We help IT Professionals succeed at work.

Can neural networks be used to filter bad trades?

PearlJamFanatic
on
169 Views
Last Modified: 2017-05-20
I have a trading system that is about 40% right. That means 4/10 signals generated by it yield in profit and the remaining 6 result in loss. Can I use neural networks/AI/Deep-learning to improve the chances? Can neural networks be used, in this case, to filter-out the bad trades in advance such that the trading system produces 60% good trades?
Comment
Watch Question

Patrick BogersDatacenter platform engineer Lindows
CERTIFIED EXPERT

Commented:
If you hit 40% correct your strategy is not correct. Simple.
Trading is about trends, news, sentiment, numbers which you cannot simply implment into software.

Making money while trading takes a lot of time keeping updated and dedication.

Author

Commented:
That may not be true. For example let the cost of trading (round-trip brokerage and taxes) be $1/trade and lets assume that the winner make $20 each winning trade and losers lose $10 each losing trade. Then for 10 trades Profit is (20 X 4) = $80 and Loss is (10 X 6) = $60 and Cost is (1 x 10) = $10. That would mean total profit would be 80-60-10=$10 for every 10 trades.

Anyway my question was more about neural networks and their ability to filter out noise.
Patrick BogersDatacenter platform engineer Lindows
CERTIFIED EXPERT

Commented:
I guess it would be possible a neural network can learn from trades that turned into a loss but imagine how much info it must process and interprete.
I know big daytraders that work/experiment with this technology which costs millions and they trade in hundreds of dollar cents.

But technically? Yes it can

Author

Commented:
I really don't understand your response. The neural network api is opensource/free and even if you buy customised tools tailor made to work with timeseries then it won't cost more than a few hundred dollars. I don't think cost is a factor.
Patrick BogersDatacenter platform engineer Lindows
CERTIFIED EXPERT

Commented:
Maybe i did not interpreted your question not correctly.
Can I use neural networks/AI/Deep-learning to improve the chances?
Gave me the idea you were asking if a new build, privatly held neural network can help improve your odds.
CERTIFIED EXPERT

Commented:
Can you find, even after the fact, prior data that could have predicted that a particular trade would be a loser?  
Is your neural network exposed to and trained on this sort of data?
Has the market failed to evolve to a point where this data is no longer relevant?

A neural network is an attempt to recognize hidden patterns in a complex and constantly evolving system.  A successful neural network has to be at least as complex as the system is analyzing.  All the connections, including the ones that weren't there when I started writing this sentence have to be there.  Or it will make some mistakes.

So how good is your neural network?

Author

Commented:
I don't have a neural network. I don't know what this magical word even means. But I do have a trading system coded in Java that gives buy and sell signals. And on using this system on live data I observed that it gives good trades 40% of the times. My question is can neural network be used as filtering system on top of my trading system such that the resultant system gives 60% good trades in place of the earlier 40% rate.

Author

Commented:
I guess my question was of the general kind. Take a look at this:

http://www.breakyourhead.com/2013/03/stock-prediction-artificial-neural.html

and

https://nseindia.com/content/research/FinalPaper206.pdf

and

https://people.eecs.berkeley.edu/~akar/IITK_website/EE671/report_stock.pdf

These people seem to find it very easy to use neural network on past closing price to get future closing price. I really doubt if it were that simple. I also doubt if the prediction made by their systems is useful.
CERTIFIED EXPERT
Awarded 2010
Top Expert 2013
Commented:
Unlock this solution and get a sample of our free trial.
(No credit card required)
UNLOCK SOLUTION
CERTIFIED EXPERT
Commented:
Unlock this solution and get a sample of our free trial.
(No credit card required)
UNLOCK SOLUTION
CERTIFIED EXPERT
Awarded 2010
Top Expert 2013
Commented:
Unlock this solution and get a sample of our free trial.
(No credit card required)
UNLOCK SOLUTION

Author

Commented:
TommySzalapski So I guess the people who wrote the above papers are wrong. Because they just concluded, in their papers, that it is possible to use neural nets for predicting the closing-price of an index/stock.

Author

Commented:
TommySzalapski well offcourse nobody can predict things like "airline crew beating up the passenger" etc. That would be a black swan event. But generally speaking barring any untoward incident can a neural net trained system pick 6 out of 10 winners? I am not sure it can.
CERTIFIED EXPERT
Awarded 2010
Top Expert 2013
Commented:
Unlock this solution and get a sample of our free trial.
(No credit card required)
UNLOCK SOLUTION

Author

Commented:
I agree with you on that. I have, now, collected atleast 12 research papers 'thesis' written by final year students of top IVY league on how to feed historical closing price of stocks to neural net to get future price. I wonder if these guys were awarded degrees on the basis of their thesis.

Can you point out the common mistake in the above papers? (refer ID: 42109041)
CERTIFIED EXPERT
Awarded 2010
Top Expert 2013
Commented:
Unlock this solution and get a sample of our free trial.
(No credit card required)
UNLOCK SOLUTION
CERTIFIED EXPERT

Commented:
A cautionary tale and a good explanation from the experts.
Unlock the solution to this question.
Thanks for using Experts Exchange.

Please provide your email to receive a sample view!

*This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

OR

Please enter a first name

Please enter a last name

8+ characters (letters, numbers, and a symbol)

By clicking, you agree to the Terms of Use and Privacy Policy.