Not wise to post too much in the open. To be fair, the wheel wasn't reinvented. Yet I think my implementation is solid, particularly on the risk management side.
Asset class is shares. Free variables are 7 per sub-system, although 3 are non-critical and used for simple volatility-based stops and targets. As for impossible to run this strategy, there are scenarios that would make it less profitable. Like a hike in margin requirements, or lack of shares available to borrow. Yet since it's highly liquid and easy to borrow, this seems unlikely.
Year-to-date, 2017 backtested return on $50K capital is $24K after costs. No hysteresis. Not sure what you mean with percent in/out but to clarify, performance report is un-compounded. It uses a fixed lotsize of $50K per trade (per sub-system, so leverage up to 2:1).
I was not asking for capital, I was asking for the return on turnover amount. That is usually gives you a good sense of sensitivity to the transaction costs and execution quality. Your execution assumptions are pretty aggressive, IMHO, but that is not that horrible. Ok, let's take it a different way. What are your execution assumptions, including (a) delay from the signal to fill if you are testing on barred data (b) if you are testing on tick data, do you assume a fractional fill (b.1) if you are testing on tick data, what fraction of traded size do you assume (c) are you trading the same size every time or you're using the strength of the signal for sizing? What percent of the time do you have a position in the market? That, combined with return on trade value, would tell you a lot of interesting things. (a) what is the fraction of buys out of all trades? (b) if (a) is close to 0 or to 1, take your average trade return and subtract an average change of your instrument - what is the Sharpe and Sortino ratios of the strategy after that? This, btw, is the best test for risk premium strategies. I think I know what the instrument is, you don't have to be all mysterious with me My personal rule of thumb is that you divide the Sharpe from the back test by square root of the number of free variables (including the base hypothesis).4 is a lot and 7 is really a lot, IMHO. What are you assuming as cost of borrow? Not that it matters given the returns, TBH.
I think my assumptions are conservative when I look at bid/ask spread during regular trading hours. That's 1 cent per share round-turn for spread, plus 1.5 cents commission which is higher than what IB charges all-in fixed pricing. Please keep in mind average trade is high, so there is leeway. Good point. During development, 25% of data was out-of-sample. Once completed, for final version I recalibrated over all data. The sim-trading that I offered would provide results on truly unseen data.
The idea was to demonstrate the strategy to an interested party. What you see in your own account always has more validity. Everything in life is risky. Maybe I find someone I trust over time. But yes I agree, there are ways to keep the code and execute it for a business partner or prop firm.
You do have a point here. Consider it from this angle: $10K was the worst DD since early 2009. You start trading at the worst possible time, that's what you get. However you may as well start with $50K and then not increase leverage for a while, which reduces DD in % as the account grows. Overall I agree, risk appetite is an individual thing. Thanks for the replies. Will answer more tomorrow.
You can mimic the signals from your strat to any CFD retail broker with the 200 bucks minimum deposit.