HOW TO TRADE FOR NOTHING ? - page 7

New comment
 
Swetten:


)))) hilarious..... to the point !

that's how we trade ))) everyone is the smartest ... but everyone is in the red )))))))))

 
We live in crowds, we work with crowds, we trade with and with crowds. And how do we avoid being influenced by it?
 
Richie:

You guys fight amongst yourselves while I bathe. Sveta will help you with the hard work. Otherwise, the heat will be over soon.


well let's just say - that i don't mean to offend ... not to quarrel with anyone and i don't intend to ... just interested in facts and getting to the truth.... hence the questions... but as I understand it's just words from books again....

 
The typical trader for some reason does not think about trading against or with the crowd beyond looking at the total volume of open positions, the typical trader somehow thinks it is a crowd position.
 

Here's the crowd, here's the secret of the market, notice how many "think differently" people there are!

)))

 
Richie:

The truth is in there somewhere :))))

https://www.mql5.com/go?link=https://www.youtube.com/watch?v=JDZBgHBHQT8&feature=related


Too bad...thought I might learn something new...although training on a sinusoid (or its likeness) is doomed to failure from the start...imho all things considered... good luck...
 
Vizard:


Yes... I agree there is less of a regularity here... I can try it... but I need the signal itself (line) in csf or txt file... format = 1 column time, second column signal... + a line a couple of times longer.... it's too short....


It's a shame this thread has turned into a flood. Please do, I'm just wondering if the NS can get anything out of it.

I have written values into file, first column time, second column value. here is original function

data file attached

now i've used matcad to read it and plot the spectrum, i.e. i don't know anything about what's in the input

I can see that the spectrum has these components and I can pull them out. can the neural network handle it... I'm just wondering.

Files:
data.rar  2 kb
 
Prival:


It's a shame this thread has turned into a flood. Please do, for me I'm just wondering if NS can get anything out of it.

I have written values into the file, the first column time, the second values. here is the original function

data file attached.

now using matcad to read it and plot the spectrum, i.e. i don't know anything about what's in the input

I can see that the spectrum has these components and I can pull them out. can the neural network handle it... I'm just wondering

yes I want to be constructive too... the author asked a question - he's tried to get an answer... asking a question with the intention to learn something new (and not with a sneer, as he probably thought) we get a sci-fi show instead of an answer....

I'm going to rework the format a bit and try to predict...

 

Here's how I did it.

blue is the original signal without noise, red is reconstructed from a mixture of signal + noise. The shift upwards is due to noise property, the function rnd(20) will gerenethe random number in the range 1...20, i.e. in addition to the random component, a constant component of 10 has also been added. This can also be seen in the picture.

mean(rnd(20))=10

 

much worse with noise - and the scatter plot confirms it, and so does the forecast itself (55 points ahead)... if we smooth it out and then feed it into the network, of course it will be better....

but here again we have cyclicality in the data ... if it is removed, the grid and any predictive model can not properly forecast.... (about forecasts for the future)

your approach is different - removing noise ... it's great to remove noise - and what goes up and down is not important - we need the line itself, or rather the peaks and troughs (the most informative part for decision making) ...

Maybe we should just try it on real quotes ? and then compare it with MA for instance... which one is better.... try it for your interest....

https://www.mql4.com/go?http://i065.radikal.ru/1007/f9/a83289a5b256.jpg
https://www.mql4.com/go?http://s001.radikal.ru/i193/1007/60/96290849aa4b.jpg

2.If you set the first signal (i.e. the one without noise) as a teacher for a network, it will certainly learn and find points in inputs (a signal with noise)... but it won't help - if the data changes (inputs), it will only get worse later...

...if you train on every bar, it's better - but there's not much practical use ... just to speed up an indicator (predict), not more than 100% and not 100% accurate ... far from it... 100%....

123456789
New comment