Trading Systems Leaders in this forum (automated trading systems) are winning more than 3000 pips in a month (30000$ investing one lot every time). Click here to register and get more information
I believe there is some instability in this BNNP predictor.
I ran 3 backtests on a daily euro chart. I took a very small period.
between 1june 2009 till today.
The test I run is the crossing of the BNNP with the smoothed BNNP.
I leave everything in default setting. But with only 1 fut. bar prediction.
When on the OPENING of a new bar the BNNP is above the BNNP smoothed on the next bar then so called go long on the bar that just opened.
When the opening of a new bar, the BNNP is bellow the BNNP smoothed on the next bar then go short on the bar that just opened.
I did this test 3 x times over this very short period whitout changing anything what so ever to the settings and I got 3 x times a total different result.
I had this problem already a week ago with a friend when we backtested some periods but did not pay any real attention to it ( because one does not think that if you do a test with a certain indicator that it would give opposite results).
Today I did again a test and because 1 particular bar that was a heavy loss made the overall result not looking that good.
By acccident I ran the test again and suddenly I saw that this losing bar turned into a winning bar.
So we ran the test again and started to look to the result over each single bar. We saw that the 3 test gave on each test a completaly different result. Even in this way that over 26 bars it gave 14 times a different result.
So I would suggest to be very carefull with whatever kind of backtest you do and specially with the results it would produce.
If somebody had a similar experience I would gladely hear about it.
Friendly regards....iGoR
__________________
Success comes with knowledge. Knowledge comes with experience. Experience comes with time and hard work... (iGoR) MT Intelligence - Stats for fxigor1
You need to fiddle with the bp-nn training settings to obtain *more* stable results.
You'll never get identical forecasts with a bpnn because of the way the network weights are randomised during the network's training phase.
Wow, thank you, very interesting stuff. But, like many codes relatedes to NN I put it in my charts and didn`t see what the author shows...
It seems that is needed to change something in code related with directories where are saved some informations, but is not clear for me
I apreciated any imputs...
Regards
I think Stochastic Net is more easy, I've studied it for more than 3 weeks
but I am still trying to find what's the best way to train it.
we depend on djama indicator or depend on other idea.
also I am trying to find what's the best filters we can use.
then we can let it work with any expert advisor which use a good strategy
I think Stochastic Net is more easy, I've studied it for more than 3 weeks
but I am still trying to find what's the best way to train it.
we depend on djama indicator or depend on other idea.
also I am trying to find what's the best filters we can use.
then we can let it work with any expert advisor which use a good strategy
Nice one
I have see that here is the indicator that you talk about:
But... I really found very interesting that article called Recipes for Neuronets. It seems that is needed to use Visual Studio to compile that DLL called Better... I really didn`t uderstant nothing about this softwares but it seems that I will need to learn. If I discovered how to solve that things I will post here.
But... I really found very interesting that article called Recipes for Neuronets. It seems that is needed to use Visual Studio to compile that DLL called Better... I really didn`t uderstant nothing about this softwares but it seems that I will need to learn. If I discovered how to solve that things I will post here.
Bob
I am learning how to compile files by Visual Studio too and I am studying Stochastic NN
I am using this Indicator to Know It's behavior with the good trend and bad trend
for example : I am using this code when I plan to go short
Code:
int BuildT(int i) // زٍَ َهى ًي ٍهٍ هًهَهييٌٍ
{ // هًهًنييٌٍ
int T= 0;
double Previous = djama(n,p,q,i+1);
double Current = djama(n,p,q,i);
double Next = djama(n,p,q,i-1);
if (Current <= Next) T = -1;
if (Current > Next) T = 1;
return (T);
and this code when I am planing to go long
Code:
int BuildT(int i) // زٍَ َهى ًي ٍهٍ هًهَهييٌٍ
{ // هًهًنييٌٍ
int T= 0;
double Previous = djama(n,p,q,i+1);
double Current = djama(n,p,q,i);
double Next = djama(n,p,q,i-1);
if (Current < Next) T = -1;
if (Current >= Next) T = 1;
return (T);
}
Then I see the result
I control by (lastBar) from parameter
Linear Mode
