Machine learning in trading: theory, models, practice and algo-trading - page 3529
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
This is an obvious and non-tradable pattern, and I would like non-obvious and tradable ones)
This is verified by comparing correlation charts for several different timeframes. They are all noise (I just didn't want to clutter the post with a bunch of charts). There is no "reversal schedule" either to the nearest minute or to the nearest hour. There are reversals, but there is no predetermined schedule of them)from non-obvious: you can ask ChatGPT (or other big networks) to draw a schedule for tomorrow :-) it is true that you have to learn to communicate with it, somehow explain what to take as benchmarks and that it is a continuation of the current one and interconnected with others, and in addition you have to circumvent prohibitions (using AI in explicit forecasting is prohibited to avoid trouble, it has to evade answers).
about correlation, who told you that pierson is not a window function ? tell him "fi...". The middle one has a window, it has a window, that window blows in.
on a brainstorming level, a mental-visual experiment:
- we present a chart, X is EURUSD, Y is GBPUSD.
- A point on the chart unambiguously defines both quotes and their cross at the moment of time.
- add time (animate) - the point pretends to be a Brownian movement, but inside a strange figure... there are obvious logical prohibitions: it cannot move horizontally/vertically/radially. In a quantum of time, EURUSD, GBPUSD and EURGBP change; all three of them
- like heatmap, we add and sum up OrderBook volumes - from EURUSD vertical bars, from GBPUSD horizontal, and from EURGBP - radial.
- by their sum we select isolines: heatmap isolines will no longer be a perfect ellipse/circle, but a somewhat "egg-shaped figure" (EURGBP is radial and it will give a thickening).
- due to other factors (from projections of other currencies and that EUR:GBP is not 1:1 and there are logical prohibitions) - the egg-shaped figure should turn out to be a "pregnant bean" :-) a figure with no symmetry on axes and diagonals.
- and this is without taking into account seasonal volatility, which will add its own effects. slight bends at round levels in particular.
It will turn out to be a potential field, and an eternal game of arkanoid - the quote hits one wall, demolishes a brick and flies off to another.
You can't draw this in reality - there is simply no data. 10-20 pips bets are nothing :-)
you can do it on crypto, but crypto is not fiat and is not linked to the economy.
Try spiking inefficient predictors, even random, if the effect remains, it's really interesting to think about. I also got a lot of stable clusters in March, but not all of them.
Well, I guess it's probably hard to take, as I haven't read anyone doing it....
The point is that at each iteration of the tree construction we have good variants for the next step (a set of variants from which one of the standard metrics is chosen), so, depending on the choice of the split (in my case double - since the quantum segment is chosen), some of the variants statistically deteriorate - thus damaging the probabilistic structure, which does not allow to use some of the splits at the next iterations.
Thanks, but there is no interest in black boxes - I am whitewashing my algorithm :)
You won't be able to communicate in such terminology, it is perceived as an incoherent set of words. It's a long-standing problem.
You can't communicate in such terminology, it is perceived as an incoherent set of words. It's a long-standing problem.
Yes, new things are always difficult to perceive - I guess you need illustrations.
From the non-obvious ones: you can ask ChatGPT (or other large networks) to draw a graph for tomorrow :-) it is true that you have to learn to communicate with it, somehow explain what to take as benchmarks and that it is a continuation of the current one and interconnected with others, and in addition you have to circumvent prohibitions (using AI in explicit forecasting is prohibited to avoid trouble, it has to evade answers).
about correlation, who told you that pierson is not a window function ? tell him "fi...". The mean is there, it has a window, that window blows in.
on a brainstorming level, a mental-visual experiment:
- we present a chart, X is EURUSD, Y is GBPUSD.
- The point on the chart unambiguously defines both quotes and their cross at the moment of time
- add time (animate) - the point pretends to be a Brownian movement, but inside a strange figure... there are obvious logical prohibitions: it cannot move horizontally/vertically/radially. EURUSD, GBPUSD and EURGBP all change in a quantum of time; all three
- for example in colour, as heatmap, add and sum the OrderBook volumes - from EURUSD vertical bars, from GBPUSD horizontal, and from EURGBP - radial.
- by their sum we select isolines: heatmap isolines will not be a perfect ellipse/circle, but will be somewhat "egg-shaped" (EURGBP is radial and it will give a thickening).
- due to other factors (from projections of other currencies and the fact that EUR:GBP is not 1:1 and there are logical prohibitions) - an egg-shaped figure should result in a "pregnant bean" :-) a figure with no symmetry on axes and diagonals.
- and this is without taking into account seasonal volatility, which will add its own effects. slight bends at round levels in particular.
This will result in a potential field, and an eternal game of arkanoid - a quote hits one wall, demolishes a brick and flies off to another.
You can't draw this in reality - there is simply no data. 10-20 point glasses are nothing :-)
you can do it on crypto, but crypto is not fiat and is not linked to the economy.
It seems that all reasonable people have already got rid of strange illusions about any LLM.
If I was talking about a window indicator, the result would look like an indicator. But I have not an indicator, but a graph of correlation dependence on time of day, like a graph of volatility dependence on time of day.
It is not for me to draw something, but for Nikolay Semko, no one can compare with him in this area
Yes, new things are always difficult to grasp - apparently you need illustrations.
It's not new, it's a kind of nonsense. Even articles in English on the MoD are easier to understand. It's a whole different level.
I reread it again - everything is clear and understandable for me. I will not change my terminology to explain it to you, but I can change mine if you have any questions.
It would seem that all reasonable people have already got rid of the strange illusions about any LLM.
If I were talking about a window indicator, the result would look like an indicator. But I have not an indicator, but a graph of correlation dependence on time of day, like a graph of volatility dependence on time of day.
It is not for me to draw something, but for Nikolay Semko, no one can compare with him in this area
everything new is well forgotten old :-)
about ACF (you have ACF, not correlation of two values) I think it was already said about 5-10 years ago. :-) on quotes and their derivatives, the ACF calculated using textbooks is chaotic. Beyond 6 counts it will be a complete nonsense. And it doesn't matter what counts, whether minutes, bars, returns. It simply shows nothing, because of volatility, because of the big difference in the window, because of sqrt(T) (it is probably 5-6 samples and the opinion that the market has a memory for 5 moves).
you've got an ACF.
What does that have to do with it? Correlation was calculated, in fact, the same way as volatility. For example, for X=1, a sample of all increments for the 1st minute was taken and its correlation with a sample of all subsequent increments for the 2nd minute was looked at. And so on for all X<=1440. The window function is organised differently.
Code in R
I reread it again - everything is clear and understandable for me. I will not change my terminology to explain it to you, but I can change mine if you have any questions.
I admit that there is a rational grain in your researches, but it is very difficult to perceive. Perhaps some thesaurus is needed. For example, what is probability structure damage :)