Machine learning in trading: theory, models, practice and algo-trading - page 1518
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
I was spinning something tonight. As an unquoted response, I think it would be interesting for others to read as well.
https://www.mql5.com/ru/blogs/post/728196
well this is a tc on "fuzzy" quotes, used to be very popular until the spreads were jacked up on those instruments
One of the real working TS, until you get banned2. Take a polynomial of degree 3 (3 free terms in total) and fit it to a piece of graph as long as a kilometer....A few free terms, usually not more than 3, describe any market curve.
I don't understand how with three degrees you can make a great fit? Such a polynomial has no more than two local extrema.
Well this is TS on "fuzzy" quotes that used to be very popular before the spreads on those instruments
One of the real working TCs, until you get bannedA little about the history of TS creation. TS was created as one of the theoretical ideas. At the time of creation it was not attached in any way to the research on any symbols. And on EURDKK it was launched by accident (using all symbols from Market Watch - MT5-tester mode). The TS has exactly three degrees of freedom.
Undoubtedly, the fluffiness is a fundamental regularity, which is exploited. Another thing is that the TS itself didn't take the fluffiness into account when writing it. It just so happens.
I would like to know out of curiosity, not as a stumbling block. Have any of the MO researchers and classical TSs run their algorithms on EURDKK? Find any similar fuzziness there? If not, why not?
And extremely interesting, how much are MO methods able to improve the result shown? I mean, I'd like to see with my own eyes how much better results MO specially applied to this pair can provide than randomly written TS the creation of which had nothing to do with EURDKK?
I don't understand how with three degrees you can fit magnificently. Such a polynomial has no more than two local extrema.
A little about the history of TC. TS was created as one of the theoretical ideas. At the time of its creation it was not attached in any way to the research on any symbols. And it was launched on EURDKK by accident. TC has just exactly three degrees of freedom.
Undoubtedly, fluffiness is a fundamental pattern, which is exploited. Another thing is that the TC itself did not take fluffiness into account when writing. It just so happens.
I would like to know out of curiosity, not as a stumbling block. Have any of the MO researchers and classical TSs run their algorithms on EURDKK? Find any similar fuzziness there? If not, why not?
And extremely interesting, how much are MO methods able to improve the result shown? I.e. I would like to see with my own eyes how much better results can be obtained with the MO specially applied to this pair than with some randomly written TS the creation of which has nothing to do with EURDKK?
1. I mean the time trend... 0, 1, 3, 4... polynomial trend. It fits perfectly and then it doesn't work. Just as an example that even 3 degrees of freedom can be a lot, over time the deviations increase real from the forecast.
2. I have not done, but the topic is interesting, for example Alexander, with whom you were snacking there, about this is what he does. But he converts any initial BP to, conventionally, "fluffy"
Try such a cast symbol EURGBP+GBPUSD-EURUSD, and open trades on EURGBP but in the opposite direction. The custom series is clearly "fluffier". I don't have time to experiment yet, but it seems to be appropriate
1. the time trend is meant... 0, 1, 3, 4... polynomial trend. It fits perfectly, and then it doesn't work. Just as an example that and 3 degrees of freedom can be a lot.
I don't get it. Still in the original question meant the TS with a certain number of degrees of freedom. Let it be three. So, in my opinion, you can not create a TS that fits any curve perfectly.
2. I did not do it, but the topic is interesting, for example Alexander, with whom you had a snack, about this does. But he transforms any initial BP to, conventionally, "fluffy".
Try such a cast symbol EURGBP+GBPUSD-EURUSD, and open trades on EURGBP but in the opposite direction. The custom series is clearly "fluffier". I don't have time to experiment yet, but it seems to be appropriate.
If you look at an expression on ticks, where each summand is a logarithm, it is one. Accordingly, such a cast symbol (especially considering commissions) will have Ask all the time higher than unity, and Bid lower. A classic arbitrage triangle.
I don't remember Alexander, much less the clues to him. But if he has spikes on the chart, it is the result of unsynchronized open bar price. For example, the open-price of one of the bars may be a dozen seconds later than the others. Hence the skewness. That is why we should use such diagrams based on ticks. And not by formulas in MT5.
From all of the above, to use this fluffy (and in fact it is not a fluffy, because Ask is up and Bid is down) for EURGBP openings has no basis, to put it mildly.
But it would be interesting to strip EURDKK through MO. Probably, the pros here should do very well. I have not prepared anything specially for this pair. If someone would like to make it without MI, it would be even more interesting.
I don't get it. Still in the original question meant the TS with a certain number of degrees of freedom. Let it be three. So, in my opinion, it is impossible to create a TS that will fit any curve perfectly.
If you look at an expression on ticks, where each summand is a logarithm, it is one. Accordingly, around the logarithm such a cast symbol (especially taking into account commissions) will have Ask all the time above unity and Bid all the time below. A classic arbitrage triangle.
I don't remember Alexander, much less the leads with him. But if he has spikes on the chart, it is the result of unsynchronized open bar price. For example, the open price of one of the bars may be a dozen seconds later than the others. Hence the skewness. That is why we should use such diagrams based on ticks. And not by formulas in MT5.
From all the above, to use this fluffy (and in fact it's not really a fluffy, because Ask is up and Bid is down) for EURGBP openings has no basis, to put it mildly.
But it would be interesting to strip EURDKK through MO. Probably, the pros here should do very well. I have not prepared anything specially for this pair. If someone wants to make it without MO, it will be even more interesting.
What is the difference in TS or curve fitting, the principle is the same, namely that even with 3 paramatrices it is possible to retrain easily and effortlessly on a large piece of history
Alexander is in the comments to my article and his topic is "From Theory to Practice". He did a tremendous job there on tick filtering and made stationary rows without outliers, for different syllables
It's not about the triangle, it's about the fact that the series will be fluffier and, as a consequence, a bit more predictable, although it will continue to correlate with EURGBP. I.e. to analyze the cast. series and trade EURGBP. Well, maybe that's a screwup.
example:
Well what is the difference in the TC or in the curve fitting, the principle is the same, namely that even with 3 paramaters can be retrained, easily and effortlessly, on a large piece of history
Strange, I see nothing in common.
Alexander is in the cams to my article who, and his topic is "From Theory to Practice"
I do not remember, but it does not matter.
It's not about the triangle, but about the fact that the series will be fluffier and, as a consequence, a bit more predictable, although it will continue to correlate with EURGBP. I.e. to analyze the cast. series and trade EURGBP. Well, maybe that's a crock.
It is. It's not even fluffy. The definition of fluffiness - the average Bid and Ask knee of ZigZag is very close to the minimum knee: Sum(ZZ[i])/Amount < k * min(ZZ[i]), where k is much less than two. The closer k is to one, the fluffier the symbol.
It's strange, I don't see anything in common.
Well for example the articlehttps://www.mql5.com/ru/articles/3795/63714#!tab=article
You can opt for just 2 parameters sigma and position and it's not a bad fit, though not perfect. The simplest MOS on fuzzy logic. The adjustment results exactly for the trend, i.e. the amount of trades begins to skew in some direction. And it is possible to fit it without any skew in the number of trades if we think about it. So the number of free parameters does not indicate anything (it can be adjusted with a small number), if the fundamental law is unknown. As I understand it, the question was exactly why a small number of parameters lead to fitting anyway. It turns out that easily and there is no contradiction.
In another thread I gave a comparison of SB and kotier through entropy. The forex majors showed pure SB, except for volatility clustering. I.e. essentially everything in forex is a fit.
For example the articlehttps://www.mql5.com/ru/articles/3795/63714#!tab=article
You can opt for just 2 parameters sigma and position, and it's not a bad fit, although not perfect. The simplest MOS on fuzzy logic. The adjustment results exactly for the trend, i.e. the amount of trades begins to skew in some direction. And it is possible to fit it without any skew in the number of trades if we think about it. So the number of free parameters does not indicate anything (it can be adjusted with a small number), if the fundamental law is unknown. As I understand it, the question was exactly why a small number of parameters lead to fitting anyway. It turns out that easily and there is no contradiction.
In another thread I gave a comparison of SB and kotier through entropy. The forex majors showed pure SB, except for volatility clustering. That is, essentially everything in forex is a fit.
For example: NBBBB-forex has shown pure Sberbank, except for volatility clustering, so everything in forex is actually a fitting.
Hello all!!! I want to ask a question to the experts. Suppose that my indicator receives data from the EA. How do I test this indicator in the tester? I get an error when i see that my EA has not loaded. Has anyone had problems with it? How have you solved this problem? Thank you in advance for your feedback.
If there is no source code, there probably is no way.