One for all. General exeprt. - page 9
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
Hello to all the trout mechanics out there. I'm still an assassin, digging into different combinations of turkeys. And here are my thoughts.
1. Classic MAs in their pure form do not work as signal turkeys - because of the strong lag. The signal indices must be non-lagging - if you want to make a decent MAF that should catch at least 70% of the market moves. CodeBase has a lag-free MA (ZeroLAG MA, 'ZeroLAG MA'). When I saw it, my jaw just dropped. I do not agree with the comment about extremums coinciding with the EMA: there is still a shift backward. Anyway, the formula of the "lag-free" indicator is given - you can calculate the delay theoretically...
I was getting to something similar a long time ago when I was experimenting with neural nets and constructing all sorts of MAs. Why not use it? The Indy is relatively new, not so many people know about its existence (judging by the widespread platitude about the inextinguishable lag of MAs)...
2. The MACD without lag based on the same moving averages was also posted there. It is not the Grail at all but the idea is correct. Of course it is naive to trust the testing results of this Expert Advisor with a simulation quality of 5%.
3. Certainly there will be more signals than necessary and they must be filtered. Here are the filtering options:
a) You don't have to filter by another indicator; you can use simple arithmetic. For example, you can enter not just at crossovers, but, say, when the МАška difference after the crossover reaches a certain threshold. You'll get something like a hysteresis system that reacts not to zero difference, but to a difference that is modulo greater than the threshold. Thus, most false entries in fluxes (and false exits in trends as well) can be easily filtered out. I can tell straight away that an EA with such a "hysteresis" filter based on standard МАшекшек, fails on a timeframe of about 1 hour.
b) Another possible filter is the RSI. It is considered leading, so the lag problem is automatically solved here as well. Of course automating a signal in the form of a divergence (when RSI makes a new local extremum not exceeding the first one, but the price makes a new one) is very difficult, but we can also filter it simply by breakdown of boundary levels 30-70.
c) Another variant of the exit filter is the Parabolic. Its output is considered one of the best of all indices.
If there are several filters, too few inputs may be left. OK, the problem of critically low frequency of inputs can be solved by reducing the TF.
I am waiting for your comments.
b) turbulent (when the price changes fast and strongly) Respectively, the EA that works in a good trend will ruin itself on milk (the author's statement) and vice versa
2) You said that you were dealing with neural networks, I also deal with this issue deeply and quite a lot, and I noticed one very subtle thing that neural networks are artificial intelligence, but artificial not because they are created by people, but because it looks for logic in things, that is, a match for a certain coefficient. I think that it is not news if I say that Forex market or rather the price changes with the help of people, but with a certain algorithm. So, one should not look for the indicator outpacing the time; this does not and will never happen, but it is quite realistic to find coincidence of one candle position from another one with the coefficient. This is what Kohanin's charts are based on
What do you think of my thought?
and b) turbulent (when the price changes quickly and strongly) Accordingly, the Expert Advisor that works with a good trend will be killed by milk (the author's statement) and vice versa
.
Yes, I understand, of course. Ideally one should be able to switch between two strategies, and use, say, some Mashka as a criterion of trendiness/flatness of the market. But now the task is simply to weed out entries into flat areas. If this succeeds, we automatically solve the problem of determining the market condition.
2) You said that you dealing with neural networks, I also deal with them deeply and extensively, and I have noticed one very subtle thing: neural networks are artificial intelligence, but artificial not because they are created by people, but because they look for logic in things, coincidence of certain coefficients. I think that it is not news if I say that Forex market or rather the price changes with the help of people, but with a certain algorithm. Thatis, one should not look for an indicator that is ahead of time and never will be, but it is quite realistic to find coincidence of one candle's position from another with the coefficient. Kohanin's charts are built on this
What do you think of my thought?
It's a good idea, but the problem is that the neural network is opaque, and it takes a lot of work to transfer it to MTS. After NS I was interested in GAs for a while; they are clearer and easier to code. I may go back to NS in the future, but not right now.
b) turbulent (when the price changes fast and strongly). Respectively, the Expert Advisor that works with a good trend will ruin itself with milk (this is the author's statement) and vice versa.
2) You said that you were dealing with neural networks, I'm dealing with this issue deeply and extensively, and I've noticed one very subtle thing that neural networks are artificial intelligence, but artificial not because they are created by people, but because it searches for logic in things, that is, the correlation at a certain coefficient. I think that it is not news if I say that Forex market or rather the price changes with the help of people, but with a certain algorithm. That is, one should not look for an indicator that is ahead of time and never will be, but it is quite realistic to find coincidence of one candle's position from another with the coefficient. Kohanin's charts are built on this
What do you think of my thought?
The system is based on neurosystems already created and she is here in zbornikom EA I am now doing her test so far I can skozat one thing she made me (on the demo in real time) zdelala 3 days 480 $ from 5000 $ when betting in 1 lot. I'm not going to dispute the benefits of the system, but i will tell you if i have learned it before and it will be good for you if you stay offline.
A system based on neurosystems has already been created and is available here in the Advice collection.
...................
A neural network is just an apparatus, a tool...
The whole point is in the formulation of the problem and the interpretation of results.
And here it requires a human brain, and it is all very non-trivial.
With the NS itself, it's just a calculator, only a bit more complicated...
It is important to know which buttons to push.
The thing is, if you know which buttons to push, the calculator becomes a trading tool. But just look for something that works is not realistic, you do not have enough money or time. All that remains is to invent a bicycle that flies!