Market model: constant throughput - page 6

 
hrenfx:

With gold, I can guess the reason. Since I took absolute rather than relative price differences in the compressed data, gold, unlike all other FIs, changes much more in points than other FIs. Although the reasons may be different.

I didn't see the anomaly of silver.

I was referring to the absolute magnitude of the reaction to its addition. Although perhaps "anomalous" is too strong a term :)

In any case, this experience will shed light on the property of the transformation called commutativity. That is, whether the proposed transformation has it or not.

The impact of scaling would need to be clarified, of course.


As for the participation of someone else in the experiments, this is unlikely at this stage, imho. Because the practical implications are really quite unclear.

The way of using it for trading, mentioned in the first post, doesn't look promising. As I understand it, it implicitly assumes that the result for 50 minutes will fluctuate, while for 1 hour it will be significantly more stable. Quite an exotic assumption, imho.

What's left at this point? Using correlation as an alternative?


The topic is interesting, but as long as the research looks purely academic we should not expect mass participation.

 
sanyooooook:
I say again, how is it supposed to be used in practice?


Nerd . What about the art? In general, it seemed to me that this work is not a study of the market using compression algorithms, but a study of the properties of compression algorithms using in this case

market quotes .

 
Mischek:


Nerd. What about art? In general, it seemed to me that this work is not a study of the market using compression algorithms, but a study of the properties of compression algorithms using in this case

market quotes .

The author himself doesn't seem to know what his idea is, if he doesn't answer or doesn't consider it necessary to explain to a simple layman the high point of his thought.
 

Conducted a study on random BPs with a normal distribution of increments. To more or less repeat the result with real price BPs, I took the same amount of pseudo financial instruments (9), the same amount of data and RMS of real financial instruments ("AUDUSD, EURUSD, GBPUSD, USDCHF, USDJPY, USDCAD, NZDUSD, SILVER, GOLD" = {51, 56, 67, 48, 39, 49, 42, 2, 73}). Result:

Files:
4analyserand.rar  423 kb
 

As we can see, random BPs are shrinking more and more with the addition of new BPs. "Gold" has shown itself to be real.

Why random BPs are compressed (contain general information) is not clear from information theory. At the level of compression algorithms the process is more or less clear.

 
Could it be the properties of a random number generator?
 
hrenfx:

As always, it's all negativity. Some in the open and some in the guise of caution.

Again, nothing on topic, just phrases that might as well be inserted into any thread at all.

I repeat, how is it supposed to be used in practice?
 
sanyooooook, don't be in a hurry. It's the wrong stage. The consequences can be most unexpected. I don't think the author has anything sensible to say on the subject yet either.
 
Mathemat:
sanyooooook, don't be in a hurry. It's the wrong stage. The consequences can be most unexpected. I don't think the author has anything sensible to say on the subject yet either.
Can you? Presumably.
 
hrenfx:

As we can see, random BPs are shrinking more and more with the addition of new BPs. "Gold" has shown itself to be real.

Why random BPs are compressed (contain general information) is not clear from information theory. At the level of compression algorithms the process is more or less clear.

Yes, it is strange. I was expecting the opposite effect - the more random the data, the less compressible it should be.