What do you think of multicurrency indicators? - page 4
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's not an indicator, it's a mass grave.
I'm just now dealing with correlation dependencies and have come to some pretty proven conclusions:
1. Correlation can hardly claim to be an independent factor in making a relatively long term decision. Apparently there are a huge number of arbitrage machines in the market that blur the real long term "intentions" of currencies, but as a filter in short periods correlation works well.
2. Reaching a correlation peak is a signal that confirms not the beginning of a general trend, but on the contrary, a certain sign that the micro-trend of the instrument being traded is about to change. It means that I use correlation only as a condition for reversal (position opening) and the rascorrelation as a condition for position closing. I cannot judge why it works better like that, maybe due to the same pair trading influence or maybe due to the fact that prices several times "trade" the same levels, but the fact remains - the correlation works better on rebound, i.e. I observe the real synergistic effect of bounce indicators and correlations, both for opening and closing positions.
3. Correlated instruments should be about the same volatility. A comparison with a less (more) volatile instrument (such as EURUSD) only confuses the issue, but the fact of exceeding (lagging) the volatility of the more volatile instrument can be used to restrain (stimulate) trade on the basic instrument. In other words, if the "older brother" gets nervous, do not make hasty decisions - wait until this excitement will be transmitted to your symbol.
4. Correlations with different instruments work better at different times of the day! and the more different nature of these instruments, the better. Usually the break occurs in the middle of the US session, but the Asian session is not "even" either. To be more exact, Asian session is "obscure" in my opinion, I have not found any clear correlations for it yet (probably I did not search well), therefore I try to build my owls in such a way that positions will not endure, and I even skip this session.
5. The correlation with the so-called currency indexes works not better, but worse than the comparison of two independent pairs, even worse than comparison with the usual stock indexes, i.e. averaging the properties of a large number of instruments brings not profit, but harm and the more pairs are taken into account, the more senseless becomes the indicator. Correlation to work should be as specific and specific as possible, and the currency indexes in this sense - is an indicator of nothing. Trading a bunch of different currency pairs simultaneously, using a common indicator, in the hope that all the errors will average and disappear, and the general trend will miraculously emerge - this is madness, in reality, everything happens exactly the opposite.
6. Even correlations that work become obsolete relatively quickly. I don't have exact measurements, but from the feeling of tests - a year and a half. So now I make a "screening" machine, when the auto-optimizer runs different correlated combinations in turn and automatically selects promising ones. Hopefully, monthly analysis will allow to catch "incipient" correlations.
I have a few more ideas, but they are still being tested.
I.e. the general conclusion - in order to use the correlation we should not only search for the coincidence (opposition) of movement directions, we should search for the general rhythm of oscillations, then the pinches of two instruments are dampened, and the turning points become more obvious, because the real relational specificity of the instrument appears. It is almost impossible to predict the long-term trend with the help of correlation, because it doesn't depend on it. I.e. the market is rather a performance of choreographic pairs and entire groups on the dance floor than a race of track-and-field athletes on the stadium. And the correlation simply shows who is dancing with whom at the moment.
And the main multi-currency indicator could be something like this: https://www.mql5.com/ru/code/897
I agree with that., these are long-term correlations, medium-term correlations are somewhere in a month
it is necessary to look fora general rhythm of fluctuations, then the pinches of the two instruments are dampened and the turning points become more evident
absolutely
but imho, this is not a correlation as such, but regression or another algorithm is relevant
though it does not give stationarity i.e. there is no drawing grail
And what do you mean by regression relationship and how does it differ from "linear relationship closeness", as Pearson correlation is sometimes also called, for example? And what might its use look like in practice?
Well, I wouldn't want to brag too much (or maybe it's just the opposite of my secret desire?) - https://www.mql5.com/ru/code/11859
regression is like a correlation, but it solves the problem more directly - it selects lots so that the deviation of the two instruments (or baskets) was minimal
i.e. regression equation attracts points of two charts as closely as possible, correlation cannot do that
2. Reaching a correlation peak is not a signal that a general trend has begun, but rather a sign that the micro-trend in the instrument being traded is about to change.
There is one difficulty here. The trend may continue in one way, but it may end in N different ways (the same way we decide to open buy or sell for one pair, i.e. 2 different directions, and to trade by "correlation" we will have N pairs, each one can "break" in its own way). So it is not clear which way to open a position with this strategy.
There you go! And the trouble is that a breaking trend can make a fool of itself several times - going up and then down several times
I've tried a reversal strategy on synthetics and got a lot of unpleasant feelings (although it looks good on the story)
i cannot choose the size of the "box" for reversal, it very often goes up and down
It is easier to trade on the classic lines - simply following the trend continuation or rollback or spread (return to the average).
There is one difficulty here. The trend may continue in one way, but it may end in N different ways (in the same way we decide to open buy or sell for one pair, i.e. 2 different directions, while for the correlation trade we will have N pairs, each of which may "break" in its own way). So it is not clear which way to open a position with such a strategy.
There you go! And the trouble is that a breaking trend can make a fool of itself several times - going up and then down several times
I've tried a reversal strategy on synthetics and got a lot of unpleasant feelings (although it looks good on the story)
i cannot choose the size of the "box" for reversal, it very often goes up and down
It is easier to trade on the classic lines - continuation of the trend or rollback or spread (return to the mean).
The regression trading strategy seems to be obvious:
- At each bar, we calculate the new optimal shares of the instruments.
- Rebalance the portfolio according to these shares and trade until the next bar.
Ideally, we get a linear regression as a trend in our account. In practice, our portfolio will always differ from the optimal, calculated on the history, portfolio function. However, all we need is for our portfolio not to crash too much in time and it will last until the next bar, because we will recalculate the shares on the new bar.And the main multi-currency indicator could be something like this: https://www.mql5.com/ru/code/897
That's it! And the tricky thing is that a breaking trend can fool around several times - it can go up and then down several times.
This is the advantage of working on a bounce - statistically, the price rebounds from levels several times more often than it breaks through them.
When breaking the channel I, on the contrary, decrease the influence of correlation, so it does not prevent my earlier entry. In general, I see some psychological phenomenon here - people perceive the correlation to be similar to a crowd or herd - where everyone goes, we go. In reality, even if a herd moves in one direction, it is very chaotic within itself, as it consists of different kinds of animals. Therefore, it is more logical to focus on animals similar in morphology and, if they have reached some extreme state, expect them to return to the body of the herd, rather than to predict their movement based on the general average line of all animals.
For me, multicurrency means accounting for the behavior of many currencies, and not simultaneous trading on many symbols. For each instrument I construct a unique combination of correlations, ie one pair-one EA. It makes no sense to decrease or increase the share of an instrument in the portfolio if its correlations are in principle in conflict with the Guide behavior. I.e. I change the share based not on the correlation, but on its profitability.
This indicator, unfortunately, cannot be used.
I have made synchronization with some cool coders. I got it the second time. -) The price series classically works, too.
В этом и есть преимущество работы по отскоку - статистичеки цена отражается от уровней в несколько раз чаще, чем пробивает их.
yes, bounces are more likely to happen
but the bounces are more along the length of the stroke