Better NN EA - page 10

New comment
 
zupcon:
Explain to me in simple terms why you cant back test a neural network ?

I don't know ............

So perhaps you can explain why

You should be asking the question to New Digital

OILFXPRO

 
oilfxpro:
I don't know ............

So perhaps you can explain why

You should be asking the question to New Digital

OILFXPRO

Now I'm as confused as you appear to be !

I suspect that there are only 2 basic approaches you can use. The first approach involves designing and training some form of network external to MT4. Once the network topology and weights are established, the output from the network, for any given set of inputs can be calculated by either hard coding the weights into the EA, or by calling a DLL. This is no different than using any type of indicator, and I cant see why back testing would be an issue.

Alternatively, you could design a solution where you require the network to be continually retrained every bar, or every N bars. This approach could cause a few problems with the back tester, I don't know, cost I've never tried it. The approach sounds like the ultimate in curve fitting to me.

 
mrebersv:
I'm not sure that the way Chessmaster and Deepfritz work can be applied to the financial markets. Those programs essentially compute as many iterations of potential moves coming up and then decide their move based on the perceived best defense/offense in comparison to the other player's position. That doesn't exactly apply to the markets. In a broader sense, the market's next "move" is either up, down, or sideways. There isn't many reactionary "choices." There are varying degrees of each "move," however (example long trend up, short trend down, kind of sideways, etc.). As a programmer, I can see how to create a "good" chess playing program by simply brute-forcing potential moves. As a programmer, I cannot see how to create a "good" trading program by trying to brute-force potential market movements. You need to also realize that chess plays in a very limited parameter set (the board only has 64 squares, knights can only move this way, you can't deliberately put your king in check, etc.). The market has no such rules. Theoretically, it has no boundaries, and it can just as easily switch directions on itself as it can continue in the same direction.

Thats your opinion.Support and resistance is like a game of chess and all the others support the kings and resist the opponents.

Oh boy!

Oilfxpro

 
oilfxpro:
Hi

So developers will wait for 6 months to 2 years of forward tests before blowing their real accounts?

So they got time to waste?

they are better off scamming newbees indirectly

oilfxpro

Sorry to say but I have no idea what you are trying to say if this was directed to me.

Probability theory is very simple. The real problem is that calculating all permutations takes time and finding the combinations that really do have some correlation is quite problematic.

Forward testing (if you are talking about the genetic algo part of my idea) is not really needed if you separate the history to two sections where the other is the evaluation group and other is the validation group as is usually done in situations where there is a possibility of curve fitting.

By the way I already have some results for people interested, following are very simple correlation one rule combinations for couple of months of EURUSD minute bars, enjoy .

DISCLAIMER: the program still probably has bugs so the results may not be accurate!

The format is: {[rules], result, probability%, hits}

{[d1_down],d1_down,66.1904,33384},

{,d1_down,58.6867,39077},

{[true],d1_down,58.1938,78419},

{[m5_up],d1_down,58.1112,34181},

{[m5_down],d1_down,58.0894,33791},

{,d1_down,57.4442,36082},

{[d1_up],d1_down,52.1656,44652},

{,h1_down,51.8846,36082},

{[m5_down],h1_down,51.5019,33791},

{[d1_down],h1_down,51.1562,33384},

{[true],h1_down,49.7890,78419},

{[d1_up],h1_down,48.7794,44652},

{[m5_up],h1_up,48.0823,34181},

{,h1_up,47.8901,39077},

{[m5_up],h1_down,47.8160,34181},

{,h1_down,47.7621,39077},

{[d1_up],d1_up,47.3865,44652},

{[m5_down],m5_down,46.6929,33791},

{[d1_up],h1_up,46.5892,44652},

{,m5_down,46.5107,36082},

{[m5_up],m5_up,46.3532,34181},

{[true],h1_up,46.0564,78419},

{[d1_down],h1_up,45.3541,33384},

{,m5_up,45.3233,39077},

{[d1_down],m5_down,44.6441,33384},

{[m5_down],h1_up,44.5207,33791},

{,h1_up,44.2464,36082},

{[true],m5_down,43.5838,78419},

{[d1_up],m5_up,43.4673,44652},

{[true],m5_up,43.0929,78419},

{[d1_up],m5_down,42.7909,44652},

{[d1_down],m5_up,42.6432,33384},

{,d1_up,42.1678,36082},

{[m5_down],d1_up,41.4755,33791},

{[m5_up],d1_up,41.3505,34181},

{[true],d1_up,41.3191,78419},

{,m5_down,40.9448,39077},

{,m5_up,40.8154,36082},

{,d1_up,40.7324,39077},

{[m5_up],m5_down,40.5254,34181},

{[m5_down],m5_up,40.2859,33791},

{[d1_down],d1_up,33.2674,33384}

 

I read/hear the same. NN or EA's that use proabilities based on historical prices cannot be backtested, only forward tested as results are different each time.

OILFXPRO how are your EA's coming along? Have you breeding several different stratgies into one so that all market conditions can be traded? Testing EA's does take months you're right and if you want one that is profitable even longer, one that is consistently profible - who knows...one that is consistently profitable for all market conditions, almost endless?

There is a software program that is standalone and clculates trades from historical prices, the EA is not yet fully completed for it.

 
FseXy:
I read/hear the same. NN or EA's that use proabilities based on historical prices cannot be backtested, only forward tested as results are different each time.

OILFXPRO how are your EA's coming along? Have you breeding several different stratgies into one so that all market conditions can be traded? Testing EA's does take months you're right and if you want one that is profitable even longer, one that is consistently profible - who knows...one that is consistently profitable for all market conditions, almost endless?

There is a software program that is standalone and clculates trades from historical prices, the EA is not yet fully completed for it.

Fsexy

I need 1 month for designing the master EA and another month for testing.

Lots of probllems in combining different types of EAS EG they often duplicate similiar PRICE orders from each EA

It is nothing like the 2 pages EAS u get on this forum , more like 60 pages +

OILFXPRO

 

I just looked at some of your demo results from your EA. Will you keep us updated on the results no matter what? I am interested in following progress of trades. I see that some trades show no SL. Will you need to code EA so that SL is hidden from broker or not?

 
FseXy:
I just looked at some of your demo results from your EA. Will you keep us updated on the results no matter what? I am interested in following progress of trades. I see that some trades show no SL. Will you need to code EA so that SL is hidden from broker or not?

All my trades show stop losses , they don't show take profit on some of them due to advanced p/t take methods.My EAS use three different types of t/p and they are well advanced .

The new hedge manager will dynamically change stop losses , during the middle of trades, so following stop losses may be useless and too complicated to follow.

I am testing the early versions here

https://www.mql5.com/en/forum

The new EA will be far more advanced than anything you have ever seen.It has a hedge manager,exit manager,profit manager,entry signals manager and trade manager.

Most EAS are simple enties and exits but not the new EA FROM oilfxpro

OILFXPRO

 

interesting qoute from nn website

Many financial institutions use, or have experimented with, neural networks for market prediction, so it is likely that any trends predictable by neural techniques are already discounted by the market, and (unfortunately), unless you have a sophisticated understanding of that problem domain, you are unlikely to have any success there either!

 

neural networks

Hi,

do someone know some neural tools for free?

Thank you.

1...3456789101112131415
New comment