Welcome to Forex-TSD!, one of the largest Forex forums worldwide, where you will be able to find the most complete and reliable Forex information imaginable.
From the list below, select the forum that you want to visit and register to post, as many times you want. It’s absolutely free. Click here for registering on Forex-TSD.
Exclusive Forum
The Exclusive Forum is the only paid section. Once you subscribe, you will get free access to real cutting-edge Trading Systems (automated and not), Indicators, Signals, Articles, etc., that will help and guide you, in ways that you could only imagine, with your Forex trading.
Elite Section
Get access to private discussions, specialized support, indicators and trading systems reported every week.
Advanced Elite Section
For professional traders, trading system developers and any other member who may need to use and/or convert, the most cutting-edge exclusive indicators and trading systems for MT4 and MT5.
I dont think you are understanding... The output of the NN is not going to change as long as the NN weights do not change so no matter if EUR/USD is uptrending/downtrending/horizontal the indicator will still be able to give Buy/Sell signals.. But the actuall buy and sell levels will change based on what the market is doing "more".
For example the EUR/USD market has been in a strong uptrend since 2001 that is why default buy level is closer to 50 then the sell level, but for the last couple of months the market has been horizontal. So the Buy/Sell levels would be more even'd out.. Test this youself, re-optimize the buy/sell levels for the more recent "horizontal market". You will see how the buy/sell levels will change based on market trend.
If you optimize for too long of a period you will see more random results like you have seen.. Because you have optimized from jan07 to present how many times have the market changed? Look at the eur/usd 1day since jan07.. How many times has market changed? i can count more then 4. If you optimize too long you will be over-generalizing. If you optimize for just the last month you might be curve fitted to just that months market. The idea is to find just the right amount of timeframe to optimize for.. Not too much, not too little. This is why i said 3 months.. you will then see what im explaining..
BTW i can guarantee the weights/logic (line for line) of the bogie indicator and indicator from the cortex site are the same..
Infact i can guarantee the entry logic from bogie and the open source EA are the same..
With respect, I do understand the indicators workings. And I also agree with what you are saying! My main point was that all the indicators are using weights that were calculated in 2003 - are you telling me that these weightings are as relevant now as they were then?
FWIW, I have registered Cortex, so can now save the NN file. Unfortunately, this is not a simple list of weights as seen in the indicator, in fact the miriad of neuron weights bear little resemblance to what one might expect they 'should' be. (at least I haven't figured it out yet!) The other annoyance is that Cortex itself is limited in the type of 'lag' file configurations it can generate, and the only way to aquire the required flexability is through scripting - but this means that you must pay again to enable this feature! My advice to anyone thinking of registering Cortex - don't, it simpy allows saving the inflexible NN file! Instead, for the same price, register the scripting feature, you then can access all of Cortex's features...
With respect, I do understand the indicators workings. And I also agree with what you are saying! My main point was that all the indicators are using weights that were calculated in 2003 - are you telling me that these weightings are as relevant now as they were then?
FWIW, I have registered Cortex, so can now save the NN file. Unfortunately, this is not a simple list of weights as seen in the indicator, in fact the miriad of neuron weights bear little resemblance to what one might expect they 'should' be. (at least I haven't figured it out yet!) The other annoyance is that Cortex itself is limited in the type of 'lag' file configurations it can generate, and the only way to aquire the required flexability is through scripting - but this means that you must pay again to enable this feature! My advice to anyone thinking of registering Cortex - don't, it simpy allows saving the inflexible NN file! Instead, for the same price, register the scripting feature, you then can access all of Cortex's features...
Honestly i do not know if the weights are as efficient in todays market as they were in 2003.. I am assuming they are not. I guess only an update set would tell.. but it is a fact the current set of weights are still able to produce profitable results..
There are some many links. Which link you are referring pls.
With this link you purchase registration details for Cortex (you click the 'Buy Now' button and pay via Paypal). If you want to buy the Scripting features of Cortex, you would select that option from the Order page instead...
I have registered scripting and Cortex, but i haven't good results. First in samples have errors - replace all MVMIN, MVMAX with MV_MIN, MV_MAX.
For me usefull are script 4b, 5a and 5b, but i dont't know how author is generated these weghts.
I have registered scripting and Cortex, but i haven't good results. First in samples have errors - replace all MVMIN, MVMAX with MV_MIN, MV_MAX.
For me usefull are script 4b, 5a and 5b, but i dont't know how author is generated these weghts.
Yep, I've hit the same impasse!
Having spent several hours looking for 'bugs', I've reached the conclusion that the 'working' weights were generated with something else, (or some other code) as the code as listed works fine (apart from the error you pointed out) but the weights are a mess! In fact, the more you train the network, the larger individual weights become - something is definitely wrong!
I could run the batch file with no problem (see pic)
I’m testing only with fw 10
The image shows 100% acuracy in classification, this cause doubt, there is no solution that will give 100% acuracy, there must be some thing wrong.
To barnix, I have a question:
In your code for "5_bar_SVM_input.mq4":
Code:
string Line1(
int nb1,
int FW,
int Price1,
bool train,
int pr1,
int i
)
{
int sig0;
string ss0;
string ssig0;
if(train)
{
sig0 = iCustom(Symbol(),0,"5_bar_sig",
nb1,
FW,
0,
i+0);
}
else
{
sig0 = 0;
}
string x1=DoubleToStr1(iMACD(Symbol(),0,FastPeriod,SlowPeriod,SignalPeriod,Price1,MODE_MAIN,i+0)/Point,pr1);
string x2=DoubleToStr1(iMACD(Symbol(),0,FastPeriod,SlowPeriod,SignalPeriod,Price1,MODE_MAIN,i+1)/Point,pr1);
string x3=DoubleToStr1(iMACD(Symbol(),0,FastPeriod,SlowPeriod,SignalPeriod,Price1,MODE_SIGNAL,i+0)/Point,pr1);
string x4=DoubleToStr1(iMACD(Symbol(),0,FastPeriod,SlowPeriod,SignalPeriod,Price1,MODE_SIGNAL,i+1)/Point,pr1);
you are having a signal on bar "i" using iCustom, then you are registering a feature (which is MACD) calculated at bar "i" too.
In my openion, you could not use a feature value at the bar you want to predict the signal at, you should calculate iCustom for "i-1".
In Cortex have something wrong.
I train with my own backpropagation code this indicator with data from 2007 to today. May be have forecast for one bar.
Linear Mode
