Machine learning in trading: theory, models, practice and algo-trading - page 1145
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
To teach on the future, and test on the past, this is only found in this forum)))
There is no difference, it has been discussed. No peeking into history
Show me at least one forum where there is a normal discussion of MO to the marketI would say that it depends on the Expert Advisor. If it generates a clear sequence of trades, that is, when the position is opened and closed, and its volume does not change between opening and closing - it is better to count by trades. If the position volume changes smoothly over time, then identifying the moments of a trade is less meaningful and can be calculated according to your method.
The pantural method is more good for selling the TS and finding investors) So over time, I suppose, will switch to it)
I think it's too much of an honor to rewrite a forum upstart's algorithm for calculating the annual Sharpe Ratio))
And seriously, "annual" and trades have no sense as a metric, it can not be optimized, because every time the strategy will be different by the number of trades, so for example a strategy generating 1000 trades will have three times less Sharpe than the strategy with 100 trades, with the same profit and max drawdown.
May be the equation of the individual trades profits and losses using "Q" learning and "Bellman equation"
It means that it should be taken in the future considering the profits in each trades.
I don't know :)
"Q" learning and "Bellman equation"
It means that it should not be taken into account the individual profits and losses in each trade.
q-learning it's a table method, so individual profits corrected with Bellman
I approximate policy directly, same thing but faster
there is no difference, it was discussed.
there is no difference, it was discussed. No peeking into history
I don't know if you're right, but I'm sure you're right.It's not about peeking, although it may also be under certain conditions, but the fact that the OOS should be as close to reality, because you want the result of OOS to repeat + - on the real, and if you test for the more distant past, it will be close to the past, and the market during this time and can change more or less. Your method can lead completely to the absurd, for example if you separate OOS and real for years)))
There are not many forums wilmot, quantnet, elite and ... I can't think of anything else, mql5 is the most liquid, but there's no "proof" of "profitable" trading with martin or sb on western forums, maybe because only qualified investors are trading there, who have never "raked out a deposit", stupid idiots...
there is a difference, it's been discussed. someone just let it pass by.
there is absolutely no difference
I recently tried to explain, but I forgot why he thinks so ))
If the policy is approximated by some model (say, linear) then we just get a solution on the new data and that's it, substituting them into the model
What you're describing is the process of finding the highest reward.
the main problem with non-stationarity is when it stops working on new data. There are non-stationary bandits described there, but I haven't gotten to them yet. Admittedly, there's nothing there that I don't already know, as it turns out :) But I need some ideas\solutions on how to properly give rewardsYou overestimate me) I haven't progressed beyond the introduction.)
As they themselves write - it's a kind of subclass of nature (environment) games. I'm sure almost all of our models lie within the nature game, but I don't know how appropriate these "bandits" are.
I like latent Markov processes better. There, the non-stationarity can be a consequence of the fact that we are not observing all the variables. Roughly speaking, a process that is non-stationary for us will be derived from a stationary process but only known to the market maker.
This is not about peeking, although it may also be under certain conditions, but that OOS should be as close to the real, because you want the result of OOS repeated + - on the real, and if you test for the more distant past, it will be close to the past, and the market during this time and may change more or less. Your method can lead completely to the absurd, for example if you separate OOS and real for years)))
There are not many forums wilmot, quantnet, elite and ... I can't think of anything else, mql5 is the most liquid, but you can't find "proof" of "profitable" trading with martin or sb on western forums, maybe because only qualified investors are trading there, who have never "raked out a deposit", stupid youngsters...
yeah that's bullshit, i'm just testing what i did (generalizability), i'm not trading that
It's bullshit, I'm just testing what I made (generalizability), I'm not trading it
I got it, bullshit is bullshit))