A common assumption in automated trading is that a robot should remain active whenever the market is open. I do not think that is always a useful design goal.
A strategy is normally developed around a particular type of market behavior. If the recent market stops rewarding that behavior, forcing the same level of participation can turn normal strategy variance into unnecessary exposure.
This was one of the reasons I added an adaptive evidence gate to Aurum Vector Motion, my XAUUSD M1 Expert Advisor for MetaTrader 5.
The gate has three operating states: NORMAL, REDUCED and OFF. It does not replace the entry strategy. A setup must first pass the normal directional, structure and execution checks. The gate is an additional question asked near the end of the process: has the recent evidence from otherwise-eligible opportunities been strong enough to justify normal participation?
The standard observation begins with the latest four completed trading days. The system requires a minimum evidence sample and gives more weight to recent days. If the evidence supports normal participation, the gate can remain NORMAL. If it is weaker but still acceptable, REDUCED mode lowers entry frequency. If it falls below the required conditions, OFF can block normal new entries.
The important part is what happens while the gate is restrictive. Simply turning a strategy off creates a recovery problem: how does it know when to start again if it stops collecting information?
Aurum Vector Motion therefore maintains shadow observations of otherwise-valid opportunities. These are not extra live orders. They are an internal evidence mechanism designed to keep observing the type of setup the strategy would normally trade. If sufficient evidence improves, limited participation can be restored according to the gate rules.
This approach is not a guarantee that the gate will always identify a bad regime in advance. No regime filter can remove uncertainty. It is a way of structuring the decision to participate so that the EA is not forced to treat every recent environment as equally favorable.
There is also a trade-off. More gating can reduce drawdown in one sample while accidentally filtering profitable opportunities in another. That is why a regime control should be tested separately and on unseen data rather than judged only by how well it improves one historical equity curve.
For me, the useful question is not “how many trades can the EA find?” It is “what evidence justifies taking the next trade?”
Aurum Vector Motion is available in the MQL5 Market with a Strategy Tester demo. Historical testing is not a forecast, and users should reproduce tests with their own broker data before considering live use.
