Lets say a system uses one or two signals. It should be not too difficult with open source software that is presently available. Watch the Ted talk
1 Like
Good luck with that. For me, I have little enough confidence in other people’s strategies in general that the thought of trusting AI to figure out how the strategy runs and then using that as a strategy of my own makes me want to laugh and cry at the same time. Or drink. Or run away. Or just about anything except pour time into such a concept and then pour money into it when I get a backtest that more or less matches that strategy. Perhaps see if you can reverse engineer your own strategies first. We’ll wait here, come back and tell us how it goes. 
2 Likes
As i look at your system, It helps me to understand the basis of your response
1 Like
Touché. I’m not happy that I started out the gate with a drawdown and then a second one as soon as things got going, but they’re both within the expected range for my strategy, and I’m satisfied with how Battle Axe is working out. I’m in it for the long run… feel free to come back and see where I am in a year or two.
I have contemplated trying machine learning techniques to highlight additional possible strategy components for me in the volatility asset class, but that’s wholly different from trying to reverse engineer what someone else is doing with their strategy. Battle Axe is a quantitative strategy based on sensible, logically considered signals. If there isn’t a logical reason for the alpha in a strategy, it’s not a strategy I want to put my money into.
Thus to complete my thought from earlier, even were I to use AI to reverse engineer someone else’s volatility strategy that interested me, I would have to take that generated strategy apart further and find the logical basis (if one exists) behind it, else it would not serve me. Likely at best, it might give me a new avenue for consideration. But truthfully, knowing how many different datasets I’ve examined looking for logical meaning that actually produces alpha in a backtest, and how many more possible (similar, but unique) datasets are available for use, I highly doubt that I could provide the machine learning enough datasets to use to actually find a realistic match to someone else’s strategy that interests me.
Could machine learning reverse engineer someone’s trend-following moving average crossover strategy that is based only on price action of the instrument being traded? Sure, likely it could. But if the strategy adds a little wrinkle like a trend confirmation based on some other instrument or index, you’ll likely only get a reliable match if the AI has access to that data and knows what to try to do with it (ie has been trained to look for such confirming trend indicators to better match the trade history).
There, now I’ve written too much, but I hope this now helps you to understand the basis of my response much more than could the lackluster equity curve of my Battle Axe strategy in its short life span to date. And once again touché, as I did not intend an attack on you nor on your strategies (I haven’t even looked to see what you trade or how), and although I may have written my prior response in a flippant fashion, I meant what I said: Good luck, and by all means come back and tell us of your experiences if you try this exercise.
1 Like
