Thanks for posting them as long as you did. I did pretty well following your signals (with stops) and don't think they looked that bad at all. Good luck in your endeavors going forward!
James, thanks for this interesting study and the posting of all those signals. What do you think are the main reasons for the poor performance in this case? From memory the backtest results looked fantastic but the live trading never lived up to expectations. Even though Dakota does walk forward testing do you think that there may have been curve-fitting in parameter range and input selection while performing the original model development? Was the data in the out-of-sample period just too uncharacteristic of the in-sample data?