can I speed up testing with lower timeframes?

Partly. Dropping to M5 gathers trades faster, but the cost ratio rises sharply and you are then testing a different strategy, not the same one at speed.

Does a losing streak mean the edge is gone?

Only if the streak exceeds what the historical distribution predicts, or if the losses cluster in a condition that used to be profitable. Otherwise it is ordinary variance.

SAMPLE SIZE ON GOLD

Gold strategies require 200 trades for scalping, 100 for intraday XAUUSD and 50 for swing trading before results carry statistical meaning.

The sample must also cover at least three CPI prints, three NFP releases and one FOMC decision so that macro-driven behaviour is represented.