a trading strategy based on Elliott Wave Theory - page 196
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
Получается, что чтобы добиться Р>80%, нужно аж 8 индикаторов с р=0.55, 4 с р=0.6 и только 2 с р=0.7.
I think, in fact, the more indicators, the worse.
I cannot agree with you.
Neutron' s post shows this very clearly. Apart from profitability there are also things like drawdown. At predictive validity р=0.55 the probability of large drawdowns is very high. If the indicator produces its signals quite often, then it would be a good idea to limit it with another indicator to reduce the number of trades and increase the likelihood of a positive outcome.
Imagine an indicator that gives on average one signal per 5 bars. This is one trade a week on D1. And on M1? This is a direct violation of the MQ (and, presumably, brokers) requirements of a reasonable number of deals per unit time. For such an indicator on M1 it is not bad to have a filter, which a lot of people by the way are looking for. This is that second indicator !
Like this - http://forum.alpari-idc.ru/thread30284.html
At how many tables to play (or how many lots to buy) is determined not by my desire, but by my risk capital. And I care about investing that capital with 55% or 80% reliability.
I think instead of arguing about it, we could talk about predictive validity and methods of its determination at least for standard indicators, for example.
I totally agree with Rosh. That's why I've been trading all broker's currency pairs at once for 4 months with 19 lots at a time, so as not to violate MM principles for the deposit as a whole.
Yurixx, I am not going to argue, but the question is interesting enough and I cannot leave it just like that. And I want to ask you and Neutron some details. Especially as the results diverge greatly from my modest experience (it's certainly not a disaster).
I guess (may be wrong) what I meant before is the total probability of prediction of a group of indicators under the condition of all indicators being activated simultaneously in a group. It was supposed that every indicator is independent and able to make prognosis (ATR, for example, can't do that without additional logics, or am I mistaken?).
First of all, I'd like to describe the experiment conditions, if Neutron doesn't mind my asking.
1 what indicators were used
2 what was the condition of simultaneous "triggering". Was it a time range within which after the first one in the group was triggered the system waited for all the others to arrive, or was it forced to query the status of the others.
Or was the Monte Carlo method simply to simulate the triggering of the indicators?
PS:
Imagine an indicator that gives an average of one signal per 5 bars. This is one trade per week on D1. And on M1? This is a direct violation of the MQ (and, presumably, brokers) requirements of a reasonable number of deals per unit time. For such an indicator on M1 it is not bad to have a filter, which a lot of people by the way are looking for. This is the second indicator !
This is of course the second indicator, but what does it have to do with simultaneous triggering of the whole group?
What difference does it make to me to buy 10 pairs at 0.1 lot or 1 pair at 1 lot, if p=0.55 for all instruments?
Just for the sake of diversification, to reduce possible drawdown? It makes no sense.
The slope angle of the profit growth curve will not change, and therefore the rate of equity growth will not change either. But p=0.8 will give a completely different picture.
Moreover, I doubt that any indicator will have the same p-value at all pairs. The nature of price movements is different, and I suppose p will also be different. In this situation it's better to choose the pair where p is maxed out rather than diversifying.
Sergei, relax. :-)))
They weren't indicators, but their models, which had some known properties: lack of correlation, constant p. And the experiment itself was based on the Monte Carlo method as a source of numerical series. No real quotes were used.
And the question was numerically abstract: how to determine a priori cumulative probability, if there are several correlated sources to estimate it.
What Neutron got was the result of a numerical experiment with independent sources, so my question remains unanswered for now. Only the fact that the joint use of indicators in principle increases the accuracy of the predictive estimate remains positive and significant.
Grasn, we used the same-type non-correlated indicators in the code, and Monte Carlo method just simulates their operation. All indicators were forcedly polled on each bar, and if the signal for market entry was available for all of them simultaneously, a position was opened. Then the amount of successfully opened positions was calculated and related to the total amount of open positions. In this way validity of the P-forecast by a group of indicators was determined.
But why do you think our results differ?
You wrote:
Rosh, I wrote:
I wrote a post above:
So we all, practically and theoretically, have come to the same result:
There must be as few indicators as possible, it's better to have one.
Yurixx, however, invented a situation when an entry was made almost at every minute and justified the necessity of the second indicator, but not by increasing the profitability of TS, but by limitation of trades frequency imposed by brokerage companies. This is a slightly different case.
What difference does it make to me to buy 10 pairs at 0.1 lot or 1 pair at 1 lot, if p=0.55 for all instruments?
Just for the sake of diversification, to reduce possible drawdown ? It makes no sense.
The slope angle of the profit growth curve will not change, and therefore the rate of equity growth will not change either. But p=0.8 will give a completely different picture.
Moreover, I doubt that any indicator will have the same p-value at all pairs. The nature of price movements is different, and I suppose p will also be different. In this situation it is better to choose the pair where p is maximal and not to diversify.
Here, you, are missing a very important point:
Yes, the returns (in PUNDS per unit time) will not change. Yes, the drawdowns will decrease (as 1/SQRT{n}, where n is the number of pairs). But, because of this it is possible, without exceeding a predetermined market risk, to increase the capitalization of open positions! And it dramatically increases the profitability (in $ per unit time) of multicurrency TS compared to any single-currency TS from this portfolio.