Don't tell me then that TA doesn't work - page 23
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
Uh-huh. This is the result for signal addition (trigger=0)
And here is for logical signal multiplication (trigger=2)
Both results, all other things being equal (pair, timeframe, optimisation periods etc). Same 11 years.
OK, I'll think about it tomorrow, it's time for bed.
This is just what I noticed when optimising your EA:
- I left pass = 1 by mistake, and set perceptron parameters 0 for optimization.
In this case perceptron parameters 0 should not be calculated, but they were calculated, i.e. control was passed to the function perceptron0().....
I am pleased that the duration of the profitable OOS is over 7 years, however it is strained that almost all of it is backwards on course.What's at the back of the course is not a strain. The point is that even assuming we are dealing with a stable prediction of the past, i.e. some sort of time machine that can take us back and allow us to trade profitably there, there is a solution. With conventional perceptrons on the difference of opening prices at inputs we can manipulate them any way we want: put the cart both behind the horse and in front. I.e. if a perceptron can confidently predict the past, the weights can be recalculated so that it will "predict" the future. Graphs can also be flipped around the vertical axis, i.e. time can be reversed.
I.e. we need any kind of time machine, regardless of where it takes us: forward or backward. What matters is the profitable result in the very place in terms of stability, where we get. The rest is not a problem - elementary mathematics. The Perceptron is a typical linear inequality.
What's at the back of the course is not a strain. The point is that even assuming we are dealing with a stable prediction of the past, i.e. some sort of time machine that can take us back and allow us to trade profitably there, there is a solution. With conventional perceptrons on the difference of opening prices at inputs we can manipulate them any way we want: put the cart both behind the horse and in front. I.e. if a perceptron can confidently predict the past, the weights can be recalculated so that it will "predict" the future. Graphs can also be flipped around the vertical axis, i.e. time can be reversed.
I.e. we need any kind of time machine, no matter where it takes us: forward or backward. What matters is a profitable result in terms of stability, where we get to. The rest is not a problem - elementary mathematics. The Perceptron is a typical linear inequality.
Because your unsubstantiation looks even worse than Reshetov's. And anyway, it seems to me that it is incorrect to demand that he prove something to you.
In the post I wrote about the p-value. It's the first azz of mathematical statistics.
The man put forward the idea, and even illustrated it technically, is not enough for you to roll up your sleeves (if you consider it promising)? You may even ask him for seed money. ;-)
I would like to remind you about the authors of ideas that 100 pundits can't comment on.
Fill in the gaps. Or at least try to.
Reshetov claims that his system is a proof of TA. This he proves with TC. But Reshetov is not the first - such proof has been given for about 400 years, starting with the Japanese with their candlesticks.
By the way, the lack of proof of working or proof of non-working TA does not detract from the fact that it is possible to develop TA based on TA and make TA a source of income. TA is an art, as in any other kind of art there are folk artists for a huge crowd of losers nobody knows about.
I understand Reshetov's TA is based on NS. If so, this is significant, as the success of applying NS depends entirely on the person teaching NS. Reshetov has succeeded, maybe he is a genius, maybe he fell out of his tree, maybe he drank a lot of beer - we don't care - his skill will not transfer to us. All TA is like that. TA is unproven in principle.
Conclusion.
Maybe you can clarify one observation of mine. This forum (and others as well) discusses TA, some exotic things like fractals etc., but never discusses the application of econometrics and its sister, matstatistics, to TC. Please note that the word "econometrics" is grammatically incorrect in this forum.
At a glance, I can recall a discussion of the regression equation, which was quickly shot down to formulas for calculating regression coefficients - no discussion of the application of regression in TC ever came to fruition. Is the ignoring of statistics an accident? Or is it thanks to Reshetov and co.
:)
there has never been any discussion of the use of econometrics and its sister, matstatistics, in TC
Offhand
https://www.mql5.com/ru/forum/105771
https://www.mql5.com/ru/forum/105740
https://www.mql5.com/ru/articles/222
Just off the top of my head
https://www.mql5.com/ru/forum/105771
Not relevant to econometrics as it does not have a clearly stated model.
https://www.mql5.com/ru/forum/105740
A new supposedly market model has been proposed. Proposed by a DSP specialist, which I am deeply convinced is not applicable to the market. There has been no actual study of this model on the forum.
https://www.mql5.com/ru/articles/222
Pardon me, I agree. If you run a search on "econometrics" it doesn't find it. The article is very recent and it's an article, not a forum. I have to insist that I am correct as of January 1, 2011. A discussion of this piece or similar would seem to me to be quite interesting. At any rate, specific algorithms and figures would have been discussed, rather than the skills of individual, albeit genius, personalities.
Commenting on the article, I'm generating a lot of different interests and suggestions. First, the author used his own programs, but there is Eviews and most importantly Matlab. If we take these packages, we get a more systematic view of the problem.
Thank you for the last link, it was quite sad from the ignoramuses, ignoramuses and DSP and NS specialists.
Colleagues, adapting model parameters with anything is an old and correct idea, I, for example, use Bayesian networks + a couple more ideas are now in testing, you can adapt it with tambourine dancing and ritual offerings. That's not the question. Check for randomness of new profitability graphs obtained. The naked eye can see obvious problems. In fact there is no reason to rejoice.
Martingeil:
What's at the back of the course is not a strain.
...
The rest is not a problem - elementary mathematics.
I don't believe it's spring in Tashkent - it's snowy and cold here.
That it's cold in Almaty, I believe - it's February outside.
As for maths, it is not a religion to believe in:
Suppose we have four adjacent sections of history in order from the past to the future: A, B, C, D
If the signals from A, B and C are added up so that D = A + B + C, then the signal is uncertain on D.
We get confident trade signals on section A by simple summing up signals on the other three sections: A = B + C + D
But we don't need section A - it is the past, while the future can be obtained only on section D, if we know the signals on A, B and C.
Then from the above formula we get: D = A - B - C
Let's carry out an experiment on a trading system which is based on forecasting of the future direction of quotes movement by
fitting weight coefficients of an elementary one-layer neuron network - perceptron on historical data. The principle of this trading system was described in details in my article "How to find a trading system". Let's take historical data for the EURUSD currency pair for 9 previous months or more on H1 timeframe chart. We will divide it into three independent sections of three months each. The first one will be used for the final test, while the other two will be used to fit the history. To avoid running the trading system separately, I immediately combined two perceptrons in one trading system.
And I created the function Supervisor() with the help of which the trading system has three modes of operation, depending on the input pass parameter:
1 - fitting and testing of the first perceptron,
2 - fitting and testing of the second perceptron,
3 - filtering by sifting out inconsistent readings of both perceptrons in the testing mode without optimization or in the auto-trading mode on a demo or real deposit.
The weight coefficients of perceptrons: x11, x12 ... x42, as well as p and sl are fitted to the historical data. The input parameter sl is a constant for all sections. Stop Loss and Take Profit levels are set for all open positions by this very value. Another input parameter p - the lag time period for the opening price difference, is also a constant. Entry into the market is made at the beginning of a new bar formation, i.e. according to the bar opening prices and perceptron readings, depending on the value of the pass input parameter, and exit only upon triggering of a Stop Loss or Take Profit. Optimization is performed using the genetic algorithm for identifying extrema and the balance maximum is taken as an extremum. The initial amount should be very large, e.g. $1000000, to prevent the algorithm from getting stuck on margin calls during optimization. Input parameters: lots - volume of open positions in lots and mn - unique magic number, so that the EA will not confuse handling of its own (which it has opened) orders with orders of others (which it has not opened).
At thefirst step , we need to find out what the values of the input variables p and sl should be. To do that, we select the last two parts of the history, i.e. from 6 months ago till today. We set all parameters of perceptron weights to values from Start = 0 to Stop = 200 in steps of 1. Set the p-value from Start = 3 to Stop =100 in steps of 1, the sl-value from Start = 100 to Stop =1000 in steps of 10 (or from 10 to 100 in steps of 1 for four-digit quotes). Set the pass value to 1. Tick the following parameters to be optimized: x11, x21, x31, x41, p and sl. All other checkboxes should be disabled. Activate the optimization. Once the fit is complete, set the input parameters to the best pass.
Second step. Fitting the weights of the first perceptron on the second section of historical data. We set the optimization date and time from 6 months ago to 3 months ago. Uncheck the optimized parameters only from the input variables p and sl. Run the optimization. Once the optimization is completed, set the input parameters according to the best fit.
Third step. Fitting the weights of the second perceptron on the third section of historical data. We set the date and time of optimization from 3 months ago to the present day. Uncheck the optimized parameters: x11, x21, x31, x41 and set them for x12, x22, x32 and x42. The other checkboxes must be unchecked. Set input variable pass to 2. Start Optimization. Once the optimization is completed, set the input parameters by the best pass.
That's all, our trading system has been adjusted to the historical data from 6 months ago up to the present day. Let's save the values of the input parameters in the settings file. Set the pass input variable to 3. Uncheck the "Use date. Start the test. We look at the testing chart. We can see that the balance and equity curve tends upwards in the right part of the chart and tends downwards in the left part. Now we must make sure that the balance tends upwards at the area outside the adjustment sample. We bring the cursor of our mouse to the balance line, where the rise of profit begins and look at the date in the tooltip. It turns out that the balance curve tended upward almost nine months ago, counting from today, excluding 10 days, i.e. 8 months and 20 days. And the adjustment was carried out on a stretch of 6 months. Hence, there is a successful test outside the optimised sample. We highlight this out-of-sample area to analyze it in more detail. On the whole the results are quite satisfactory, although significantly inferior to the record of J. Soros, but superior to that of W. Niederhoffer.
In order to make sure that we have dealt with fitting in some parts of the history, it is necessary and sufficient to uncheck the "Use date" checkbox. And run the test of the Expert Advisor with values 1 and 2 through the whole available history. In each of these modes, we can see that the upward growth of the balance curve is observed only within those periods, on which certain perceptrons were fitted. For all other periods of history there is no positive trend, except for individual humps ending in troughs.
As we have seen, despite the fact that both perceptrons did not pass the forward tests outside the optimized sample of historical data, nevertheless the filter of their joint signals, gave positive results on historical data of which nothing was known at the time of fitting. You can also experiment with other trading systems, such as those based on the breakdown of simple moving averages or on more advanced multilayer neural networks. If the trading system is robust, it is more likely to produce positive results on filtered trading signals outside the optimization period. If it is not robust, it will not give positive results on the optimized period with the filter enabled. However, the robustness of the TS is secondary compared to overhead costs of spread, swap and broker's commission. Therefore, with significant overhead, one can only dream of positive results in forward testing, because the expected payoff will be obviously negative.