Market model: constant throughput - 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
How exactly did the addition take place?
I had to experiment with it. At first I took relative price changes. The archiver was compressed by 30%.
Then I tried absolute price changes (in pips) - the archiver compresses by 60%.
Further it turned out that if you locate data on financial instruments not by rows, but sequentially (first one, then the second one, etc.), then the compression increases by another 5%. I didn't experiment any further.
Pay attention to the graph of change of the sample window size from one financial instrument(AUDUSD) and its distribution.
Completely symmetrical and the variance is relatively low. If we compare with other charts where the sample contains more financial instruments, we can see that there are almost no spikes.
It is the first time I encounter it, usually on the contrary, the more financial instruments, the closer to stationarity. In our case it is vice versa, one financial instrument through elementary compression algorithms (no mathematics, we find the simplest patterns) gives excellent stability.
P.S. I measured the correlation between AUDUSD relative changes and AUDUSD compressed window size changes. The correlation is zero (~0.004).
I had to experiment on this one. At first I took relative price changes. The archiver compressed by 30%.
Then I tried absolute price changes (in pips) - the archiver compresses by 60%.
Further it turned out that if you locate data on financial instruments not by rows, but sequentially (first one, then the second one, etc.), then the compression increases by another 5%. I did not experiment any further.
Information is a set of bits that cannot be compressed in any way to transmit.
Achtung!
There's been a conceptual substitution!
Information is structured data. No matter how you compress the data, it does not become information as long as it does not inform you about anything.
That is, the amount of data in the window is 9 times larger for 9 symbols than for 1? And what will happen to one symbol when the window size is 9 times larger?
Yes, the amount of data grows linearly.
If I increase the window, then only for sampling from one financial instrument, because with large samples even with a constant window I had to compress gigabytes of data, which is very resource-intensive. I will try it for AUDUSD.
If I increase the window, then only for sampling from one financial instrument, because with large samples even with a constant window I had to compress gigabytes of data, which is very resource-intensive. I will try it for AUDUSD.
One more detail. There is indeed no reaction to the addition of gold. But there is an anomalous reaction to the addition of silver. What if you add gold first and silver later?
And of course joo is right, it would be nice to do this for random rows.
One more detail. There is indeed no reaction to the addition of gold. But there is an anomalous reaction to the addition of silver.
You can experiment a lot and it is not easy to do it alone.
With gold, I can guess what the reason is. As I take absolute price differences instead of relative ones in forming data for compression, gold, in contrast to all other Instruments, changes much stronger in points than the rest. Although the reasons may be different.
I did not see any anomaly in silver.
If I increase the window, then only for sampling from one financial instrument, because with large samples even with a constant window I had to compress gigabytes of data, which is very resource-intensive. I will try it for AUDUSD.
It is sampling from only one symbol. I change size of a sliding window (from 1 to 5 days). Charts with the same meaning:
Three-dimensional representation of compressed window size change when AUDUSD sliding window itself changes size:
The compression algorithms for zip and rar are not universal. The algorithms themselves also have certain window sizes.
The research is bold, pulled by thirty-kilometre ears. With an abstract approach, you can come up with so many hypotheses that even ten thousand people in fifty years will not study. Which is, by and large, often the case in science.
If you make it a rule to have at least the simplest justification for your hypothesis, you will save enough time for three lifetimes.