Menu

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

Neural Network Trading, Serious people only! 127 replies

Neural Network anyone? 12 replies

Lets build a neural network based scalper together 23 replies

Systems, Systems and Systems... 1 reply

The "Better" neural network 53 replies

Attachments: neural network based systems in forex
Exit Attachments

neural network based systems in forex

  • #1
  • Quote
  • First Post: May 24, 2007 1:31am May 24, 2007 1:31am
Did anyone here used custom neural network based systems in trading forex? The literature i found is mostly the attempts to predict the daily/weekly stock market returns (as represented by SP500 or Dow). The studies showed that the best neural network solutions predicted the sign of the market (if it is up or down) with something like 80% correctness. Together with trading rules those NN solution gave way to trading systems that outperform SP500 or Down by 2-3% annually. Did anyone here applied NN in forex? If so, what are the results.
  • #2
  • Quote
  • May 24, 2007 1:40am May 24, 2007 1:40am
I think you are referring to Vantage Point software. My advise: Save that money and use it in trading instead. You do not need a $4000 software to make money in this market. A few simple lines will help you, as long as you are disciplined and have a strong psychology.
Bullish Attitude!
  • #3
  • Quote
  • May 24, 2007 1:57am May 24, 2007 1:57am
Quoting marketsnipers.c
I think you are referring to Vantage Point software. My advise: Save that money and use it in trading instead. You do not need a $4000 software to make money in this market. A few simple lines will help you, as long as you are disciplined and have a strong psychology.
An expensive program will make you more determined to win and probably help with psychology with confidence in picking a direction.
  • #4
  • Quote
  • May 24, 2007 2:03am May 24, 2007 2:03am
Quoting marketsnipers.c
I think you are referring to Vantage Point software. My advise: Save that money and use it in trading instead. You do not need a $4000 software to make money in this market. A few simple lines will help you, as long as you are disciplined and have a strong psychology.
Actually i was refering to someone who had tried to build the system him/herself and not using tailore made. Anyways than you for your advice. I now know also what available system to use as a benchmark for performance. As i view it, one does not need to follow his/her NN based market prediction in a straight forward fashion. It could be used as an indicator together with other things in building the actual system. NN can give a statistical edge over long term. Correct me if i am wrong on this one.
  • #5
  • Quote
  • May 24, 2007 2:07am May 24, 2007 2:07am
The only neural network you need is between your ears.........so save your money on crap software and put into a trading account.
Same Whore .... Different Dress
  • #6
  • Quote
  • May 24, 2007 2:11am May 24, 2007 2:11am
Quoting AlexanderS
Actually i was refering to someone who had tried to build the system him/herself and not using tailore made. Anyways than you for your advice. I now know also what available system to use as a benchmark for performance. As i view it, one does not need to follow his/her NN based market prediction in a straight forward fashion. It could be used as an indicator together with other things in building the actual system. NN can give a statistical edge over long term. Correct me if i am wrong on this one.
as to a statistacl edge ...... follow the trend, thats the best edge you can have.
Same Whore .... Different Dress
  • #7
  • Quote
  • May 24, 2007 2:19am May 24, 2007 2:19am
Alex my two cents here is that everyone on this forum has a different opinion. Some folks swear by the trend. Some folks swear by the stochastics....

And yes, there is a thread of die-hard neural network traders (VantagePoint.) A few months ago, I posted some challenging remarks to the VP community, and my posts were deleted. I guess I wasn't being "positive" enough.

Anyway, I DO NOT subscribe to NN or Gann Theory or many other predictive methods for that matter. You will find over time that everyone around here swears they are "right." lol
  • #8
  • Quote
  • May 24, 2007 4:10am May 24, 2007 4:10am
An expensive program will make you conscious of your trades enough to scare you and will help in losing your money to the market.

Bottomline, you dont need to predict to make money.

Quoting nitman
An expensive program will make you more determined to win and probably help with psychology with confidence in picking a direction.
Bullish Attitude!
  • #9
  • Quote
  • May 24, 2007 4:12am May 24, 2007 4:12am
There is no need to predict to make money.

Quoting AlexanderS
Actually i was refering to someone who had tried to build the system him/herself and not using tailore made. Anyways than you for your advice. I now know also what available system to use as a benchmark for performance. As i view it, one does not need to follow his/her NN based market prediction in a straight forward fashion. It could be used as an indicator together with other things in building the actual system. NN can give a statistical edge over long term. Correct me if i am wrong on this one.
Bullish Attitude!
  • #10
  • Quote
  • May 24, 2007 4:18am May 24, 2007 4:18am
Quoting AlexanderS
Together with trading rules those NN solution gave way to trading systems that outperform SP500 or Down by 2-3% annually. Did anyone here applied NN in forex? If so, what are the results.
SP500 from what i know grows 10% a year and outperforming with 2-3% doesn't look like NN has a 80% performance to me...
"Abandon all hope, you who enter here" La Divina Commedia, Dante Alighieri
  • #11
  • Quote
  • May 24, 2007 10:12am May 24, 2007 10:12am
Quoting semar
SP500 from what i know grows 10% a year and outperforming with 2-3% doesn't look like NN has a 80% performance to me...
If you read carefully the first post, you would see that i said the 80% applies to the sign(whether it will be up or down) of the market. Obviously you need to add trading rules to that indicator. While adding those rules, researchers got those 2-3%
  • #12
  • Quote
  • May 24, 2007 11:15am May 24, 2007 11:15am
Quoting AlexanderS
Did anyone here used custom neural network based systems in trading forex?
Yes, I look at it from time to time and enjoy it very much. It is not my main way of trading but I find the whole neural network concept very interesting and believe it does have a place. Not as a "black box" but where you control the inputs and know why you take certain decisions.

So if you like it, investigate it for yourself. The best results I had was not with prediciting the trend but with the range of the next day.

Enjoy
  • #13
  • Quote
  • May 25, 2007 1:17am May 25, 2007 1:17am
Quoting marejp
Yes, I look at it from time to time and enjoy it very much. It is not my main way of trading but I find the whole neural network concept very interesting and believe it does have a place. Not as a "black box" but where you control the inputs and know why you take certain decisions.

So if you like it, investigate it for yourself. The best results I had was not with prediciting the trend but with the range of the next day.

Enjoy
I am going to give it a try, i was just curious if people are using them to some success. As to "Black Box", neural network themselves are essentially black box or as i read in one traiding book:" A system with unknown number of rules". I agree however that using them on their own without giving a tought is a path to disaster.
  • #14
  • Quote
  • May 25, 2007 9:02am May 25, 2007 9:02am
you can google wardsystems and look for the neural net site.
They probably have what you are looking for.

I have a neural net from Calsci and it is to labor intensive for my needs. I have resigned to having it pick my lottery numbers. By the way the numbers outperform quick picks. On one draw I missed the jack pot by a +5 / -5 deviation from the draw. I have also matched 3 of 6 on mulitple occasions and never play more than a buck or two.

Anyway, give it a shot. You can probably right it off as a work tool, if you are stuctured correctly.
  • #15
  • Quote
  • Aug 10, 2007 6:11pm Aug 10, 2007 6:11pm
Hi Alexandar and company,

I am creating autotrade systems using machine learning methods, although not neural networks. I DID, however, use neural networks when beginning my work until I got better results with other methods.

You can get free neural network software for R, Matlab and C++, although you need some expertise to learn and use those programs for testing.

The main reasons I don't like Neural Networks so much are 1) They are black boxed by nature 2) It is common to iteratively find the optimum network architecture (layers, nodes, etc), and this can give skewed results due to large sample size (ie. at least some look good due to normal distribution of results) and 3) There are simply better methods out there!

Regards,
-Duke
  • #16
  • Quote
  • Aug 10, 2007 6:21pm Aug 10, 2007 6:21pm
There is an impressive NN for forex called Flexibile Neural Trees (FNTs.) Graduate students in China have posted their results for the USDJPY pair. Google "Flexible Neural Trees" to read the article.

FNTs are a second generation NN, but I don't have specifics of the mechanics.
  • #17
  • Quote
  • Aug 19, 2007 7:49am Aug 19, 2007 7:49am
Quoting duketoga
Hi Alexandar and company,

I am creating autotrade systems using machine learning methods, although not neural networks. I DID, however, use neural networks when beginning my work until I got better results with other methods.

You can get free neural network software for R, Matlab and C++, although you need some expertise to learn and use those programs for testing.

The main reasons I don't like Neural Networks so much are 1) They are black boxed by nature 2) It is common to iteratively find the optimum network architecture (layers, nodes, etc), and this can give skewed results due to large sample size (ie. at least some look good due to normal distribution of results) and 3) There are simply better methods out there!

Regards,
-Duke
Thanks for the info, what machine learning methods are you using?
  • #18
  • Quote
  • Jul 7, 2010 4:45am Jul 7, 2010 4:45am
:-)
  • #19
  • Quote
  • Sep 24, 2010 4:43pm Sep 24, 2010 4:43pm
Quoting AlexanderS
Did anyone here used custom neural network based systems in trading forex? The literature i found is mostly the attempts to predict the daily/weekly stock market returns (as represented by SP500 or Dow). The studies showed that the best neural network solutions predicted the sign of the market (if it is up or down) with something like 80% correctness. Together with trading rules those NN solution gave way to trading systems that outperform SP500 or Down by 2-3% annually. Did anyone here applied NN in forex? If so, what are the results.

I did my own form of neural network that has its roots on SOM Kohonen architecture. I call it Reverse Mirrored SOM Neural Net. I use it as a manual system and I have great accuracy trading with it. Here is a pic of how it looks.
I can tell you that Neural networks is a great solution to recognize price patterns to trade.
Attached Image (click to enlarge)
Click to Enlarge Name: eurusd24-sep.gif Size: 45 KB
  • #20
  • Quote
  • Sep 24, 2010 4:53pm Sep 24, 2010 4:53pm
Best Neural network is your brain..you wont believe it brain is the most powerful PC in an entire universe with 100 billion neurons connected by 1000 trillion synapses.

At the time of our death, 80% of our brain is not utilized. We only use 20%.

To use more neurons and synapses in brain, it requires intense hard work to make journals, learn, revise, form patterns, memorize, practice, feedback etc.

We are willing to do sex for full 8-10 hours in night to get pleasure but we are not willing to do hard work and develop brain processing capabilities.

Needless to say, we are running for automatic softwares.
0 traders viewing now