Article: Price forecasting with neural networks - page 3
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Can you tell us a little bit more about the millionaire's journey?
Well, I don't look like a Soros :)
And you, Schwager :)
I just wanted to say a few words in defense of neural networks. Because of the widespread belief among Runet traders that NS is just a useless toy. In fact it's a powerful tool for forecasting financial markets, but it's also very complicated. That's why after 2-3 months of trying to master some neuro-package and no results traders have wrong opinion about NS. It's like that joke:
You don't like cats? You just don't know how to cook them!
You don't like cats? You just don't know how to cook them!
No, I haven't dealt with cats yet (haven't cooked or consumed;o))
Have you used neural networks at all in your trade to achieve such an impressive result? Or do you just know them well and apply them in some other field?
- the use of grids to predict exchange rates and even the direction of exchange rates proves less effective than the use of simple classical methods of technical analysis. Predictions of relatively simple grids do not exceed 70-75%.
Predicting the direction of a currency at 70-75% is from the realm of fiction.
I have been making such predictions for a long time, working through a bookmaker who accepted bets on currency appreciation/depreciation over a fixed period of time (intraday). A rate match at the beginning and end of the period was considered a loss. At first, the bookmaker's commissions were so small that strategies with only 52% of correct predictions yielded profits. At first I used a simple system based on tehanalysis, which gave me about 54-55% winnings.
Then the bookmaker's commissions increased and I had to improve the trading system. I took all the indicators I was using and put them in .
The winning percentage increased to 59-60%. So there are tasks in whichneural networks rule, despite the opinions of sceptics!
With the "winning percentage" of 60% at a spread of ~5pp you can live not even in Sochi, but somewhere in Hawaii... :)
The only problem in which networks rule is remembering the pattern and searching for it in the future.
And by the way, it's not enough just to put a bunch of indicators into NS.
The NS require an exceptional skill from the user, to get something meaningful from the NS
It is possible only by correctly posing the question, and correctly interpreting the result.
Neural networks can do a lot, the problem is in the preparation of the initial data, as an example - EURUSD M5 chart, but the network has not yet been fully optimized, I am working on it now. All lines on the chart are the result of neural network operation without using any indicators.
Have you used neural networks at all in your trade to achieve such an impressive result? Or are you just familiar with them and have you used them in some other field?
No, I have not used them in other areas. I used them in financial betting since 2005. Then I switched to Forex and tried to use neural networks on 1-minute data. But profitability of these systems depended on quotes of certain brokerage companies and that's why they worked for a very short period of time - from one week to two months. Everything depended on the brokerage company. I managed to take profit in honest brokerage companies, but in some other ones it was still hanging in a blocked account :).
After that I came to a conclusion that the stable long term operation requires a system with more timeframes, that it works in any brokerage company and does not depend on quotes. And that is what I am doing now.
Neural networks can do many things, the whole problem is in preparing the initial data, as an example - EURUSD M5 chart, but I have not optimised the network yet, I am working on it now. All lines on the chart are the result of the neural network operation without using any indicators.
I' m working on it now. All lines on the chart are the result of neural network operation without using any indicators.
At this stage I am still testing the neural network itself in automatic dynamic mode on a demo account, it is recalculated on each new bar. These lines are the forecasts, and in the picture if you look carefully, you can see that they show the dynamics ahead, but the optimization of forming these forecasts, and also the block that will interpret them and form the control trading signals, I'm not involved yet, I'm debugging the overall algorithm of the system.
I tried to use them for three years unsuccessfully, then I gave up for several years. After I understood the market better and created adequate trading strategies I came back to neural networks a couple of months ago with encouraging results.
Piligrimm, would you like to work together on the NS, in terms of writing
- the training block
- interpretation unit
- decision block
- the position control block, in terms of the NS.
If yes, please write kirillovi@gmail.com.
I am not a specialist in this field, but I know some things, I am familiar with BrainMaker, MathLab (Neural Network), I have some experience in writing LE in MQL4. I have some experience in combining MS and DataMining - the latter performs processing and sample preparation for training.
Regards, kirillov.
Piligrimm, would you like to work together on the NS, in terms of writing
- the training block
- interpretation unit
- decision block
- the position control block, in terms of the NS.
If yes, please write kirillovi@gmail.com.
I am not a specialist in this field, but I know some things, I am familiar with BrainMaker, MathLab (Neural Network), I have some experience in writing LE in MQL4. I have some experience in combining MS and DataMining - the latter performs processing and sample preparation for training.
Regards, kirillov.
Thanks for your suggestion, but I have no problem with creating models based on neural networks as I do on other packages.
If there are any problems, it's with linking models to the terminal, I'm not a programmer and I don't know MQL4, but I'm solving this problem at the very least.
At this stage I am still testing the neural network itself in automatic dynamic mode on a demo account, it is recalculated on each new bar. These lines are forecasts and in the picture if you look carefully you can see that they are showing the dynamics ahead, but I have not yet optimized formation of these forecasts, and also the block that will interpret them and form control trade signals, because I am still debugging the overall algorithm of the system.
I tried to use them for 3 years unsuccessfully, then I abandoned them for several years. After I understood the market better and was able to create adequate trading strategies I came back to neural networks a couple of months ago and received encouraging results.
That's great.
I'd like to see how it looks to the right of zero bar. I've got it like this ('Did you see this picture?', p. 1 message of 28.03.2007 22:34). But this is also a crude version, on a small sample. And not on grids, but on corr. analysis.