Neural Network system in action - Live

Intel_Bvc

Junior member
Messages
41
Likes
3
In the short time I've been on this forum I've seen a great deal of skepticism around the performance of Neural Networks in Forex trading. I believe that this is due to an awful lot of systems out there that have tried to exploit the hype and mystery around this concept and they've ultimately fallen down or been straight up bogus and burnt a lot of people.

However, my system is nothing like any other system out there and I'm prepared to showcase it to all those interested in seeing a truly revolutionary system in action.

I setup a Demo account yestrdy afternoon, to access via a read-only login simply use the below details:

MetaTrader 4 Client
Server: MetaQuotes-Demo
Account: 1813127
P/W: 6twsqlv

Start Balance: $120k

Just under 2 weeks from now the system will reach a point of equilibrium and they'll be anything from 250-400 open trades on at any point in time so I will be updating the thread with the aggregated volumes and positions so that people can get a better view than from the MT4 client as to exactly how the system is positioned going into all the major data releases.

Due to the demo minimum trade size being 0.1 lots I've had to scale up the system by a factor of 10 (So initial deposit was 120k). However the system can be traded using micro-lots (0.01 lots) and min account of 12k.

Once it has settled the system should average about $9k profit per week. The projected profit for the next year is $480k. That's a 400% return on investment.

Any questions please feel free to ask!

Note: You can download the MT4 Client for your PC here: http://www.metatrader4.com/ or goto app store and download on your iPhone or Android device.
 
Last edited:
...and more to the point has your neural network taken consideration of the virtual dealer plugin and other similar applications?
 
...and more to the point has your neural network taken consideration of the virtual dealer plugin and other similar applications?

Do you really think an MT4 broker is going to let you make 400%?

pboyles, its not a scalping system, I'm typically holding trades for 4-10 days with approx mean profit per trade of 70 pips. I have to admit I've never heard of this virtual dealer plugin but slippage of a few pips should not have a detrimental affect on the performance/results of the system.

I've ran an earlier version of this system on a metaquotes demo account before only a few months back. I started with $50k and made $5k a week for 5 weeks straight. I'm sure it should be fine.

I'm being transparent about what I'm doing, so I suppose if there's anything untoward going on on the server then I'm sure the experienced members like yourself could point it out.
 
My main point is that most, if not all, MT4 brokers are taking the other side of your trades, if you win they lose. I seriously doubt they will let you take 9k a week off them. The virtual dealer plugin is designed to carry out a number of functions, not just artificial slippage. For instance it is used to create price spikes.
 
Also the assumption that the live trades will be the same as demo trades is way off the mark. There will be a world of difference.
 
If you have managed to program it into MT4 then you should be able to do the same on platforms that trade by DMA. These will not have the same issues as you get with MT4 especially if you are swing or position trading.
 
I dont think it really matters. Its a neural network, they will blow of their own accord. I've never had one network, even with stunning "hindsight" equity curve, give me a positive number in real trading
That is of course just me, so i will look with eager anticipation to be proved incorrect
 
Also the assumption that the live trades will be the same as demo trades is way off the mark. There will be a world of difference.

I dont think it really matters. Its a neural network, they will blow of their own accord. I've never had one network, even with stunning "hindsight" equity curve, give me a positive number in real trading
That is of course just me, so i will look with eager anticipation to be proved incorrect

Guys, I fully understand your cynicism here. I truly do. Neural Nets have gained a bad reputation.

However, I've created an account and shared the login details for everyone to see. All we can do is watch this space.

pboyles, I do not agree that the demo and live trades are wayyy off the mark. Please provide an explanation of exactly what you mean? I've traded this system on a live account and the only difference I see between demo and live is that the prices are a tiny bit different, the occasional spike, and that I cant trade GBP/AUD because its not available on MetaQuotes Demo server!

I've been honest on this forum as to the issues I've had trading live and they're all my own issues like fear through incorrect account sizing and meddling with the trades/strategies in the run-up to big data release like NFP etc.

However, the system keeps grinding away performing in line with the stats which is why I'm in the process of halving my risk and going in on a joint £12k account with a partner which I'm just waiting to be opened.

Malaguti, this is not like any other Neural Network system you've ever come across. That's all I can say really. Its more than a "trading system". It gives 1 week of 15-min interval future price predictions for each of the currencies I trade. Over time I have noticed certain consistent patterns on the predictions.I've then created code that looks for these patterns in the prediction outputs (which are generated every hour on the hour) and if that pattern is found a trade is placed based on the expected behaviour.

So I didnt data-mine these strategies, I studied the outputs literally for years and then I was able to go back and find all the occurrences of these patterns and quantify just how accurate or not these 'events' are. If the stats are good and they give good Mean per Trade, Std Dev, Max Drawdown, T-Stat etc then I hook it up and add it to my portfolio.

If you look at the Comment for every trade you can see the Strategy that generated that trade. Each Strategy / currency-pair combination has its own optimal hold time which varies between 4-10 days.

I'm not saying this is some supa-dupa crystal ball system. I know its not.

But what I do know is that it will eventually settle on a 60-65% win ratio and that the amount it makes on winning trades will be less than the amount it loses on losing trades.

I've heard someone say to me that Nets are slow to react, my nets build up a massively complex web of hedges, it doesn't necessarily need to "turn around". Of course it generates wrong signals but often another strategy or another set of trades on a different currency-pair altogether quickly off-sets those losses.
 
There are 2 main issues

1. You have a vendor badge.

2. You are promising 400% a year return

I have an awful feeling this will not work out well.
 
Guys, I fully understand your cynicism here. I truly do. Neural Nets have gained a bad reputation.

However, I've created an account and shared the login details for everyone to see. All we can do is watch this space.

pboyles, I do not agree that the demo and live trades are wayyy off the mark. Please provide an explanation of exactly what you mean? I've traded this system on a live account and the only difference I see between demo and live is that the prices are a tiny bit different, the occasional spike, and that I cant trade GBP/AUD because its not available on MetaQuotes Demo server!

I've been honest on this forum as to the issues I've had trading live and they're all my own issues like fear through incorrect account sizing and meddling with the trades/strategies in the run-up to big data release like NFP etc.

However, the system keeps grinding away performing in line with the stats which is why I'm in the process of halving my risk and going in on a joint £12k account with a partner which I'm just waiting to be opened.

Malaguti, this is not like any other Neural Network system you've ever come across. That's all I can say really. Its more than a "trading system". It gives 1 week of 15-min interval future price predictions for each of the currencies I trade. Over time I have noticed certain consistent patterns on the predictions.I've then created code that looks for these patterns in the prediction outputs (which are generated every hour on the hour) and if that pattern is found a trade is placed based on the expected behaviour.

So I didnt data-mine these strategies, I studied the outputs literally for years and then I was able to go back and find all the occurrences of these patterns and quantify just how accurate or not these 'events' are. If the stats are good and they give good Mean per Trade, Std Dev, Max Drawdown, T-Stat etc then I hook it up and add it to my portfolio.

If you look at the Comment for every trade you can see the Strategy that generated that trade. Each Strategy / currency-pair combination has its own optimal hold time which varies between 4-10 days.

I'm not saying this is some supa-dupa crystal ball system. I know its not.

But what I do know is that it will eventually settle on a 60-65% win ratio and that the amount it makes on winning trades will be less than the amount it loses on losing trades.

I've heard someone say to me that Nets are slow to react, my nets build up a massively complex web of hedges, it doesn't necessarily need to "turn around". Of course it generates wrong signals but often another strategy or another set of trades on a different currency-pair altogether quickly off-sets those losses.

then trade it, show us the neural signal, your entry and lets put it to the test
 
then trade it, show us the neural signal, your entry and lets put it to the test

That's the whole point of this thread!

MetaTrader 4 Client
Server: MetaQuotes-Demo
Account: 1813127
P/W: 6twsqlv

Start Balance: $120k

Btw, I will be trading a scaled down version of this on a real money account i.e micro-lots and £12K start balance before the end of this week.
 
Last edited:
That's the whole point of this thread!

MetaTrader 4 Client
Server: MetaQuotes-Demo
Account: 1813127
P/W: 6twsqlv

Start Balance: $120k

no, the point of the thread is neural network live..
and for all those lucky people who have don't have metatrader..
 
no, the point of the thread is neural network live..
and for all those lucky people who have don't have metatrader..

You can download the app for your phone for free on android or iPhone, or you can download the client for your PC which would probably give you a better insight into the complex web of trades the system generates.

I will be posting how its performing and how the system is positioned hopefully on a daily basis, but its very early for now, I just switched it on. Trades are going on every hour on the hour so certainly wouldn't be practical for me to post every signal.

Its currently up 0.75% over the last 24 hours but that doesn't mean anything, the floating P&L of the system actually tends to swings a lot and is not necessarily indicative of how the systems performing. A better indicator I've found is the daily Change in Equity. This gives an indication as to whether the system is currently doing well. And I'll also try and provide the updated chart for this going forward.

It will take approx 2 weeks before the system reaches equilibrium (trades are being opened at the same rate they're being closed).

Then it should settle into a steady rate of closing off trades for overall profit.
 
You can download the app for your phone for free on android or iPhone, or you can download the client for your PC which would probably give you a better insight into the complex web of trades the system generates.

I will be posting how its performing and how the system is positioned hopefully on a daily basis, but its very early for now, I just switched it on. Trades are going on every hour on the hour so certainly wouldn't be practical for my to post every signal.

Its currently up 0.75% over the last 24 hours but that doesn't mean anything, the floating P&L of the system actually tends to swings a lot and is not necessarily indicative of how the systems performing. A better indicator I've found is the daily Change in Equity. This gives an indication as to whether the system is currently doing well. And I'll also try and provide the updated chart for this going forward.

It will take approx 2 weeks before the system reaches equilibrium (trades are being opened at the same rate they're being closed).

Then it should settle into a steady rate of closing of trades for profit.

gotcha, apologies Intel
 
Ok, so I'm attaching a screenshot of the current positioning of the system.

Will try and update daily if possible.

Its still very much early days, but currently the system has 79 open trades (each 0.1 lots). Once its fully settled I'd expect the system to maintain approx 250-450 open trades at any point in time. At which time these positioning updates will prove even more beneficial to attempt to untangle the complex web of trades that the system generates.

My main concern at the moment is that going into NFP on Friday the system is unlikely to have had enough time to fully established its web of trades leaving it potentially vulnerable if the data release "doesn't go its way"... But we'll see how it goes, we have some very big data releases throughout the whole week to be honest.

Despite this being a purely technical system I've found that fundamental factors like data releases and press conferences often provide the catalysts that push price in the right direction to make profit. Hence I've found its best to just leave the system to run and not try to second guess it or interfere with the trades it has on.
 

Attachments

  • 05062013_Positioning.bmp
    911.2 KB · Views: 248
Actually this week is pretty news-full, with NFP ADP having been announced today (which sometimes gives a clue as to the Fri announcement). So your system will be sailing into a tough first week of trialing - so good luck and I look forward to hearing results as they unfold. Big respect to you for making the platform available.

Regards

Nigel
 
Actually this week is pretty news-full, with NFP ADP having been announced today (which sometimes gives a clue as to the Fri announcement). So your system will be sailing into a tough first week of trialing - so good luck and I look forward to hearing results as they unfold. Big respect to you for making the platform available.

Regards

Nigel

Thanks Nigel. So far, so good, but its very very early days.

You don't need to wait for me to publish results though, you can download the MT4 client for your PC, iPhone or Android device and keep tabs on it as and when you wish.

Forex Trading Platform MetaTrader 4

Login details provided at the start of the thread.

If you install Mt4 client on your desktop you actually have an option to see a detailed statement of the account which will include every single trade, whether pending, open or closed, and a section with detailed stats like win rate, max drawdown, average win, average loss etc etc.

It is a crazy week for data, Beige Book out later, EUR and UK Interest rate announcements tomorrow then of course NFP on Friday.

Like I've said before these releases act as the catalysts which can push price in my direction or on the other hand, they can quickly reverse gains. However, although they present a definite risk, my statistics have proven that by leaving all the trades on, over time, the system will ultimately produce consistent profit.

One of the things that have hurt me in the past when trading this on a real money account is that going into these big events I have some massive positions on and I get scared (due to incorrectly sized account) and I start meddling with the trades. But this is a no-no because as you'll eventually see what the system does is build up a complex web of hedges and if you stop one side of the hedge from going on then you leave yourself massively vulnerable if price goes the wrong way.
 
There is a thread on T2W by Arreyex in which he claims to have bent neuroshell NN to work with MT4.
Are you using an off the shelf NN system ?
 
No, this is a 100% original neural network system that I've built from scratch. I use no other 3rd party software.

At university back in 2002, my final year project involved using Neural Networks to map the output of Patch Antennae.

I took what I learnt from that and applied it to mapping the price movements in FX
 
Top