I also created a new python only simulation simulate-funding-v2.py that is similar to the bitcoind based simulation.py which I forked from achow101’s coin-selection-simulation project.
Running simulate-funding-v2.py is much faster because it doesn’t test bitcoind’s complex coin selection algorithms. It simply picks the best single input for each simulated funding attempt. My goals is to find a heuristic that pays the least total fees for each simulation given estimates of the feerate probability distribution and funding request probability distribution.
I’ll post my results from the simplified simulations once I’ve validated them using the full bitcoind simulation.