Machines would have an even greater advantage here. They know exactly the second best move, and would easily calculate it based on any set of constraints. Humans are worse at the increased complexity
But then they'd just always play second best until they at best drew, or (more likely) meet with a forced move that's inherently the best engine move and lose.
You need a different engine that's focussed on not only avoiding conventionally top moves itself, but also forcing its opponent into them.
The catch here is that the engine doesn't understand the objective in its search. To the engine, getting the king in check with an obvious response is no issue, in human chess its game over.
My bet (uninformed, very novice at chess) is that it's likely there's guaranteed setups that would always catch an engine.
