Confluence stacking: how many conditions is too many?

[Shadow Trader]

Every trader eventually discovers that adding conditions improves the apparent quality of setups. Sweep, plus order block, plus fair value gap, plus premium discount, plus session timing, plus higher timeframe bias. Each filter removes some losing trades, but each also removes winners, and the number of valid setups drops fast. At some point you have a rule set so tight that you take three trades a year and cannot tell whether the results mean anything statistically. I aim for three to four independent conditions, meaning ones that measure different things. For example, higher timeframe bias, a liquidity sweep, a structure break with displacement, and a premium or discount location. Filters that are variations of the same idea, like two types of zone drawn at the same place, do not add real information. To check whether a filter earns its place, remove it and compare results over the same sample. If nothing changes much, delete it. Simplicity also makes the system easier to execute under pressure, which matters more than a slightly higher backtested win rate.

[FXS HongKong Scalper]

Three to four independent conditions is a good target. Many traders stack conditions that are essentially the same, like order block and fair value gap in the same spot, which adds complexity without new information. Checking how often each condition changes the outcome in your data shows which ones actually matter.