Machine learning in trading: theory, models, practice and algo-trading - page 1679
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 won't do you any good, you can't do proper
preprocessing (before the inputs were not discussed at all, but picking
alone, in silence) The last cartoon had a correlation
matrix- showing the min lin dependencies between the inputs
(there are other dependencies as well), the algorithms are fresh,
Not more than two years old. The results are average to great.
As in all new rattles there is a random component
(can be removed). For the sake of interest rattle more...
Have you seen my article on preprocessing? OK, or do you need something more tricky? Gaussian mixes are used, I saw that not seasonal, but by clusters, but I did not use
https://www.mql5.com/ru/articles/5451
If the markets were repeatable, the patterns would work.
They do work. And the fact that neural networks do not see them, that's the problem of neural networks.
The weather was inspired by a good example on the hubrahttps://habr.com/ru/post/495884/ yesterday.
about a magic formula, well, as if only from the theory of games, you can look for something, the market can not be in memory, but it may depend on the previous move of other players
Well, it's like in chess - we take someone else's game, then try to use it "head-on", like everything according to plan, we beat him, but then for some reason the opponent did not want to follow our pre-learned plans
In the case of the weather, diffusers give us confidence in the very existence of a pattern. Looking further for a particular kind of this pattern can be done in different ways.
Game theory usually reduces everything to probabilistic models through Nash equilibria. In our case, such an equilibrium is unstable, because staying in it means a gradual drain (prices - SB), but staying away from it can lead to a catastrophically rapid drain. Thus, we are talking about some fluctuations near the equilibrium position, which do not fade, but do not become too large either.
They do work. And the fact that their neural networks do not see, that's the problem of neural networks.
I do not deny, I knew people who showed good results with graphical analysis, but it's not my thing.
In the case of the weather, diffusers give us confidence in the very fact that a pattern exists. Looking further for a particular kind of that pattern can be done in different ways.
Game theory usually reduces everything to probability models through Nash equilibria. In our case, such an equilibrium is unstable, because staying in it means a gradual drain (prices - SB), but staying away from it can lead to a catastrophically rapid drain. Therefore, we are talking about some fluctuations near the equilibrium position, which do not fade, but also do not become too big.
The point is that you can not even find a TS that will show a long time profit around the initial deposit with a given maximum/minimum deviations, the spread can be discarded - it will be an equilibrium TS, right?
I do not deny it, I knew people who showed good results using graphical analysis, but it's not my thing
The point is that you can not even find a TS which will show a long time profit around the initial deposit with a given maximum/minimum deviation, the spread can be discarded - it will be an equilibrium TS, right?
But the graphical analysis can show the result. The patterns, yes. I try to use them most effectively. Without indulgences and other entities.
The thing about the duration - why do I need it? If you set yourself a goal, it's easier to go to it. The market doesn't let you play long to the upside, by definition.
I must be repeating you, in my own words... Or is that not what you're saying?
Tell me a couple of words, so that I understand...
Tell me a few words so I understand...
That's what I mean:
The market does not allow to play in the plus for a long time, by definition...
To find a TS which will show a good result during optimization (training if on the topic of the topic) - no problem, more than half of the members of this forum are able to do it )))
Just look for a TS, that will pass a forward testing after optimization - everyone does it, not many, but they do it ;)
But to find an estimate that TS has stopped working in real time in real trade - I do not know how, I suspect that very few people know,
I think it's not our way :))
The point is that you can't even find a TS that will show a long time profit around the initial deposit with the specified maximum/minimum deviations, the spread can be discarded - it will be an equilibrium TS, right?
I suppose the equilibrium strategy will be a bit unprofitable even at zero spread. I assume that on average the price moves in such a way that most players usually lose - because the spread, in my opinion, is not enough to support the entire market structure. Therefore, you have to be in the minority with your decision, and that's always less than 1/2 as likely. Hence the negative expected payoff.
I do not know how to find an estimate that the TS stopped working in real time in real trading - I suspect that few people know,
SZZ: to determine that the TC does not work is possible except to observe that the balance has become regularly decreasing - well, I think this is not our way)).
There are a lot of tools. In different markets. One collapses, to ride on others. Crossed the line below which it is not profitable: close shop. Maybe so. "Valuation" at risk.
Not the balance, the "result" is intermediate... imho. Balance will show plus/minus when everything is closed.
My point is that equity should be in plus the main thing) Not in one place, but in another... Cumulative.
I suppose an equilibrium strategy would be a bit of a loss with zero spread as well. I assume that, on average, the price moves in such a way that most players usually lose - since the spread, in my opinion, is not enough to support the entire market structure. Therefore, you have to be in the minority with your decision, and that's always less than 1/2 as likely. Hence the negative expected payoff.
Spread is enough for everyone who earns on it, well, if not enough, then involve "ingenuity" - requotes, large slippages - although the point is quite different - their earnings are guaranteed and risk-free, so to speak, pure mediation
Not about that, it's not our lot, or we should drop everything and do it)))
As for the optimum TS I mentioned, there is a trivial one - by opening a bar by turns buy-sell-buy-sell, regardless of the outcome with a constant lot
with this TS is pretty quickly found the optimal StopLoss and TakeProfit, but this TS works more confidently only with increasing TF - I checked a long time ago, it seems on the daily bars more or less steadily
BUT... and there will be problems - crises and changes in rates of the Federal Reserve System cause a series of defeats/victories in such TS
That is, I think that even to find a TS "constantly dangling" near zero, most likely not real to find - tomorrow I'll check GA tester
There are a lot of instruments, after all. In different markets. One goes under, ride out the others. Crossed the line below which it is not profitable: close shop. I guess it's like this. "Valuation" at risk.
Not the balance, the "result" is intermediate... imho. Balance will show plus/minus when everything is closed.
My point is that equity should be in plus the main thing) Not in one place, but in another... Cumulative.
I do not know the TS that may work without modification (or optimization) for different instruments, i.e. the problem is the same - to find out that the current TS has stopped working for the instrument, but not to drain the deposit and what to do in the future - to search for another TS or another instrument - is of secondary importance.
I do not know the TS that can work without modification (or optimization) for different instruments, i.e. the problem is the same - to find out that the current instrument has stopped TS working, but not to lose the deposit, and what to do in the future - to find another TS or another instrument - is of secondary importance
For me, TS - is a set of all positions in all instruments, markets. Open positions. And their "result" by risk/reward. If one element of the system has paid for itself, then its risk is reduced to zero and its value is not important now (we close part or all of it). Or leave this element, but so that there is potential. Changes the "balance of power", the main thing that one or more of our elements in the TS does not pull all the others too much ...
Modification is not necessary. It then immediately requires modification of all others. One lemon gave its juice and enough. Why pump him up again? Only if you want it to be the only one that works, this one. As if there is a lot of it, the effect in the form of the final result is better...