Discussion of article "Developing a cross-platform grider EA (part III): Correction-based grid with martingale" - page 4
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
Used the virtual, loved it. It didn't start with half a turn, though, but I always have a file handy.
It depends a lot on the tasks to be solved. This task, I think, should be easy to manage.
My experiments on this subject have shown the following.
Getting ahead of the curve. Thanks for the cool comment!
Fractional orders, of course, gave a higher profit, but they had a worse ratio of profits to losses (which is also natural - all the main trades closed on SL were duplicated by fractions, which were also closed on SL, but not all profitable main trades were duplicated by fractions, as the price simply did not touch them).
That is, the losses, though smaller in absolute terms, were 100% taken by share orders, and profits - only partially. Because of this, it made no sense to skip the main trades.
I used to think like that myself. As if without dolivoye it is lost time, when it is possible to earn. But this is actually an explanation that is comfortable intuitively, but not mathematically. And mathematically, if the main entries will give a negative result in total, then they are definitely not needed.
Essentially, this distance from the main trade to the refill is a function of the quality of the main strategy's entry point. If the distance is 0 (the result does not improve with fills), then the entry points are perfect. If the distance is large, then there is something wrong with the signal.
I screwed myself when exactly this kind of performance was ruined by backtests. It would seem that if the same channel is not perfect, tweak it by 1-2 pips and the result will be better with hundreds of trades. But no way. I put very tricky variants of doles only as a study. And I expected to see something quite different from what was theoretically imagined. It turned out that if you refill in 5 pips (half a point of four-digit), the FV increases by almost 50%. But if you refill at a greater distance - badly. This is a peculiarity of channels, which were calculated at their optimisation limit. We will have to repeat the experiments.
And the conclusion from all this is the same - the fills themselves (with or without martingale) do not make any sense. If the strategy is good (the entry points are perfect), refills will not improve the result. And if the strategy is bad, they will not improve the result anyway (but they can temporarily pull it to the plus side due to higher load on the deposit and drawdown).
The factor of uncertainty plays here. And it is necessary to talk about the portfolio of the TS. A replenishment is a new TS in an implicit portfolio. Portfolios are created because of the uncertainty principle. Yes, at the stage of portfolio selection, all TSs show great results. Moreover, together they make equity very attractive. But the main thing is that you realise that some will break, and some will shoot. Which ones - you don't know. But the portfolio reduces the probability of all of them at once.
And mathematically, if the main inputs add up to give a negative result, they are definitely not needed.
Well, that's the conclusion I came to at the end. You can't improve a bad strategy with additions, but you don't need it without them.
It turned out that if you refill in 5 pips (half a point of four-digit), the FV increases by almost 50%. But if you refill at a greater distance - badly. This is a peculiarity of channels, which were calculated at their optimisation limit. It will be necessary to repeat the experiments.
Well, the channel has not been optimised just by these 0.5 points. And it is necessary to finish it, not to add a refill.
Although, if we "smudge" the entry point by detecting the reversal, we may do better. The only question is the lot distribution between all trades.
The factor of uncertainty plays here. And we should already talk about the portfolio of the TS. A replenishment is a new TS in an implicit portfolio. Portfolios are created because of the uncertainty principle. Yes, at the stage of portfolio selection, all TSs show great results. Moreover, together they make equity very attractive. But the main thing is that you realise that some will break, and some will shoot. It is not known which ones. But the portfolio reduces the probability of all of them at once.
Then it is more logical to add fundamentally different systems, other instruments, or other, very different, parameters of this strategy to the portfolio.
The correlation with the main system is very high, and any black swan will take the whole portfolio to a steep peak.
Then it is more logical to add fundamentally different systems, other instruments, or other, very different, parameters of this strategy to the portfolio.
The correlation with the main system is very high, and any black swan will take the whole portfolio to a steep peak.
I have had many situations when 1 pip did not reach the rollover, for example. And it significantly affected the result. That is why "smearing" always had a positive effect.
Moreover, even if it did, there might not be enough liquidity or for some other reason there might be a re-jack. In such situations one realises especially well how important is the portfolio of TC even from different sets of input parameters of the same.
Well, just by these 0.5 points the channel has not been optimised. And it is necessary to finalise it, not to add a refill.
I added 1 pip artificially. The result was worse.
There are a lot of controversial statements in the article, starting from "the best results in the Forex market are shown by geometric increase of lotness in the chain" and ending with "the best performing financial instruments are USDCAD, NZDUSD, SBUX, XOM, INTC, CMCSA, PG".
Firstly, where did such bold claims on lotness come from? My experience does not confirm it.
Secondly, the most successful according to the results of testing most strategies (many years of experience in testing various TS) is the pair "EURUSD". But for some reason it is not in the list of used ones.
Therefore, despite the interesting and lively presentation of the article, you should be careful with general conclusions.
I have always been against such systems. Especially if the lot increases more than 1.5 times.
For some reason nobody pays attention to charts where the profit is up to 7% for several years. What are you discussing here?
At the same time, everyone forgets that the tester finds the entry and dosage parameters, so as not to lose, i.e. there is a complete adjustment to the history.
Let me explain on my fingers. For example, there is a news spike on which such robots lose. So the tester will choose such parameters to skip this hairpin, as a rule, the start of this hairpin will be from the initial lot. And if next time this hairpin will be say on the 3rd knee or on the 5th. And that's it ! Hi Kolyan !
The second reason is that on a series of trades we earn little and lose a lot.
Thank you for the excellent work! It is very much appreciated.
--Dave
Every time I try to test this on historical data I get:
2019.08.17 20:38:02.707 2018.11.18 22:23:30 failed market sell 0.01 XXX [Unsupported filling mode]
It happens for any symbol I choose (FX or Stocks). No trading is performed. What could be the problem?