You would need something like 1000 instances of each LLM putting on trades and have a 1000 random walks to judge an average sharpe ratio or something along thos

You would need something like 1000 instances of each LLM putting on trades and have a 1000 random walks to judge an average sharpe ratio or something along those lines. As is, this means absolutely nothing and not understanding the problem. Adding a random walk to this would mean you have 4 random walks instead of 3. There is also the problem that it is tough to make a prediction for tomorrow that is better than today's close.