What to do with my algorithm
Hello All,
I am not completely sure that I am posting to the right forum, Sorry in advance if it is not very related.
I am senior PhD student in computer science (machine learning), over more than 1 year, I was working in my free times on a market trading algorithm. I implemented my algorithm on a popular forex platform which is given by a bank (But the core of algorithm is not dependent on forex). I tested my algorithm over different long periods (2 years usually) and given that the algorithm doesn't have training, testing on history is not logically different from the live test. But live test was also performed. The bank agreed to give me a live demo account that the report can be visible by any client.
After this brief introduction I have these questions:
1) My algorithm returns around 120% per year profit with maximum use of leverage of 25%, use of leverage is less than 15% in ~95% of times . How does this stand in comparison with state of the art in this area ?
2) What is the best idea to use the algorithm, given that I can not work on it full time and I don't have much money to start hedge fund (if I understood the concept correctly!)
Thanks in advance,
How long have you tested it in real time? You can come up with a million strats thatll backtest well but will not perform in real time with comparable results, and you'll need at least a couple of months of live testing to get anywhere with it.
Is your algo based on machine learning tools or a MACD crossover?
@ seabird
As I said, there is no training, so no overfitting (http://en.wikipedia.org/wiki/Overfitting), therefore no difference between history and live test. Anyway It was tested around 6 months live. In addition to that, As I said, I am constructing a live demo account by the corresponding bank that the account statistics is available live by a url.
My questions are specifically those in the post, I have already passed the evaluation and validity part.
@macro arbitrage I think none, although machine learning is a broad term. It mainly does somehow online optimization.
In comparison with almost any other money management scheme for this year you would be considered a top performer and people would be very happy with you....this year. Most important thing in money management is consistent returns, and sadly you can't prove that yet.
I'm sure you understand the risk of an algorithmic trading platform if you can't dedicate the appropriate time to the system. I don't know the strategy but one day of neglect because you have to finish a research paper and you could be completely screwed. For that reason I dont think it would be smart to try and raise money in any way but that's up to you.
If it's your passion I would say just try to pique the interest of a quant hedge fund and see if they will give you an opportunity for your strategy.
@ jktecon
Actually that maximum use of leverage that I mentioned (25%) was related to this year, you are right, in general the performance of my algorithm is also better on previous years. It is possible to be more aggressive but then it takes more risk.
Again you are right, now I can not prove consistency (unless (maybe!) if I describe the algorithm in details for a professional), and the idea of the live report was to cover this, but it takes time.
Thanks for your advises.
Live trades have something called slippage, how your actions affect the market. If you arbitraging your product
e.g. A is a US commodity, B is it's Dual Listing in the UK, You can buy one in the US, sell in the UK and hedge with the currency THEN this does not matter and your model is valid
If you are trading stocks and you open your trades on a single product and close them later on your model cannot take that into account and as such will not represent the true picture.
If you have scenario A, trade your own account, if you have B, then you have not more than a good demonstration of your technical abilities, I'm sorry to say.
Have sent you a PM, we may be a solution to an opportunity.
T
Rerum ratione a et distinctio eius ab. Corrupti nemo necessitatibus nihil ullam consequuntur quasi. Eos inventore laborum voluptatem praesentium dolorem totam enim. Dolores dicta molestiae vitae ratione cum omnis.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...