PiNNacle - a Neural Network EA - 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
I'm working on techniques/trade rules to make it multi-timeframe, well the simplest trade rules. That way it will be able to best decide which timeframes to make the trade calculations with.
The idea that makes the most sense right now is to monitor ADX on multiple timeframes and trade on the next lower timeframe below, or two timeframes below, the one that is trending. So if ADX is >20 on the 1hr then it's probably ok to trade on the 15m signals. Or, if ADX h1 is ranging, but 30m ADX>20 then trade on the 5m.
Something along those lines. I'll post it once I get it working in code.
Bcktesting
nittany1,
I realized you r not answering people keep asking why your NN does not show good in backtesting...is there any data fed differnrt between fresh and old while your system calculating the neural?
Many thanks for sharing this EA. I really appreciate it, because it is the first time I see how to lay out a NN and that helped me a lot. I have some questions for a better understanding.
1) Hidden nodes weights. As you said, they have been calculated externally and then hardcoded. In case of additional training/optimization, they could differ from the actual values. That means they should be changed in the code?
2) Same question about buy/sell levels (in this case the values are defined as external, so no need to recompile)
3) Buy/Sell nodes. Why the signal generated by MFI is included only in the Sell hidden nodes?
4) Weight again. Absolute values of Buy and Sell weight are equals. Is it a must, it is a choice of this particular perceptron, are they derived from training/optimization or what else?
Many thanks for your reply
skipper
Thanks very much ,I like all EAs which depend on neutral network .
Do neural network EAs work?
Do these neural network EAs work in today's market?
I'm working on techniques/trade rules to make it multi-timeframe, well the simplest trade rules. That way it will be able to best decide which timeframes to make the trade calculations with.
The idea that makes the most sense right now is to monitor ADX on multiple timeframes and trade on the next lower timeframe below, or two timeframes below, the one that is trending. So if ADX is >20 on the 1hr then it's probably ok to trade on the 15m signals. Or, if ADX h1 is ranging, but 30m ADX>20 then trade on the 5m.
Something along those lines. I'll post it once I get it working in code.
Firstly I want to thank you for sharing your EA, I was researching various ways of implementing the NN and I think your's is the simplest way, the weights can be calculated in python using the NUMPY stack, I like your idea to manage the timeframes, did you have any luck avoiding the ranging times?, I backtested it a lot and of the drawdown areas were also ranging zones. Comes back with positive results but is probably due to curvefitting, I'm testing it live now along with another set with some changes added to the code I'll keep you posted of how that goes.