Hello, it is an interesting study in terms of the general framework. But the results look quite mixed. I would especially suspicious for the fact that the best performing results in the table are seemingly randomly distributed; i.e. none of the strategies can be definitely associated with a clear outperformance compared to the others. This means that the results could be random outcomes. Also, many of the trajectories seem to be highly correlated to the buy&hold strategy. This is a further indication that the algo might not yet make strategic buys and sells so well as to circumvent drawdown periods etc. But in general, I think the framework is a good idea and probably needs better policy models to function properly. May I ask how exactly you modelled the trading cost?