Discussion of article "Developing a cross-platform grider EA (part III): Correction-based grid with martingale" - page 2
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
run the signals for a couple of months
From a mathematical point of view, this makes no sense as there is a Tester. Signals here can be useful only for marketing promotion.
...
In fact, I come to the conclusion that if there are two profitable equities, their combination (with weighting coefficients, the sum of which is one) gives a more attractive equity. Actually, gridder is a combination of several TS with the same coefficients, martin - with different ones. And the stupidity of these approaches consists only in the fact that in the portfolio of such combinations plum (first values of majics) summands are allowed. I.e. grider/martin is a special case of combining a portfolio of TSs with stupidity enabled (there are unprofitable ones in the portfolio).
It turns out that all "stupidity" is the result of wrong thoughtless entry. And all "not stupidities" are the result of correcting "at random" the previous "stupidity". All entries "at random", where luck comes after several unsuccessful attempts.
Conclusion - you need to calculate the inputs. If you use the average number of "foolishnesses" - just skip them and enter with alarger lot to level out the missed but unsuccessful entries. And if you break it down by timeframes - from smaller to larger, you can get rid of a small number of entries on the larger timeframe - missed entries on pullbacks of movement on the older timeframe will be filled with not missed entries from the younger timeframes, where this pullback for the older tf is no longer a pullback for the younger one, but a movement for successful profit taking.
Conclusion - you need to calculate the inputs.
There are two approaches to refills
The first variant should give a result not worse than the second one, because the second one is a subset of the first one. But the first variant is so huge that it becomes computationally impossible. And GA will give nonsense (hypothesis). So the second option is reasonable.
The author of the article seems to have used the first option. And if he did it through the GA, the nonsense input was present with high probability. If he did it via full overshoot, then it seems to be absent.
We need a toolkit, which shows the chart of profitability for each majic. For example, such a toolkit can be applied to the history of dolivotnyh Signals, showing the degree of presence of stupidity.
About timeframes, I don't understand the idea.
Signals here can only be useful for Marketing promotion.
And an indicator that the author has seriously worked on the topic and has not spared time/resources to test the idea in real life.
There are two approaches to top-ups
The first variant should give a result not worse than the second one, because the second one is a subset of the first one. But the first variant is so huge that it becomes computationally impossible. And GA will give nonsense (hypothesis). Therefore, the second option is reasonable.
I don't understand the idea about timeframes.
A flat to work in the channel on each tf has a different size. If on H4 it is a flat, then on M5 it is a sequence of decent trends. If on H4 we have five entries, four of which are unsuccessful, and the fifth one was successful and reached the opposite channel boundary with a good profit, and the previous 4 entries were on pullbacks and were taken down by the stop, then it is logical to move to a smaller tf, where the channel boundaries are smaller and entries are more frequent. I.e., on H4 we have eliminated four of five entries - the number of entries has become five times less. On M5, we also threw out four of five entries, but the entries there are more frequent, and every fifth successful entry on M5 fills the four thrown out entries on H4, only they will be with a profit, because they are closed inside the five-minute channel. Yes, with a smaller profit than on H4, but we have thrown out four unsuccessful entries on H4 and replaced them with four successful ones on M5, although with a smaller profit.
Timeframes are not a market object, but a completely artificial formation.
Therefore, the topic of timeframes is closed forever, if the TF is taken as an input parameter.
I.e. it is just an input parameter to be optimised. Well, since there is no sense in discussing input optimisable parameters, the same thesis applies to TF.
Timeframes are not a market object, but a completely artificial entity.
Therefore, the topic of timeframes is closed forever, if the TF is taken as an input parameter.
I.e. it is just an input parameter to be optimised. Well, since there is no sense in discussing input optimisable parameters, the same thesis applies to TF.
No. It doesn't. It is just one of the private methods for selecting the size of the working channel width.
You are reading the explanations, but you are not seeing in them what you are being told. Try to go away from looking at a timeframe to looking at another type of chart with smaller parameters - this TS will work differently on it than on the older chart. But about fills and the accuracy of their entries - it is just about the smaller timeframe, where the range of movement is smaller, the width of the channel is smaller and the profit is smaller. But in comparison with the older tf, the entries are more frequent. Hence it follows that in order to make losing trades on H4 profitable - they should be opened in a smaller channel, with a smaller channel span. And close them more often.
Draw two channels on the chart with your losing trades - big and small. For each of them - different parameters. For each of them - your losing trades. And see which profitable trades on M5 (for the small channel) coincide approximately with unprofitable trades on H4 - then you will understand what I am talking about. However, you always said that you don't need indicators - you keep everything in the tester and in your head. Then it will probably be difficult for me to explain my thought - I am more used to drawing than visualising. I don't have time for drawing :(
It turns out that all "stupidity" is the result of wrong thoughtless input.
The conclusion is that you need to calculate your inputs.
Unfortunately, the "conclusion" will be good only on history, I don't want to argue, but I know for a long time that the price doesn't owe anything to anyone, imho!
so it turns out that it is not "nonsense - the bottom line", but the bottom line is risk management (you want MM), you need to clearly define what you want from your TS - or a "chart at 45 degrees up" - usually it is an increase in risks or a chart in the form of a broken line - it is a risk = a series of losses.
in general, I have long ago told myself that the problem is not to place the order correctly according to the signals of the TS (in fact, fitting or you want to guess where the market will go...), and the main problem is to make the right decision, what to do if the signal of the TS turned out to be false (erroneous).
There are two approaches to the replenishments
every "5th member" of the forum has already done this, the problem will be that the main TS will wait for the grid of orders to work - i.e. the main TS will "sleep",
we need some universal solution to add a grid to the existing TS to each placed order - then there will be an increase in TS profitability, but it will also require an increase in deposit, risks.... etc.
ZYZY: the TS itself should be able to generate a convenient correct grid of orders in the tester.
ZYZY: wrote 10 times already grids of orders, the authors of the TS that only did not invent, and that I noticed that the reopening of already triggered orders (new orders in place of closed ones) usually significantly increases the profitability of the grid...to accompany such grids is a troublesome matter )))))
No. It isn't. This is just one of the private methods of selecting the size of the working channel width.
Discussion of TF here is offtopic, of course. Perhaps it is reasonable to put it in a separate discussion.
TF is an input parameter of any TS. The fact that it is not assigned as an input does not make it not an input parameter. But for some reason you do not perceive it as an input. Probably, you do not perceive the trade symbol as an input either. Although it is also an input parameter.
In my opinion, TFs and symbols have nothing to do with this topic.
Unfortunately "conclusion" will be good only on history, I don't want to argue, but I know for a long time that the price doesn't owe anyone anything, imho!
so it turns out that not "nonsense is the bottom line", but the bottom line is risk management (you want MM), you need to clearly define what you want from your TS.
Let's clearly define what Stupidity is. Stupidity is when optimising in the Tester to put into the portfolio a TS that showed a loss on the tested interval. This is an absolutely clear wording.
Saying that Stupidity is done on purpose because we can't be sure that it won't show a profit on OOS, of course, we can. But absolutely any approach can be justified that way.