Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
Hey guys, just wondering, what's a good way to make sure that your optimized NT strategy works going forward on a certain set of stocks? I'm curious to know whether an optimized strategy "sticks" to its optimized instruments. Would working on different timeframes be enough to tell?
Can you help answer these questions from other members on NexusFi?
In my experience (and I had to learn this for myself because I wouldn't believe people who told me otherwise) optimization of trading systems rarely works except on a very basic level.
Optimize on half your historical data - taking precautions that you are not curve-fitting (you should know what that is before you start optimizing), and then run the optimized system on the other half of your data. If it is profitable on the unseen data, then it could be profitable in the future.
You have only one chance with any particular set of data. If the optimized system fails to reproduce the expected profits on the blind data, that's it. The system either doesn't work or is not optimizable. You can't go back and choose a different optimization to see whether that works on the blind data.
And after that is done and you created a system which worked on blind historical data, you can put the system into simulated trading for a year to see if it still holds up.
You can discover what your enemy fears most by observing the means he uses to frighten you.
Thanks, how does one know if something is or is not curve-fitting? I am assuming I do have something that works because it works on different years but to varying levels of success.
I wouldn't optimize, it's that simple. When you have to choose a value for a parameter in your trading system, choose one that makes sense from your knowledge of the markets you're trading it on. If you really have no idea what would be sensible, then make a graph of the tested values vs the profit that value results in. If the graph is nicely sloping up to a brow and then back down again, it's pretty obvious you have found an optimum value.
If your graph looks like a seismograph read-out from the Fukushima earthquake, then you can't optimise the parameter.
You can discover what your enemy fears most by observing the means he uses to frighten you.
That's sounds like a good thing to try, might take some time though, given that I have a good # of parameters. Of course this means keeping all the other parameters the same right?
The honest answer is you've asked a question that takes years to properly understand the answer to.
There are hundreds of posts about this subject on the forum, spread out, maybe even a few good threads that focus on answering it. I know one thread top of my head where it was discussed:
I'm creating this thread with the purpose of a 'catch all' for any trading related question that futures.io (formerly BMT) members want to ask, but don't want to create a new thread for -- or find an existing thread to reply to.
I've …
I am sure it is primarily mentioned in other Elite threads where most all the automated discussion takes place.
But long story short, I look for thousands of trades over a multi-year period. I use market orders, and always incorporate 1 tick slippage to make sure I am hitting the offer price as a realistic fill. From there, you need to consider the macro economy during your backtest period --- was it bullish or bearish, etc.
Last, a single strategy is not the holy grail by any stretch of the imagination. The best way to approach it is from a portfolio standpoint, combining multiple uncorrelated markets and strategies into a single portfolio, analyze the total risk and total correlation between trades. Focus should be here. Ernie Chan recently did a good webinar on this subject:
Best advice heard this week. Simple truth is the lump of fat and water between the ears is still 1000X better at feature extraction than any hot silicon. Its not so much about having an edge as edge detection in the first place, minus all the illusions.
Too bad no one listens. The OP probably hasn't even come back to this thread after posting. I noticed he hasn't Thanked a single post since he's been on the forum --- not a good sign that he is paying attention