EA N7S_AO_772012 - page 58
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
Flop or no flop, there are loss-making weeks. Every fourth or fifth week is a loss.
In 19 weeks of general testing on demo with different versions and number of signs I have had 5 losing weeks, including two weeks with -$21 and -$6 ($1=1py). To be fair, there was a week with a force majeure -$400 and two more -$370 and -$190. With the positive weeks I got confused in the accounting with the move to 5 digits, but the balance seems to have almost tripled in the end from a starting $2000 to $5780. Average profitable week $340. Average profit about $200 per week, i.e. we can talk about 8-10% per week with a maximum drawdown of $400 (20%).
What to do? ANALYSIS! Look at the charts. There's a reversal in many instruments, plus relatively weak volatility and a trimmed week. I think I would have traded no better with manual trading.
If there are any doubts, I'll reiterate and share my observations.
IT IS NOT SO IMPORTANT TO SELECT PARAMETERS AFTER THE OPTIMIZATION,
IT IS MORE IMPORTANT THAT THE OPTIMIZATION ITSELF HAS A SIGNIFICANT AMOUNT OF POSITIVE RESULTS, PREFERABLY MORE THAN 70 PERCENT.
IN ORDER TO ACHIEVE SUCH RESULTS IT IS NECESSARY TO CHOOSE A RANGE AND EVEN A BASIC INDICATOR FOR EACH PAIR, DEPENDING ON THE MARKET CONDITIONS.
One more small observation
WHEN WORKING WITH EXPERT ADVISORS USING NN (NEURAL NETWORKS), SUCH AS THIS ONE, STRIVE TO USE A GREATER NUMBER OF INSTRUMENTS.
Another silly question: can I use the old sets for the next week if the current week traded well?
I.e. the optimization is actually confirmed and the week that traded well is kind of a good forward?
As they say: good deeds never do good.)
It is interesting to know the opinion of the author and other gurus.
What exactly should be done for this purpose, everyone decides for himself.
For example, the simplest one is to prohibit trades in the direction of maximum loss, i.e. if Trd_Up_X is optimised well, and Trd_Dn_Y is very bad, then after optimisation we experiment with these values. Other variants of actions are if you have desire and time. We change the scheme to 3&2 (or even 2&1), perhaps the market has changed dramatically, i.e. both the first and the second range are shiftless and shorter. Make sure that the number of deals is acceptable (not 1-2 per week and not 50))). Finally, if you are confident in the strength of your PC, then try other indicators (parameter Indctr), maybe in the current market, it may be that any other indicator works better. Before optimization, you can visually evaluate different indicators, including their parameters, play with them.
Another silly question: can I use the old sets for the next week if the current week traded well?
I.e. the optimization is actually confirmed and the week that traded well is kind of a good forward?
As they say: good deeds never do good.)
I am interested in opinion of the author and other gurus.
Yes I did. Last week I had sets that had been running for a fortnight in a row, i.e. last week was the third week for them. This was a forced measure, with the sets working again in the failed previous week as well. This gave me an idea of what to do if the week was plus or minus, and whether to keep the setting. Theoretically I should get fresh data, so I cannot give a clear answer. It needs to be tested.
My opinion - it is necessary to fix profit in some other way... too long deals float in Shuter versions... I am trying this "tool", though of previous versions, slightly modernized by me (with another indicator) and a different output (added trawl and position closing every day) on CFD... I haven't got enough history to share, but first results are very satisfying =)
Followed the topic from the beginning, I think the Expert Advisor is really worthwhile!
For some reason, I must have over-optimised it "a lot".
variant 4+2
The first forex week is losing often
tried EURUSDm5, EURGBPm5, NZDUSDm5
since the beginning of the year
I think I may try to build revision deals into this EA
i.e. i optimise as described in the brochure and forward is 1-4 weeks, but reversed
i.e. buy and sell and vice versa
switch off trawl
Eurobucks set from 04012009 - 14022009
forward from 14022009 +3 weeks
i have a -18 on the forwards from this revier to make a possibly profitable
i can't get a ratio of 7 profitable to 1 losing trade ....
To the entire community of profit-seekers and sufferers, hello!!!
Yesterday I spent the whole day reading the thread and studying the Expert Advisor. It is a nice EA, respect to the author.
I am new to this platform MQL4, although I have experience in programming, and in particular in NS projects.
I would like to make some comments, which may be helpful in further development of the project.
Constant retraining leads to more and more new presets and does not lead to an accumulation of already acquired information.
From my experience I can say that neither indicators, nor NS can indicate reversal points with a high probability, but only signal about such a possibility. Therefore, the drawdown and possible losses are not curable. And they will be even larger if the Expert Advisor will not be able to identify the change of the trend direction, different from a trained sample.
I see only one way out - to enter switchable presets, depending on the readings of the senior chart. Or, better said, finding prices in +/- trend or flat.
For this purpose we can introduce one more NS, working with the large frame indicator, and conduct its training in a separate step and on a much larger sample than for training of the shared presets and do it not so often. After the training its signals should be used for teaching the appropriate preset, which should be at least three - it should be a flat, + trend and - trend, and probably more, well, something in addition similar to -correction and +correction. They will have to be used by the signals of the senior network.
It is true this approach will require rewriting the engine, but it is worth doing.
I have run the Expert Advisor on various symbols and faced the situation of "unlearning". But in my view, it appears in this version when falling into a trained sample of market reversals where the trend detection approach used does not give signals for action and reduces the number of deals to zero. Perhaps this can be corrected by some, insignificant!!! increase in network signals.
Taking the above into account, I suggest a learning algorithm consisting of several stages as well, at least one for the network that recognizes trends and at least one for each preset. I do not think the learning time will increase. But the frequency of retraining should decrease, and ideally just fall away for a well-trained pair.
Putting additional bows in the form of MM and processing of errors of order execution is a purely personal matter for each user. Do not waste your time on it.
The "grail" of this version is still very, very far away. And continuous retraining is unlikely to allow for such a chance to reach the 2009 Championship.
But you still have time)).
Profits and lots of them!!!
How about this expert work, for example:
We have a history of sets, we divide these sets into three groups: "worked in an uptrend with profit", "worked in a downtrend with profit", and "worked in a flat with profit". We make 3 copies of the Expert Advisor with different majors and put them simultaneously on the same (positively proven) instrument - or on several can.
Then it can happen that only one of them will work correctly and bring profit that does not cover the losses of the other two.
Still, we should strive to switch these sets depending on the market situation. But it may require some additional efforts in the code.
Without the intervention, I suggest according to your scheme to hang three EAs and switch only the one that we consider most suitable for the market. But not simultaneously!!!