Absolute courses - page 38

 
Dr.F.:

I can do it myself. The topic is interesting, about the distinction between quotes and hspc. My algorithm seems to be able to do it, although it wasn't planned to do so.
Take the real series. Calculate by your algorithm for the first 144 samples. Then for next and so on. See how the QC (distribution) changes. Do the same for the random walk series. Calmly, no tricks))
 
Avals:
take the real rows. Calculate your algorithm for the first 144 counts. Then for the next and so on. See how the QC (distribution) changes. Do the same for the random walk series. Calmly, no tricks))


That's exactly what I'm going to do and post the material. Although the amount of calculations is large, and I will have to purely manually run no more than 10 points there and there.
 

By the way, here are the pictures for your situations. I have converted them into ED1, EY1, ED2, EY2 files for easy reading - I am posting them here.

Files:
ed1.txt  7 kb
ey1.txt  8 kb
ed2.txt  7 kb
ey2.txt  8 kb
 

It worked out for the first one:


The figures are approximate because the algorithm is worse than on the computer at home.

 

For the second case, however, it turned out like this:

 
0.98 for real and 0.997 for HSPC is a noticeable difference. I think in the evening with a normal algorithm it will even become clearer. like 0.985 and 0.9999 for example.
 
The physics of it is actually quite simple. According to what I said earlier. In real quotes we will be able to see some "general shape" and against it "individual differences". Differences of shapes with a "common movement". And in the HSPC we will simply see a "common shape". And if we either count more accurately (longer) then nothing, or absolutely negligible effects simply because we count approximations. No "individual differences" forming ED, EY, DY relations. In a way, this is a proof of my model. More details in the evening.
 

Remember that there is no such thing as an ideal random process in nature. PRNG is a "man-made" event, so there must be some (not necessarily significant) "functional" dependencies. Algorithms that detect/reduce seemingly random series to regularities have long existed. So ....

music pause

 
essentially twisting currency ratios to match the increments of the pairs. All currencies are going in the same direction. It's all about picking proportions. There is more than one solution, as has already been said here. But on the condition of correlation of currencies there should be more or less optimal condition/ratio, probably, we should look through the relations and choose among them the one that has minimal parameters in the set of relations. Maybe we should take into account the minimal modulo value of the normalized series of pairs, let's look at what pair has the minimal increase and choose the nearest relation, so the difference between co-directional indices in the normalized form would be minimal.
 
Joperniiteatr:
essentially twisting currency ratios to match the increments of the pairs. All currencies are going in the same direction. It's all about picking proportions. There is more than one solution, as has already been said here. But given the directionality of the currencies, there should be more or less the optimal condition/ratio, apparently, we need to go through the ratios and choose the one that has the minimal parameters in the set of ratios.

There may be many solutions, but the limit transition is satisfied by one. We are looking for such E, D, Y which would correlate in relations with known relations with coefficient = 1, and between themselves would approach maximally close to unities (I mean correlation coefficients). Reached the maximum possible ceiling corr(E,D)+corr(E,Y)+corr(D,Y) -> 3 as close to 3 as possible. There is, of course, only one such solution. On real courses it has a limit, it cannot reach 3. On the HSPC it can.