What do you mean about your profits? All $4500 were from shorting stocks and none were from going long? I'm also in the middle of developing an automated trading system but mine is based on statistical probabilities (quantitative I believe) and summarizes which indicators are appropriate in a particular situation. Good luck on your new system.
Chabah on Automated Trading
- Thread starter Chabah
- Start date
Are you walk-forward testing these optimizations? I probably missed it because I only read the last 2 pages. I can create the smoothest equity curves with 99.5+% confidence intervals in backtesting but when tested foward they break even or don't produce enough to even bother trading.... or sometimes even worse.
But I'm much like you. I have a small range of systems which are individually optimized for each stock and I optimize only 2-3 variables to avoid over fitting as well. I usually test them foward at a 20 to 1 ratio (20 history, 1 forward) and that gives me good results and gives me a good indicator as to whether I should activate/deactivate my system for that day/week/month.
Also, do you have an standard deviation or confidence numbers for your trades? If I can remember correctly 50% of your total short gross was due to 1 trade?
Good luck with your system.
Hey guys, thanks for commenting. For some reason I didn't receive my normal email notifications.
Regarding my earlier results, there were plenty of long profits and losses, and they were in balance. My short trades were where the profits were at, which was remarkable a few months ago because the market seemed to be going straight up. Since then I've been making money mostly on the long side trading names like VMW and CME using my updated trading tools. Simple stuff, I will share soon but right now I am on the road and don't have my TS.
The equity curve thing, well, if it is really smooth then you probably have TONS of trades with a somewhat unrealistic system behind it. In particular, over-optimized on too many variables. My systems usually only have a few dozen trades per year so it won't be smooth.
In general, the way you want to handle optimization is to choose a period of time, optimize, then test against a completely different period of time. Don't include the time used to optimize in the second test or it will shade your results too much.
A good example is to optimize twice, once using Jan-Mar and once using July-Sept. Assuming both results are fairly close (if not, that's a problem), then take the optimized system and run it against Apr-June and Oct-Dec. If your results still look good, then you very well might have a system. If it falls apart, well, then you probably don't.
For more details, check out the Acrary thread, he talks about how to evaluate different profit factors, combine systems, check for correlation, etc.
Using past time frames is also a good way to complete your testing without waiting for any kind of walk forward. Your "walk forward" is still historical data so you can do more research now.
Enjoy,
Chabah
Regarding my earlier results, there were plenty of long profits and losses, and they were in balance. My short trades were where the profits were at, which was remarkable a few months ago because the market seemed to be going straight up. Since then I've been making money mostly on the long side trading names like VMW and CME using my updated trading tools. Simple stuff, I will share soon but right now I am on the road and don't have my TS.
The equity curve thing, well, if it is really smooth then you probably have TONS of trades with a somewhat unrealistic system behind it. In particular, over-optimized on too many variables. My systems usually only have a few dozen trades per year so it won't be smooth.
In general, the way you want to handle optimization is to choose a period of time, optimize, then test against a completely different period of time. Don't include the time used to optimize in the second test or it will shade your results too much.
A good example is to optimize twice, once using Jan-Mar and once using July-Sept. Assuming both results are fairly close (if not, that's a problem), then take the optimized system and run it against Apr-June and Oct-Dec. If your results still look good, then you very well might have a system. If it falls apart, well, then you probably don't.
For more details, check out the Acrary thread, he talks about how to evaluate different profit factors, combine systems, check for correlation, etc.
Using past time frames is also a good way to complete your testing without waiting for any kind of walk forward. Your "walk forward" is still historical data so you can do more research now.
Enjoy,
Chabah
Individual stocks are subject to fundamental issues as well as overall market sentiment. Why not just trade indexes. Pure price action. The edge in stock trading IMHO comes from fundamentals and event driven issues.
Dude chill with resurrecting old threads just to advertise your blog.... at the very least make your blog look more professional before spamming