Over the weekend, the real currency market is closed. A few demo servers keep showing prices, and the charts continue to move.
Those quotes are not real. Some brokers run a simulated feed for demo accounts when the market is closed, to keep the platform active. Prices might follow a random walk, or replay old data, or hold flat.
A scalper who tests a strategy over the weekend on such a demo and sees results is measuring nothing. The prices don't reflect liquidity, order flow or news. Even spreads can be set to arbitrary values.
Worse, some of this data ends up in the demo account's history, mixing fake bars with real ones. Backtests that use that history include the fake periods and produce misleading results.
Spotting it is easy. A one-minute chart over a weekend that has constant movement in a pair that should be closed is a clear sign. Checking the history for bars between Friday's close and Sunday's open shows how the broker handled it.
If the history has weekend bars, cleaning them before testing is essential. If the testing software has an option to exclude weekends, it's worth enabling.
Any period without real trading shouldn't be part of the evidence.
Has weekend demo data ever gotten into a test or an analysis?