Discussion of article "Developing a self-adapting algorithm (Part II): Improving efficiency"
The paper Development of a self-adaptive algorithm (Part II) has been published : Improving efficiency:
Author: Maxim Romanov
Good idea with correlation, I will take it into consideration. About opening on each candle, it reduces the overall profitability, you need to open, say if the fan on BUY only if the price has gone below, the next position is the same. On SELL sovetvetvetstvo the same with the exact opposite. We exploit waves. We know that there will be a reversal, and if we know, why make unnecessary entries? It is better to increase the lot towards the end, to work on the grid principle. Although maybe I don't understand something, but it seems to be simple. I need to write one for myself )) time will be I will do it ). By the way still works on a real account ? ) . I think there and up to 100 per cent per annum can be safely raised by making modifications appropriate, and maybe all 200. The system gives such opportunities. By the way, the longer the fan, the percentage of rollback can be reduced relative to the movement, so you can avoid a very large number of drawdowns, only have to build up lots towards the end to ensure this, but if you remove the unnecessary positions that I said here and a lot can appear additional for this ).
Good idea with correlation, I will take it into account. About opening on each candle, it reduces the overall profitability, you need to open, say, if the fan on BUY, then only if the price has gone below, the next position is the same. On SELL sovetvetvetstvo the same with the exact opposite. We exploit waves. We know that there will be a reversal, and if we know, why make unnecessary entries? It is better to increase the lot towards the end, to work on the grid principle. Although maybe I don't understand something, but it seems to be simple. I need to write one for myself )) time will be I will do it ). By the way still works on a real account ? ) . I think there and up to 100 per cent per annum can be safely raised by making modifications appropriate, and maybe all 200. The system gives such opportunities. By the way, the longer the fan, the percentage of rollback can be reduced relative to the movement, so you can avoid a very large number of drawdowns, only have to build up lots towards the end to ensure this, but if you remove the unnecessary positions that I said here and a lot can appear additional for this ).
There is a mode where positions will be opened only at a price lower than the previous opening price (buy positions).
Apparently you have found more profitable and stable algorithms, if not secretly hint in what direction to dig ?
Apparently found more profitable and stable algorithms, if not secret hint in what direction to dig ?
I went the way of refusing from optimisation. The algorithm should understand by itself what parameters to use to work on any instrument at any time. I will write one/two more articles, I will tell you how and show the results. I don't like optimisation, it's an eternal war with undertraining/overtraining.
Good article and the ideas are interesting. There is a lot to learn from you.
I will wait for the continuation.
Good article and the ideas are interesting. There is a lot to learn from you.
I will wait for the continuation.
Thank you, I have travelled the path myself, learned something, I share my knowledge to make it easier for others. There are too many superstitions instead of knowledge in trading. Because few people share real knowledge and experience is not accumulated.
I went the way of refusing to optimise. The algorithm should understand by itself what parameters to use to work on any instrument at any time. I will write one/two more articles, I will tell you how and show the results. I don't like optimisation, it's an eternal war with undertraining/overtraining.
You are not the only one who doesn't like it ) I initially as soon as I started to look askew at it, imho it's just a tweak. I only turn the settings by hand. Plus the sample should be 10 years at least and the ratio of the number of trades to the number of bars is also good, at least 0.05, otherwise we just encounter randomness. Self-adaptation is good, but it is a very voracious pleasure)). Optimisation should be simply divine )
2,337 settings???
What kind of adaptation is there...
By the way, the prefix "self" to the word "adaptation" in the phrase "adaptive algorithm" is completely inappropriate and cuts the ear.
An adaptive algorithm is an algorithm that adapts to the current internal and external conditions of functioning and is designed to fulfil the task at hand and achieve the goal. And the main thing is that it does it by itself.
2337 settings ???
What kind of adaptation is there...
By the way, the prefix "self" to the word "adaptation" in the phrase "adaptive algorithm" is completely inappropriate and cuts the ear.
An adaptive algorithm is an algorithm that adapts to the current internal and external conditions of functioning and is designed to fulfil the task at hand and achieve the goal. And the main thing is that it does it by itself.
It doesn't know how to adapt yet. It adjusts some parameters only. Full-fledged adaptation will appear in the next article. And about the spelling of the word self-adaptive, thank you for your comment.
1) The point of the question was: why do you need such a huge number of settings? This number is obviously excessive. And who is going to understand them? Who will be able to understand them? Not superficially, but thoroughly. Can you do it?
2) Please. Same with the adaptive system.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
New article Developing a self-adapting algorithm (Part II): Improving efficiency has been published:
In this article, I will continue the development of the topic by improving the flexibility of the previously created algorithm. The algorithm became more stable with an increase in the number of candles in the analysis window or with an increase in the threshold percentage of the overweight of falling or growing candles. I had to make a compromise and set a larger sample size for analysis or a larger percentage of the prevailing candle excess.
The article describes only the most basic and interesting modifications and operating modes. In reality, much more has been implemented, and all modes can be combined with each other. The requirements specification for the algorithm with all the details is attached below.
I am going to run the tests on the same currency pairs I used to test the first version of the algorithm in order to visually highlight the differences. Like the first version, this algorithm works by closing candles, so you can safely test it in the "control points" mode. Inside the candle, it only controls the current profit and makes sure that the current funds do not fall below the threshold value defined in the settings. As before, the tests will be carried out with an overestimated spread. I will set a spread of 40 for GBPUSD.
Like in the first algorithm version, we can trade and optimize any timeframe. The minimum timeframe is limited by the size of candles relative to spreads and commissions. The lesser the timeframe, the higher the requirements to the signal quality and expected payoff. As the timeframe increases, the size of the candles grows and, accordingly, the drawdown level rises as well. Therefore, the maximum timeframe is limited by trading style preferences.
I have performed optimization in 2017 when using the robot for trading on real accounts. Therefore, I simply took the previous settings without performing a new optimization.
Figure 7. GBPUSD H1 2000.01.01 - 2020.12.08, static lot
Author: Maxim Romanov