Discussion of article "Evaluating the ability of Fractal index and Hurst exponent to predict financial time series" - page 3
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
It is quite interesting that the topic and the article has generated discussions. However, it is true that among them the business and substantive part is smaller. To make the discussion more useful and businesslike, I suggest to get acquainted with its basis - the thesis of Starchenko N. C. FRACTALITY INDEX AND LOCAL ANALYSIS OF CHAOTIC TIME SERIES. Since Starchenko himself works and his algorithm is successfully used in CJSC "Intrust Financial Company", you should get acquainted with the work and think about it, well, and learn how to pass through doors. The thesis discusses "coloured noise", applications in econometrics, discordance and confident prediction in more detail.
Dmitriy Skub writes that the fractality index is incorrectly counted in the code. I would like to know more about what the error is in order to fix it (or better yet, a test dataset and the index value for it). Dmitriy's rating and experience is respectable and trustworthy to discuss such things with him.
When switching TFs we have the following:
2019.06.01 08:31:16.705 Fractal Index (Si Splice,M5) zero divide in 'SimpleStat.mqh' (95,34)
2019.06.01 08:32:06.808 Fractal Index (Si Splice,H1) array out of range in 'CFractalSeriesSet.mqh' (110,16)
2019.06.02 11:39:33.371 Fractal Index (Si Splice,M1) zero divide in 'SimpleStat.mqh' (95,34)
Screenshots of MetaTrader trading platform
Si Splice, H1, 2019.06.02
JSC ''Otkritie Broker'', MetaTrader 5, Real
And about the algorithm itself - I compare the picture with my indicator. There is little in common. Mine is tested - it counts correctly.
Here is the comparison for window 64. Perhaps I am looking at something wrong or not taking into account when comparing - then correct me.
Dmitriy Skub writes that the fractality index is incorrectly calculated in the code. I would like to know in detail what the error is in order to correct it (or better yet, a test dataset and the index value for it). Dmitriy's rating and experience is respectable and trustworthy to discuss such things with him.
The size of the zipped file is 32M - there are four such files. It does not fit here. Please write where to upload them.
The size of the zipped file is 32M - there are four such files. It doesn't fit here. Please write where you want to upload them.
***
One more question. In the provided picture the bottom one of the three looks like my indicator, but Noise Level is bound to amplitude 1. This is strange, it should be equal to 0.5 and the purple colour corresponds > 0.5. I am also quite surprised as the shape of the graph is similar but the amplitude is shifted.
***
Another question. In the picture provided, the bottom one of the three looks like my indicator, but the Noise Level is tied to an amplitude of 1. This is strange, it should be equal to 0.5 and the purple colour corresponds > 0.5. I am quite surprised too, as the shape of the graph is similar but the amplitude is shifted.
It is quite interesting that the topic and the article has generated discussions. However, it is true that among them the business and substantive part is smaller. To make the discussion more useful and businesslike, I suggest to get acquainted with its basis - the thesis of Starchenko N. C. FRACTALITY INDEX AND LOCAL ANALYSIS OF CHAOTIC TIME SERIES. Since Starchenko himself works and his algorithm is successfully used in CJSC "Intrust Financial Company", you should get acquainted with the work and think about it, well, and learn how to pass through doors. The thesis discusses "coloured noise", applications in econometrics, discordance and confident prediction in more detail.
The overall impression of both texts is good - written at a good level and in clear language.
At first glance, the statistical test for the consistency of real prices with the efficient market hypothesis (in the thesis) does not seem quite correct.
The source code is obviously not the latest version: In addition to the above division by 0 and exceeding the array size, SimpleStat also has division by 0 in lines 105,191.
***
Another question. In the picture provided, the bottom one of the three looks like my indicator, but the Noise Level is tied to an amplitude of 1. This is strange, it should be equal to 0.5 and the purple colour corresponds > 0.5. I am also quite surprised as the shape of the graph is similar but the amplitude is shifted.
The source code is obviously not the latest version: In addition to the above division by 0 and exceeding the array size, SimpleStat also has division by 0 in lines 105,191.
Please specify your settings when division by 0 occurs, so that I can correct the length of the arrays immediately. I didn't consider the M1 time because I don't rely too much on fractality statistics for such intervals, but I will debug it and save the programme from arithmetic problems.
The following was curious to me. When I ran the programme to check the fractality index indication, I did not see the desired correspondence of "trends" or "reversals" to the index on FOREX. That's why I ran the programme on MICEX for "blue" chips with a daily period and there I observed the consistency of charts and index. I don't know how stable this pattern is, but one should be very careful when switching to high frequencies.