I developed a strategy that I have been testing heavily and now am at the point where I need to make the right decision on how to optimize and launch live. I am kind of new, and would really thank and appreciate someone experienced to share some tips for me ..as this would improve significantly my confidence in algo trading.
I have selected 5-6 stock instruments for this particular strategies by doing a broad optimization on the S&P 500 inex. When I do back-testing, I manage to get pretty good combined result for all 6 instruments. For 4 months I get 100-150 trades, with a cumulative profit of appox 50%-80% and a max drawdown of approx 5%.
I use slippage of 1 as I reckon it would not be possible to make perfect entries when in real market conditions. I also include commissions.
I however realized that backtesting is not the most reliable way to configure the strategy as it does not provide any consistency on of out of box sampling.
That's why I decided to do walkforward optimization.
My biggest concern is on what sample period to use for optimizing, and what would be period for out of box testing.
I did some tests with
90 days / 30 days
45 days / 7 days
90 days / 7 days
So far the best results are from the 45 / 7 days..and profits seem to be consistently positive on a weekly/monthly basis for a sample of 4-5 months (whole period) of combined out of box sample testing.
I however found out that most articles and recommendations are for keeping a ratio of 3:1 for in/out of box walkforward optimization.
My question is:
- should I aim at 3:1 walkforward optimization, or in this case 45 days / 7 (which is approx 6:1) is also ok as well? The 90 / 7 also results in consistent gains ..but with a sligthly lower net profit.
- The profit I manage to achieve with walkforward optimization is approximately 50% from the optimum I manage to get when I optimize for the whole period at once. Are these results sufficient to launch the strategy live and risk real money for this new strategy?
- Is it better to run the same strategy on one instrument with higher volume trades..or to run it simutlaneously for e.g 5 or 10 instruments with smaller volumes?
Any input would be greatly appreciated.
Thanks!
Comment