Piligrimus is a neural network indicator. - page 4
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
You'd better torture the expert in different positions))
For example here is a test without any tricks, with one fixed lot, starting balance 10 000, the currency is the same:
the test is from 2008 to this dayH4 GBPUSD
Gentlemen, and ladies, if there are any! Let's define ourselves once again, so that there is no useless demagogy and off-topic reproaches. I am ready to listen to constructive criticism, and in accordance with it make adjustments to the development to improve the result. But I want to reiterate that I haven't offered for discussion the final product, but only the first model, which is the starting point for further development of the indicator according to this strategy. From my experience I suppose something interesting may come out of it. But this process is creative, I don't know what may really be the results. Maybe I am mistaken in some way, time will show, be patient.
To begemot61 23.04.2009 22:45 : The disadvantages of MA I think you are well aware of. First of all, by training networks with embedded prediction goals, I want to achieve substantially lower lag than for AI, with a similar level of smoothing. Although that is not the only objective, this indicator will also have other features.
I cannot optimize anything at this stage. I am generally against optimization. More often than not, it leads to fitting, it's hard to track its timeliness, it wastes a lot of time, the system loses autonomy, etc. I tend to the idea that the indicator should correct its parameters during its operation according to the changes on the market, in the future I will introduce such elements of self-regulation into this indicator.
To Neutron 24.04.2009 05:11 : Last - do you mean the zero bar? If so, I have already mentioned it above. And since High, Low and Close on the zero bar are changing, the indicator calculation results on the zero bar should change accordingly. When a new bar comes, the result of calculation on the last tick is fixed. If you do not like it, enter the conditions for indicator calculation only on the formed bars.
You'd better torture the expert in different positions))
Here is a test without any tricks, with one fixed lot, starting balance 10 000, the currency is the same:
test from 2008 to dateH4 GBPUSD
>> So far, the indicator is too crude to build trading systems on it.
Continuing to work on the indicator, introduced another module, although deviating somewhat from the original idea, a funny picture emerged. The pink signal corresponds to the first module presented earlier. The possible trading logic of this model is clear: if red is static then Sell until blue and pink turn, if blue is static then Buy until red and pink turn. But this is still a rough experiment, I'll see what will result from the introduction of one more module.
For comparison, replace Pilligrimus with MA and optimise its period, all other parameters as they were. Will the better result after optimisation be better/different than with Pilligrimus?
The result is as follows:
>> more false entries and market entry delays very much comparing with Pilligrimus.
you should have seen the advisor.)
The indicator is still too crude to make trading systems on it.
Just wanted to show the importance of your development, that even with a normal MA your indicator gives a good prediction...
Continuing to work on the indicator, introduced another module, although deviating somewhat from the original idea, a funny picture emerged. The pink signal corresponds to the first module presented earlier. The possible trading logic of this model is clear: if red is static then Sell until blue and pink turn, if blue is static then Buy until red and pink turn. But it is a crude experiment so far, I will see what may be caused by one more module.
What other modules are you adding?
I just wanted to show the importance of your development, that even with an ordinary MA your indicator gives a good prediction...
I am not against such experiments, on the contrary, thank you very much, I am developing it further and I don't have time to check it. I just said that this indicator has a lot more potential, and if I manage to do it, it will be really good indicator.
To nord: I am going to make the next module to compensate for the error, to reduce the false positives.
I have added a couple of signals in the second module, now there are more accurate signals for closing positions.
I'm displaying the indicator with the second module. The signals have not been finalised yet, if you are interested, you can experiment. Please publish the results in this topic. A constructive criticism of the finalization is desirable, please refrain from idle demagogy. The work is still raw and it is too early to make final conclusions.
Setting up the indicator with the second module.
Thank you, let's test it.