In addition to back testing models like this, it’s always worth running forward simulations of what would happen if some/most/all agents in the simulation shared your new strategy.