Machine learning in trading: theory, models, practice and algo-trading - page 1270
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
these are simply different tasks.
the task is not to influence but to predict
at the level of abstractions of statistics and IO there is no difference at all
There is another problem: professional players in the UK are extremely inefficient, global markets are very efficient
No, the situation is different, here you don't need to predict many moves, you need to calculate the most profitable action here and now, but unlike trading, if the situation changes, there will be a new decision, and the cost of an error is minimal, rarely without loss of resources, while with us even an exit from an erroneous position will lead to a loss.
everything is the same, the problem is the effectiveness of the opponent's strategy. The market is harder to beat.
You can't beat the market, that's the point. Totally different requirements - in a game you control the situation, while in trading you only have to catch a wave in the necessary direction, and all you have to do is to count the crests of waves.
You don't control the situation in the game, but you evaluate your chances depending on the decisions you make
So it should be clearer that playing the game with your opponent and the market are similar things.
In the game you influence the situation, you don't need to assess your chances of winning at every moment of making a decision, you look for opportunities to spend less resources to do more damage. Also, like I said, there's a dynamic process, the decision can be revised at each iteration without any consequences, while in our game even revising the decision costs a fee.
If you play zeros in tic-tac-toe, you have no influence on anything if your opponent is efficient, he always wins.
If you're trading an inefficient market, you don't have to influence it to win.
Purely a terminological argument, I think, that doesn't change anything )
I disagree, when you play tic-tac-toe you influence your opponent's probability of action with your actions.
I'm not arguing, I don't need it, I see a difference in situations for myself, but I admit that these two directions can give something to each other in symbiosis.
he will just put a cross in another cell, the probability of his winning will remain 100%. what influence are we talking about? he had no effect on the probabilities of the outcomes. purely physical influence does not make any sense when it comes to the odds
Well, if he puts his cross or zero in the center, obviously you'll have less chances, or it's time for me to sleep...
If he plays correctly and does not make mistakes (effective), then in tic tac toe always wins crosses (ie, the first one to go)
Okay, never mind... ) Just if you catch the idea, you'll stop thinking about itWell, that's what influences, and this is very important, and it is the influence there is more training than prediction. There monitoring of foreign units, and evaluating their threat, i.e. the possibility to control them cheaper, and separately assess the branch of development - here you can just use a binary tree. And something to assess the situation. I think there is a symbiosis of different models responsible for different things and exchanging data with each other.
What did I agree with? That's exactly the opposite of what I said. There is a probabilistic picture, a Nash equilibrium for example, where no matter what you do you will not change the balance of power, you have an optimal strategy that you can stick to, and that is all. Playing for zeros with an effective opponent you will never win, i.e. you cannot change the situation.
It's the probabilistic transitions that are evaluated in the game, something like a Markov chain. Cause-Effect. Strategy-counterstrategy, and no influence. If the opponent is wrong, he loses.
Take your anticipated rate change as the market strategy, and take your anticipated rate change as your strategy. Whichever strategy is cooler, he wins. An efficient market has the best strategy, i.e. none of the participants can beat it in the long run.
That's the thing, your actions affect the probability of your opponent's actions in the game, but you can't affect the market. Plus add the stationarity of the playing field. I don't see the game as a serious prediction, just a probability calculation and risk mitigation by your actions. We can also reduce risks by analogy, for example enter the market with a small calculated stop, but then the number of deals will be catastrophically low. In general, if you had 100k rubles and you were sitting on the second echelon of Moex stocks, the analogy might be appropriate, because you can significantly influence the price and its movement, and all you have to do now is try to predict what someone with these 100kk will do. And we think that you can do this by his behavior, i.e. how he will influence the price, we have nothing else.
You can't influence your opponent's choice of strategy. You can put him in a stalemate, but that's part of your strategy. He may or may not change his strategy as the play progresses, at will. In general, if you have a flush royale like the market, you win in advance. That is, you have the best strategy in hand from that space, with any opponent's moves, as in tic tac toe, you win.
Individual moves have no effect, because they are beaten in advance by your opponent's strategy (in any combination).
Or let's say in an intellectual dispute with someone, you have a brick in your hand and your opponent doesn't. What's the odds of you blowing it?Here you give examples where two subjects influence each other, in the game they do, but you critically have little influence on the market, I tell you that. So whatever strategy you choose, the market will do what it needs.
You're thinking narrowly, it seems I can't get the point across that the game is about probability and not about interaction with another player. This is where the argument began, that you said you needed different algorithms for the market. I said no.
And don't try, I think pragmatically, substantively about these cases. You can think in terms of probabilities when you can influence them. And so, it's called hope, hope that the market will trade according to a randomly recreated algorithm... Well, there is such a probability, but how big is it?
Unfortunately, the conversation has turned into nonsense.
therver is one for all, so are neural networks.
There are no pragmatic and subject things in MO, only abstract concepts.
We apparently have a different idea of the market and different expectations of MO, including why it should work.
Thanks to dialog with you I have already imagined how I would train the bot for the game and now everything seems less fairy-tale to me.