Machine learning in trading: theory, models, practice and algo-trading - page 2861
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
It is not a complex structure - it is a coincidence that will not repeat/repeat in the future (P=0.5). That's the conclusion I've come to. All "complex" non linear relationships without theoretical justification are fiction. Just an algorithm stretched over the data. Market participants are not aware of this "complex" relationship. So this relationship has no effect on price movement.
It is another matter if there is a theoretical basis for "Thursday at 4pm sell if the daily bar is rising". See below
Well, let's say you are right and it was a random process, but is it possible to identify it in a mathematical way? After all, if there is a shift in probability, it means that some event has entered the window, which is caught by our logic in the form of predictors, and then it will be blurred with the arrival of new data. Maybe we should evaluate not the total probability shift, but at least at a number of sites?
It is possible to try to justify price behaviour by fundamental indicators, but to do so you need to have knowledge about all the intricacies of the financial system, and nobody usually has this knowledge. Besides, I proceed from the fact that in order to implement plans to sell/buy an asset or currency in large volumes, it is necessary to collect liquidity, i.e. the price needs to move in different directions and collect the necessary bids to set a position. The price moves according to the decisions of market participants, who are looking for circles on the water (using technical analysis), and with the help of MO we should determine which methods of technical analysis will be the most popular among market participants and based on this use their strategies suitable for this behaviour.
Any theoretical interest rate calculations can also be included here.
Your example demonstrates that traders are not professionals in accounting and financial management. Even not very large firms engaged in import and export look at the market and carry out currency transactions at rates that are favourable in their opinion, and free funds (if they cannot be withdrawn into the black through pads at one time) are deposited or buy derivative financial instruments, and this is from my experience of already working in these organisations.
I am just very sceptical about fundamental analysis in the Russian Federation, especially of enterprises, and any statistics, as I know the kitchen from the inside.
As much as I can try to convey - predictors are the essence. And when they are normal - MO is no longer necessary.
So the question is how to distinguish "normal" from "not normal". I am putting more emphasis on this direction.
I'm leaving the discussion - it's just that I taught my neural networks on OpenCL 10 years ago, spent a shitload of effort and time on it, but I didn't get any profit out of it. God willing, you will succeed.
Thank you. Maybe I will, maybe I won't, it's been fascinating.
But I don't understand, in 10 years you haven't developed any unique ideas in this field? If you give up everything, you could publish them, so that people could continue from the place where you put the point, instead of starting the torment from the beginning.
Well, let's say you are right and it was a random process, but is it possible to detect it in a mathematical way? After all, if there is a shift in probability, it means that some event has entered the window, which is caught by our logic in the form of predictors, and then it will be blurred with the arrival of new data. Maybe we should evaluate not the total probability shift, but at least in a number of areas?
...
So the question is how to distinguish their "normal" from their "not normal".
No, I'm out. The only method available for me to identify predictors now is my subjective knowledge of the market + observations. And if there is a hypothesis - then testing it in a tester or on the real, if it is work in the glass. Very slow, but reliable.
Only I do not understand, for 10 years you did not have any developments in this area, unique ideas? If you give up everything, you could publish them, so that people could continue from the place where you put the point, and not start the torment from the beginning.
I gave up this direction a long time ago - about 5-6 years ago for sure. I was closely engaged in it for about 3 years (in my spare time). There is nothing valuable and useful in my "developments": multilayer networks, genetics for optimisation, and I don't remember what else.... And science/practice has advanced a lot during this time - modern packages can do much more. The bottom line is the same - retraining and drain.
The only idea I'm trying to convey is that it's a torment and you don't need to start at all. It is better to take something simple and proven and try to make money with it.
Another fresh example - though not very positive for its participants :))))) I will put the main point here...
...
At Jane Street, Sam was involved in arbitrage trades in low-liquidity cryptocurrencies. He took advantage of differences in cryptocurrency rates on different exchanges, buying up cheap cryptocurrencies on Asian exchanges and then selling them at a higher price in the US. In September 2017, Bankman-Fried left Jane Street
...
in November 2017, he founded his own trading company, Alameda Research, in Berkeley, California, which immediately began generating millions in profits.
....
In March 2018, Bankman-Fried invited his Jane Street acquaintance Caroline Ellison over for coffee. That's when he suggested she join Alameda. In Caroline's own words, she was already "more experienced than many of the then Alameda traders" at the time, even though she was only 24 years old. However, in May this year, Allison admitted in a Spanish-language podcast that her entire trading strategy was based on "primary schools level maths" and intuition.
....
In a single day, on 8 November, his fortune collapsed from $10.5bn to
And I remember it being said here for a very long time that there are no fish in arbitrage. While the talk here was going on, the guy who was out of the loop made a decent amount of dough.... But because of just amazing stupidity and naivety, he will probably go to jail now....
Nah, I'm out. The only method available to me to identify predictors now is my subjective knowledge of the market + observations. And if there is a hypothesis - then testing it in a tester or on the real, if it is work in a bet. Very slow, but reliable.
I gave up this direction a long time ago - about 5-6 years ago. I was closely involved in it for about 3 years (in my spare time). There is nothing valuable and useful in my "developments": multilayer networks, genetics for optimisation, and I don't remember what else.... And science/practice has advanced a lot during this time - modern packages can do much more. The result is the same - retraining and drain.
The only idea I'm trying to convey is that it's a torment and you don't need to start at all. It is better to take something simple and proven and try to make money with it.
Another fresh example - though not very positive for its participants :))))) I will paste the main point here...
And I remember how they have been saying here for a long time that there are no fish in arbitrage. While they were saying it here, the guy who is not in the know made a decent amount of money.... But because of just amazing stupidity and naivety he will now probably go to jail....
Another recent example - though not very positive for its participants :)))))) I'll paste the main part here...
No, I'm out. The only method available to me to identify predictors now is my subjective knowledge of the market + observations. And if there is a hypothesis - then testing it in a tester or on the real, if it is work in the glass. Very slow, but reliable.
I gave up this direction a long time ago - about 5-6 years ago. I was closely involved in it for about 3 years (in my spare time). There is nothing valuable and useful in my "developments": multilayer networks, genetics for optimisation, and I don't remember what else.... And science/practice has advanced a lot during this time - modern packages can do much more. The result is the same - retraining and drain.
The only idea I'm trying to convey is that it's a torment and you don't need to start at all. It is better to take something simple and proven and try to make money with it.
Another fresh example - though not very positive for its participants :))))) I will paste the main point here...
And I remember how they have been saying here for a long time that there are no fish in arbitrage. While they were saying it here, the guy who is not in the know made a decent amount of money.... But because of just amazing stupidity and naivety, now he will probably go to jail...
well done !
Nah, I'm out. The only method available to me to identify predictors now is my subjective knowledge of the market + observations. And if there is a hypothesis - then testing it in a tester or on the real, if it is work in a bet. Very slow, but reliable.
I gave up this direction a long time ago - about 5-6 years ago. I was closely involved in it for 3 years (in my spare time). There is nothing valuable and useful in my "developments": multilayer networks, genetics for optimisation, and I don't remember what else.... And science/practice has advanced a lot during this time - modern packages can do much more. The result is the same - retraining and drain.
The only idea I'm trying to convey is that it's a torment and you don't need to start at all. It is better to take something simple and proven and try to make money with it.
Another fresh example - though not very positive for its participants :))))) I will paste the main point here...
And I remember how they have been saying here for a long time that there are no fish in arbitrage. While they were saying it here, the guy who is not in the know made a decent amount of money.... But because of just amazing stupidity and naivety, now he will probably go to jail...
I get it - you only admit what you think you understand. I on the contrary have recognised that I don't understand much, which is why I moved from handwritten strategies to automatically generated ones.
Why haven't you addressed the issue of how to reduce overtraining?
Cool!!!
You can make a fitness fukk and torture GPT3 until it creates a workout )