I also think the random deviation from "best" strategy isn't realistic.
In the real world, true randomness is kind of unusual. There's usually some kind of spatial or temporal autocorrelation.
So something might change in a quasi-random way, but it often doesn't just return to "normal" immediately. There's often something continuing (like an antagonist, or some environmental variable that decays over time, but slowly).
So these random "blips" do happen in reality, but they're not normal. What's more normal is for there to be changes, and for those changes to be somewhat, but not completely, stable with some decay. They might not decay at all.
In the real world, true randomness is kind of unusual. There's usually some kind of spatial or temporal autocorrelation.
So something might change in a quasi-random way, but it often doesn't just return to "normal" immediately. There's often something continuing (like an antagonist, or some environmental variable that decays over time, but slowly).
So these random "blips" do happen in reality, but they're not normal. What's more normal is for there to be changes, and for those changes to be somewhat, but not completely, stable with some decay. They might not decay at all.