I think the more complex the chain is however, the more data is required to train it? I’ll be working with random data, so it may actually be the case that less is more for me. I might even want to reduce the model further to interval data as opposed to pitch and duration data. It’s very general, but more specific than the heuristic I’m currently using.
It might take a similar amount of time to take the easier path and to collect random ideas until I have a suitable gene pool 