Live trades for NN EA idea - page 2

 

Leeb

Why 0.01 first entry?

Is this to test the market?Could you not code it to test the market?

Is this money management reasons?.

It looks good .Keep it up

OILFXPRO

 

lot size

Hi Oilfxpro, yes 0.01 is to test and get market info, it could be coded just to follow it and obtain same info but as most of these trades are profitable anyway I have left as is.

 
davidke20:
Thanks leeb. I have no experience with NN at all. But am interested since its a hot topic for the moment, I'd like to see some promising stuff produce by some senior member of this forum. Fair enough to keep a secret, I can respect that point until people say give me 50 bucks, so you get my EA to test. So far there is only buy positions available, so we'll see when it changed from buy => sell, that will give a bigger picture of how the EA play with market reversal.

Regards

David

There was a pretty complicated NN EA posted way back by some French guy, which seemed to do pretty well on some pairs - this included the source.

Personally, I view NN as the ultimate in curve-fitting. There is/was a trader from Australia who used to post in the Wealthlab forum ('brucev') who both used them in trading and was also doing a Masters on them - so he knew his s*it. Might make interesting (and sobering) reading...

 

Hi omelette, yep very interesting point curve-fitting and neural nets seem to be very similar at a high level

 

Hi leeb, Not sure if you are looking for ideas to add on to the EA or if it is excellant in and of itself. But if you are looking for ideas an indicator posted recently looks worthy of a look......Synthetic VIX posted by mrebersv

leeb:
Hi omelette, yep very interesting point curve-fitting and neural nets seem to be very similar at a high level
 
leeb:
Hi omelette, yep very interesting point curve-fitting and neural nets seem to be very similar at a high level

I'm sure a few may have seen a documentary where the US army decided to use NN for spotting deeply camouflaged tanks hidden in a forest. Surprisingly, it performed incredibly well, so much so that they were ecstatic - until they tried replicating the results!

Turns out all the NN was managing to do was differenciate between sunny and cloudy days

The moral of the story is that although all these really clever people were convinced it was doing what they 'expected', it wasn't! They were feeding it input data and hoping it would output the answers they were looking for. However NN are non-plussed by your hopes and dreams - they simply are fed inputs, use this data to generate outputs which are 'steered' in the direction of known 'correct' outputs by continually adjusting the node weights.

And this is also a near perfect example of curve-fitting in action...

 
omelette:
There is/was a trader from Australia who used to post in the Wealthlab forum ('brucev') who both used them in trading and was also doing a Masters on them - so he knew his s*it. Might make interesting (and sobering) reading...

If I remember correctly the guys called Bruce James Vanstone, and his PhD thesis is available for download, from the Bond University website. It's probably worth a read for anyone interested in AI, its easy enough to follow even for non technical readers.

From memory he develops a four different systems for the Australian ASX200 based on both technical and fundamental data.

From what I recall, the conclusions from Bruce's research was that NN'S where useful, there's also some good ideas presented regarding how you might go about developing a methodology for designing and testing any sort of trading system so its worth a read, but there's nothing earth shattering about the approach he uses.

I believe he's now working at Porter Capital Management

Welcome

 
zupcon:
If I remember correctly the guys called Bruce James Vanstone, and his PhD thesis is available for download, from the Bond University website. It's probably worth a read for anyone interested in AI, its easy enough to follow even for non technical readers.

From memory he develops a four different systems for the Australian ASX200 based on both technical and fundamental data.

From what I recall, the conclusions from Bruce's research was that NN'S where useful, there's also some good ideas presented regarding how you might go about developing a methodology for designing and testing any sort of trading system so its worth a read, but there's nothing earth shattering about the approach he uses.

I believe he's now working at Porter Capital Management

Welcome

That's a hell of a memory you've got!

But yes, he was really good at explaining 'complicated' stuff. What I took from his postings was that yes, NN can be useful when applied properly and probably more important, selectively.

And to do this you've got to understand what you're doing! I spent many a wasted hour playing with NeuroSolutions demo but never really achieving anything...

 
 
omelette:
I'm sure a few may have seen a documentary where the US army decided to use NN for spotting deeply camouflaged tanks hidden in a forest. Surprisingly, it performed incredibly well, so much so that they were ecstatic - until they tried replicating the results!

Turns out all the NN was managing to do was differenciate between sunny and cloudy days

The moral of the story is that although all these really clever people were convinced it was doing what they 'expected', it wasn't! They were feeding it input data and hoping it would output the answers they were looking for. However NN are non-plussed by your hopes and dreams - they simply are fed inputs, use this data to generate outputs which are 'steered' in the direction of known 'correct' outputs by continually adjusting the node weights.

And this is also a near perfect example of curve-fitting in action...

That's correct, I did my first test on NN, I used the data of almost 10 indicators to train a NN based on market patterns and Indicators signals.

Instead of learn how to trade, the NN memorized the past data.

The result is amazing performance in the training period, but totally fail in the out of sample data. The best curve-fitting results I saw in my life.

I am still a newbie in that field and more I learn more I discover that I still need to learn.

My big concern is that sooner we will see an avalanche of Sellers selling the best curve-fitting systems you can ever imagine.

Files: