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
I think that's not the way to use Hearst. I don't think it should be used to identify any event. It has to be counted on a decent sample and when that sample is formed it will be too late. As, in fact, you have written.
Also, if Hearst changes for a given channel on the same time interval when you switch to another t/f, then the sample (either of the two) is unsatisfactory. The Hearst value for a channel should not change. Maybe this can be used as a criterion for selection of samples.
IMHO
Vladislav didn't say anything about switching to different timeframes. I don't see any sense in running through different timeframes either. I understand that you are inventing your own approach to solving the problem.
On page 24 I posted a post:
An interesting observation. The optimal nearest channel on the watch is the one on 15 min on the border of curve (2006.06.09 01:15). Thus, you can either look for a series of channels on a shallow timeframe or the nearest channels on different timeframes. The confirmation is that the nearest optimal channels on the 15-minute and 30-minute timeframes are the same at the moment.
Getting the task right is the key to success. I've created one simple system based on two well-known tenets:
1) The price cannot be predicted.
2) The trend is more likely to continue than to reverse.
As it turns out, a lot of robust systems can be built on this, the basic idea of which will be the same - trading along the trend. Another issue is that the formulation of this case is a non-trivial task.
This file contains data of testing strategy with optimal parameters obtained earlier but on the sample that was not optimized as well as real trading data.
I can see a certain correlation of the curve shape. But we cannot say that the correlation is high enough. Also the most unpleasant thing is that curves direction is a bit different. We have incurred losses on the real account but almost zero losses on tests which cannot be considered as a proof of viability of this system - a system of random guessing based on the current data of bar parameters. Suppose you can imagine that the experimental results were not absolutely reliable because for the period there were about 5 blackouts / no connection to the server for 3 to 20 hours, which I think is only a significant disadvantage of the system (high sensitivity to inevitable external accidents that will happen again and again no matter what measures are taken to eliminate them). In connection with this experimentally obtained result, I can draw the following conclusions:
1. The number of deals optimized on historical data does not guarantee you anything in the future. That is, 3600 deals were performed on the history of 1.5 years, which could not guarantee the success of further 754 deals.
2. Testing data and real trading data diverge significantly, exceeding the allowable limit, which can still be considered as an error.
3. Consequently, it is obviously a dead-end street and loss of time and money to try and optimize parameters without having a strategy based on something more than just a wish "to take and try this". In this regard, the point of chasing full historical data without holes is a prerequisite for the "fit the history better than you could have done in real life with the real broker quotes that actually happened" game;o). I think that a strategy that really works will show you results already on those data, which are present on the broker's server (16k bars for each TF), and will show positive results in the future on the real market as well. You just look at the quality of entry points on the server and make conclusions. If the strategy works, the very first passes in the tester should show it. In this case, it would be more logical to change (select) the system parameters based on some logical reasoning, rather than on a simple arithmetic mean based on the results of the N-th number of deals on the history.
Perhaps these results can impress many people who were thinking in the same way as I did 3 months ago.
Although, if I am wrong in my conclusions, I will be glad to read the reports with arguments. Maybe the fact that my system lost 30% of my real account instead of increasing my deposit by 1.5 times within the specified period of time is just a coincidence and I should have waited for a few more months and then I would have had an estimated profit? So far I have doubts about the necessity of continuing to observe the depo plummeting by this strategy. That is why I stopped experimenting in this direction.
I think you are absolutely right on all 3 points.
And the question, it seems to me, is not how to improve the quality of testing or optimization, but in the strategy itself and its implementation.
So if you wonder why Vladislav's strategy led to such unhappy results, then the answer, again, seems to me to be in the details of implementation. Each strategy consists of a sequence of steps, and the probability of their cumulative positive outcome is equal to the product of the probabilities of a positive solution at each step. So in order to get the probability of an aggregate positive outcome significantly higher than 50%, you need to have at least a 90% probability at each step. It is enough that there is an eagle-eye in only one link and the total probability will fall below 50%, which is an inevitable failure.
If you remember, Vladislav has repeatedly talked about the plethora of different criteria that are used at different stages of strategy implementation. By skipping such a criterion in our implementation, we leave it to chance. This is what ruins the whole case. If it were enough to build several channels to correctly identify entry/exit points, it would have been done long ago. I think hope for elementary solutions should be set aside. Just like the hope that a kind man will come along and give us a money printing press.
If you have a sensible and, to a certain extent, well-founded ideology, then in order to make it work and be profitable you need to supplement it with intelligent management. After all, even if you have a car and fuel, but no reliable management, you still can't drive safely.
That is why we should take the situation as an opportunity for independent research. Vladislav spent two years on everything. We, on the basis of the fact that he shared his ideas, will probably need less time to bring them to the strategy. :-)
I think Vladislav's strategy, on the contrary, inspires a certain optimism, at least I have chosen it as my immediate plans. The first test on it was published by me on page 26. My last post was exclusively addressed to my strategy, which I wrote about on page 7-8. My strategy is a strategy for catching spikes in noise (a random guessing strategy), which has nothing to do with Vladislav's strategy.
In fact, the variant of the strategy I have based on the data of this thread is not very bright yet. If my variant of Vladislav's system has showed profitability of more than 6 on EURUSD, the result of GBPUSD is slightly higher than 0 with the same algorithm (the first half of the period has experienced depo growth and then the same loss). Now I am working on improving the order placing algorithm to try to get a positive result at GBPUSD as well. I think my Expert Advisor needs to go a certain way of modification to achieve positive results on different currency pairs before starting to use it for real.
1) The price cannot be predicted.
2) The trend is more likely to continue than to reverse.
As it turns out, a lot of robust systems can be built on this, the basic idea of which will be the same - trading along the trend. Another issue is that the formulation of this case is a non-trivial task.
Hi Rosh.
If you do not mind, I would like to communicate with you privately.
Please, send me your mail.
Sincerely,
Alexey.
My bad, I didn't get it.
Nevertheless, everything I wrote still stands. If you got a zero result on GBPUSD today, tomorrow it may happen on EURUSD as well.
So supplementing the ideational and computational aspect of the strategy with intellectual logic is something we all need.
If you don't mind, I'd like to chat with you privately.
Please let me know your mail.
Regards,
Alexey.
I will answer you, please send me your contact details on Alpari, or spider, or investor, or wyack. I get a lot of spam as it is, so I don't want to give out the address, sorry.
Привет, Rosh.
Если ты не против, хотел бы пообщаться с тобой приватно.
Сообщи, пожалуйста, свой mail.
С уважением,
Алексей.
Drop your coordinates in a personal message on Alpari, or spider, or investor, or viac - I will answer.
I get a lot of spam as it is, so I don't want to give out the address, sorry.
My address is the same as before.