Machine learning in trading: theory, models, practice and algo-trading - page 3540
![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
Explain.
Look carefully at the second picture, if you don't see any useful properties in the synthetic series, I'll write it down.
It's smoothing without lag. No?
No.
there's no smoothing and there's no lag.
But you can say there is a filtering of the trend noise.
You can see in the picture
that on the original the price goes up with noise, and on the reconstructed series you can see that all candles are "white" and you can filter by this feature.
property #2
Even more interesting...
pay attention to the candles with gaps
I wondered if it is possible to train AMO on indicators and oscillators to restore OHLC prices.
I submitted about 40 indicators and oscillators, trained linear regression....
It restored perfectly.
and this is what price recovery looks like from only two indicators RSI and SMA.
pretty good, but with a note that there is nothing to restore.
The price is explicitly present in the indicators. That is, you can take the calculated formulas and allow them back. If you have enough indicators, you can explicitly get the transition matrix (readings=>price).
but there is a nuance - for the case of RSI/SMA an interesting point: AMO implicitly selected areas with minimum error, because there is not enough information. "It itself" found RSI/SMA combinations where the price can be restored almost exactly (minimum volatility with non-critical slope or something like that).
No.
there's no smoothing and there's no lag at all.
But you can say there is a filtering of trend noise
You can see on the picture
that on the original the price goes up with noise, and on the reconstructed series you can see that all candles are "white" and you can filter by this feature
The price is explicitly present in the indicators.
Yes, but you can't reconstruct the price from a 20-mashka, you need enough information.
I think I wrote the same thing, only in different words. At least I meant it.
I see.
But there's no smoothing without lag, it's more like filtering.
Useful thoughts on optimisation
https://www.youtube.com/watch?v=NoZZ8evZQ18&t=555s&ab_channel=%D0%92%D0%BE%D1%80%D0%BE%D1%82%D0%BD%D0%B8%D1%86%D0%BA%D0%B8%D0%B9%D0%AE%D1%80%D0%B8%D0%B9
Yes, but you can't reconstruct a price from a 20-year-old car, you need a lot of information.
I don't really want to mess with the formulas for the sake of dialogue,
but to give you an idea: if you take the 20th MA-scheme, you will need 41 indicators to restore 30 points accurately. (only in the tail there will be 20/2 +- 1 inaccuracies).
I don't really feel like messing around with formulas for the sake of dialogue,
but to give you an idea: if you take the 20th MA-case, you will need 41 Mach numbers to accurately restore 30 points. (only in the tail will be 20/2 +- 1 inaccuracies)
without formulas, purely common sense and practice.....
you can't describe the high-frequency oscillations that it suppressed with a single flywheel, the size of the window won't help, it's irretrievably lost information....