Having 86% of success rate, but need logic correction.

 

Ehlo,

I have created a network miner, which uses trading flow to find financial signs (trader[s] buy/sell signs). I have success of 83% confirmed and 91% not confirmed success rate. I need a [wo/]man, possibly developer or just an agent which will direct me in using trade technology according to my needs.

My needs: 15$ of income perday, on M15/4H. Using 100$ as start capital.Furher every gained 1k$ will increase lot count (0.01 at start)

What i have:

1) An financial flow s ign find er (based on neural networks (FANN)).  Which creates mathematical functions to extract signs on real-time data-flow.

2) An MT4 EA c# program that uses networks generated in goto 1

3) A super-pvp-met-owned development engine. (dont ask what is it, coz i then will need to kill you so noone knows)

4) A problem: 83% of trades are profitable ~0.35$, other 17% are very improvident  ~5-10$. need trader to get an idea how overcome this. 


meta data:
mathematical functions supported (remove trader_): 

Neural network params randomization (in alpha 0.0.1 version):

2) the function itself avg / STOCH / MOM / .... 3) the parameters of this function 4) the order of the function 5) the number of input parameters for the function (i.e. how many high / low / open prices to take ...), in the code it is InputDimension 6) activation function for the incoming layer of the neural network 7) activation function for the intermediate layers of the neural network 8) the number of neurons in the layers 9) additional network parameters that are based on the train algorithm. e.g. LearningRate, Percentage of Interconnected Neurons (Connection Rate), RPROP Step size, Weight Decoy, Temp (if it’s SARPROP. there is little to say about cooling or heating, the mathematics say is accurate. but on the Internet there is conflicting information about this, someone writes about heating, someone says about the initial test of this technology during cooling, namely the optimal distribution of atoms in the crystal lattice at the moment of cooling, that is, when heated, the opposite effect will be, even it’s approximately understandable to the fool who is different. but I can’t check it with my matan, sometimes I even thought that it’s for fun (fake? fucking ??) fakes. I even found combined temperature technologies, but that’s interesting. after all, I also began to try different technologies of the train in one era according to the logic of my own imagination). and a bunch of parameters that we can change, and what’s most important, we can get a really working network with a little change, while applying other trains or fundamentally changing the train will not get the data. V00t.


Programmers/Agents/Hate help Recognizer Enabled.

alpha researches: https://github.com/kilitary/fann-related

If you're interested, you will find beta.


-- Sent from umbrella core sector. Directorate approved. Drylock council awareness e d.

kilitary/fann-related
kilitary/fann-related
  • kilitary
  • github.com
FANN networks [+ forex tasks that was applied to it using this software] This is just command line tools for FANN. No gui, but many options and real-time switching not only temps/decoys or any other params (u think) affect this. +auto-tuning may train network starting using RPROP or QUICKPROP, then when hit ratio did not changes for example...
 

4) A problem: 83% of trades are profitable ~0.35$, other 17% are very improvident  ~5-10$. need trader to get an idea how overcome this. 

In summary, you have nothing but a losing system.

 
Alain Verleyen:

In summary, you have nothing but a losing system.

Yes, for current state. But this can be repaired by using plus aspects of it. The 83% success part.
 

Sergey Efimov:
Yes, for current state. But this can be repaired by using plus aspects of it. The 83% success part.

A typical losing system with high success rate is base on leaving the losing trades open and close the winning trades immediately, which also matches with your description of winning trades ~ $0.35 and losing trades at ~ $5-$10. Unless you find a way to maintain your 83% success rate while fixing the system so that winning trades covers losing trades. 

 
Sergey Efimov:
Yes, for current state. But this can be repaired by using plus aspects of it. The 83% success part.
I wish you all success.
 
Alain Verleyen:
I wish you all success.

thx

 
Meatshell123:

A typical losing system with high success rate is base on leaving the losing trades open and close the winning trades immediately, which also matches with your description of winning trades ~ $0.35 and losing trades at ~ $5-$10. Unless you find a way to maintain your 83% success rate while fixing the system so that winning trades covers losing trades. 

I read your threads. Need time to converge this. 

I think i have no knowledge of how to limit maximum loss on the trade. Currently different test periods can be good trained with optimization for example last month to predict current week. But i am still investigating this.

If i can find a smart trader ... at least i will give him a full programmers hands.