Machine learning in trading: theory, models, practice and algo-trading - page 3266

[[Deleted]]

fxsaber #:

It's not like anyone else didn't know but me.

Pearson is invariant for the actions of multiplication and addition.

Addition didn't feel right, despite the simple formula. And the wiki specifically talks about it.

Specifically, this original matrix

has a unit correlation matrix (rows by columns).

Yeah, that's cool.

[[Deleted]]

Forester #:
I think it makes sense to normalise the columns before calculating the correlation. The reasons were discussed earlier.
Maxim - I thought you wanted to try normalisation. Were there any improvements? Or deterioration.

Everything is more or less OK there without normalisation.

[mytarmailS]

Level experiment.

1)

We take a section of the chart of the euro 1m in the size of 100 candles and count how many times the price of high stayed in the same place, that is, how many were high-rates at the same price.

Do this 500 times on 500 different non-overlapping sections, summarise the statistics.

2) Generate random series with distribution identical to prices, with the same average of real ticks, with the same standard deviation, generate a series of m1, as well as the series has the same number of decimal places (karoch made the maximum identity).

I doubt that anyone will distinguish this chart from the real one.

============================================

Let's conduct an experiment.

   n_times  real simul
1        1 19506 24036
2        2  6575  4829
3        3  2373   959
4        4   873   141
5        5   375    32
6        6   154     3
7        7    70     0
8        8    35     0
9        9    18     0
10      10    11     0
11      11     3     0
12      12     5     0
13      13     0     0
14      14     1     0
15      15     0     0
16      16     1     0
17      17     0     0
18      18     0     0
19      19     0     0
20      20     0     0

This is the absolute sum of bounces for all experiments.

The table shows that on simulation data the price almost never hits the same price 5-6 times, but the real price can hit much more times.

So it can be interpreted in the direction of the existence of levels.

[Maxim Kuznetsov]

mytarmailS number of decimal places (karoch made the maximum identity).

I doubt that anyone will distinguish this graph from the real one.

============================================

Let's experiment

This is the absolute sum of bounces for all experiments.

The table shows that on simulation data the price almost never hits the same price 5-6 times, but the real price can hit much more times.

So it can be interpreted in the direction of the existence of levels.

In real currency quotes there is (not very big, but there is) dependence on price levels. Data analysis was started, but traditionally for the branch we rushed to experiments.

ZY. in general, in any prices there is also, but even weaker, induced by currencies

[mytarmailS]

Maxim Kuznetsov #:

in real currency quotes there is (not very big, but there is) dependence on price levels. We started analysing the data, but traditionally for the branch we rushed to experiments.

ZY. in general, any prices also have, but even weaker, induced dependence on currencies

Experiment is the criterion of truth...

And experiment has shown that if the price hits some level 3 times or more, it's not a coincidence, because coincidence has completely different statistics on a similar event.

What is "induced by currencies" ?)))

[Alexandr Sokolov]

mytarmailS number of decimal places (karoch made the maximum identity).

I doubt that anyone will distinguish this graph from the real one.

============================================

Let's experiment

This is the absolute sum of bounces for all experiments.

The table shows that on simulation data the price almost never hits the same price 5-6 times, but the real price can hit much more times.

So it can be interpreted as the existence of levels.

Can you look at the code that generates the chart?

[mytarmailS]

Alexandr Sokolov #:
Can we see the code that generates the graph?

The code is in R

[Maxim Kuznetsov]

mytarmailS #:
Experimentation is the criterion of truth...

And experiment has shown that if the price hits a certain level 3 times or more, it is not a coincidence, because coincidence has completely different statistics for a similar event.
.

What is"currency manip ulation" ?)))

they are all linked. Currencies hit and all derivatives, securities, stocks repeat the movement in one way or another. Currencies are stronger and affect everyone directly, and the inverse effect is from the total aggregate.

very figuratively: EURUSD hits a level, a trader sees a weak channel on AAPL and tries to do something. So he has a lead - not related to apple business, significant change in supply/demand of the stock and investor sentiment. It's outside, it's outside, it's outside.

[mytarmailS]

Maxim Kuznetsov #:

they're all connected. Currencies hit and all derivatives, securities, stocks repeat the movement in one way or another. Currencies are stronger and affect each person directly, while the inverse effect is from the total aggregate.

very figuratively: EURUSD hits a level, a trader sees a weak channel on AAPL and tries to do something. So he has a lead - not related to apple business, significant change in supply/demand of the stock and investor sentiment. It's outside, it's outside, it's outside.

I agree.

[Alexandr Sokolov]

mytarmailS #:
R code

I'll figure it out.

... or at least tell me the principle.