The "Maybe we'll get lucky" counsellor - page 7
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
Forgive me, but do this - set the terminal from the 1st to the 6th month and find the best variant of the parameters in the tester. And run those parameters from the 6th to today. Picture in the studio.
I don't insist on anything, especially since the orders in this EA are not placed randomly, as I would like them to be, but in strict accordance with the principle: BAY is followed by SELL and vice versa, but I have fulfilled your request anyway. Test from 01.01.2011 to 30. 06. 2011, forward from 01. 06. 2011 to 24. 12. 2011 TF H4, lot 0.1:
Once again:
There is a system that produces roughly equal frequencies of profit and loss on 700 trades - no matter how its parameters change. Right?
You want to know what are the chances that, having checked all possible variants of parameters (10,000 variants in total), we won't encounter a single case of "640 times 60 and worse"?
I can estimate the probability, but what do you need it for?
2 yosuf: You are very persistent, to put it mildly. What is the SL/TP ratio in the last figure?
Yusuf, well, this is getting to be child's play. You are only self-deceiving by the ratio of the number of profitable and losing trades, without considering their size.
OK, pf = 81/(3*8.72) = 3.1. If there are 10 losing trades instead of 3 (easy!), the system will become unprofitable.
Then let's slightly modify the task, namely, we need to find an Expert Advisor that will randomly place orders by the principle "eagle-reckoning" according to their direction with TP=SL+spread+commissions+N points.
Let's say there is only a trend in one direction on a section of the chart. Up or down, but we don't know in advance where, but we know that there will be no trend change in that section. The simplest mathematical model of a trend is random wandering with drift. To capitalise on this knowledge, you can trade a system with fixed TP and SL. Take Profit should be higher than Stop Loss and trades can be opened on a coin. But the best solution would be to change - first buy, second sell, third buy etc. That's enough.
The second situation - we know that the market is in flat. The simplest model of a flat is a random walk where the RMS grows slower than the root of time. Horizontal flies like a sinusoid are special cases (RMS does not increase with time at all). In this case the ideal would be mean reversion - trading from overbought oversold areas towards the mean. For example, systems a la Bollinger Bands. But if not ideal, the entry can be done at any time and in any direction. The randomness of the entry can be compensated by Take Profit and Stop Loss values. The "wrong" entry, the smaller take profit and larger stop loss should be set. But we can assume the worst case and take a system with fixed TP<<SL and we will still be in profit on such section even on random entries.
In general, systems where the direction of entry and/or the moment of entry do not matter. But only on the sections of the chart with certain properties. I.e. it is necessary to know beforehand that prices will have these properties for some time. It is stupid to expect that it will be a long period of time of several years. Of course, this may happen purely by chance. You need a filter - when to expect the market to have the desired property.
Let's say there is only a trend in one direction on a section of the chart. Up or down, but we don't know in advance where, but we know that there will be no trend change in that section. The simplest mathematical model of a trend is a random walk with a drift. To capitalise on this knowledge, you can trade a system with fixed TP and SL. Take Profit should be higher than Stop Loss and trades can be opened on a coin. But the best solution would be to change - first buy, second sell, third buy etc. That's enough.
The second situation - we know that the market is in flat. The simplest model of a flat is a random walk where the RMS grows slower than the root of time. Horizontal fluxes like a sinusoid are special cases. In this case the ideal would be mean reversion - trading out of overbought oversold areas towards the mean. For example, systems a la Bollinger Bands. But, if not ideal, entry may be at any time and in any direction. The randomness of the entry can be compensated by Take Profit and Stop Loss values. The "wrong" entry, the smaller take profit and larger stop loss should be.
In general, the systems with the entry direction and/or the entry moment do not matter. But only on parts of the chart with certain properties. I.e. it is necessary to know beforehand, that some time prices will have these properties.
Yusuf, this is a kindergarten situation. You are only fooling yourself by the ratio of the number of profitable and losing trades, without taking into account their size.
OK, pf = 81/(3*8.72) = 3.1. If there are 10 losing trades instead of 3 (easy!), the system will become unprofitable.
The first option you mention, this EA implements, but the second option needs to be added here and tried on different parts of the market. Convinced that the market is more random than regular, so apparently the approaches should be random and we should accept the randomness of success, I think. Let's see how the market responds to random entries in terms of timing and direction. What will be the probability of a positive outcome. Of course, it will be related to the size and ratio of TPs and SLs that need to be optimized. The main thing is that if something turns out to be successful, we won't have to "guess", calculate, assume moments and directions of entering. In the future it is possible to try variants of random TP and SL setting, but I don't know how to practically implement it.
We will definitely need a filter - when to wait for the necessary property from the market. In the most primitive variant it is a trend/float indicator. On average, the market compensates for trends with a flat. I.e. if we trade only a flat or a trend system without filtering when to trade, profit obtained in one section will be compensated by loss in the other one and equity will randomly wander with a drift (spreads, commissions, etc.). The simplest filter is the time of day - when it is trending and when it is flat. There are other filters as well. Without them it's all a dabble :)
Quite right, but I have shown the optimised case, and there are many variants that give almost the same results, I have chosen one of them. Generally, I believe that the system that needs optimisation is doomed from the start, so I will try to set all parameters randomly, I think there should be more stability in this system, although the success will be random. The market leaves us no other choice. Of the regularities, only trends are possible.
It does not matter whether you optimise the system or set parameters randomly. (You can't predict market behaviour by the "colour of the jackets")
The initial error is that you do not try to build hypotheses and a logical relationship between price changes and your signals.
If a trader does not think about a causal connection between signals and price change, he or she will experience the following.
He or she will search for various algorithms that bring profit.
First, it will understand that it needs forward testing, and then it will probably need preliminary testing on a demo account.
And the Expert Advisor will show no better efficiency than the deals made at random.
Then it will complicate criteria for selection of the best strategies.
And each time he or she will be convinced that these criteria do not work, and then the programmer enters an "endless cycle" complicating the algorithms and selection criteria, until the qualification is enough to evaluate his or her own creation.
A side effect of his activity is PR among other seekers, which in principle can attract investors, but this is no longer a science - it's the art of selling.
The market leaves us no other choice. Of the patterns, only trends are possible.
Of course, if you only do algebra for a set of quotes, you will still end up with a quote. And it's like pulling one bar from the past and asking - Which way did the market move after this bar in the past? ))) The same thing will happen on a zero bar. You just never get above a lot of QUALITY quotes to get a QUALITY picture of the market. And the very notion of a trend will also fall apart. Because even those definitions that have already been pulled in by the ears (such as alternating lows and highs) are local in nature. Therefore the profit will be local too. And thus the desire of money will again and again return you to search for the highest possible expected payoff for a particular TS, so that your "avssoy" would happen more often. )))