Scold :) Interested to hear your opinion regarding... - page 22
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
2 Prival: Seryoga, PPP UUUU PPP UUU is not a Bernoulli scheme (again, I have rolled back the ideas of my article).
Non-Bernoulli systems definitely exist - e.g. Lucky with the number of simultaneously open positions greater than 1. But you can't take advantage of that. The question should be posed from another plane: how to turn a Bernoulli system into a non-Bernoulli one, i.e. turn trades into dependent ones, and so be able to use them?
For example, here's an example: we have two strictly Bernoullian systems X and Y. Is it possible to learn how to combine them somehow (Z = X*Y) so that the result of their combination becomes a non-Bernoullian system? The question is not stupid at all, unexpected solutions are quite possible.
That it's not Bernoulli, yes. I knew you'd see it straight away. And you know exactly what's behind it. And the test that you conceived, if I understood you correctly, you conceived for a reason :-)
My opinion is to get from two strictly Bernoulli systems, a non-Bernoulli system is impossible. It's like operations with two normal distribution laws, whatever you do you will come to it.
The only true way I think, is to create mathematical model of this curve, what I see on the screen, if it (the model) will be adequate, it will allow to predict the direction of movement and accordingly to achieve what we aspire to + get synthetics.
Я влез в эту тему, чтобы поддержать мысль о том, что адаптация любой системы с фиксированными SL/TP - простой, но часто не верный подход.
And in some cases it is even disastrous. And sometimes it can even be described as fitting, with essentially no justification whatsoever.
Thank you, of course, for the profound moralizing. I will use them for sure.
I was asking a very specific thing: how to determine whether TC gives a statistical advantage and, in particular, how you do it. I think it is impossible to be more specific. - Recently I came across a description of cultural differences. I quote: "This peculiarity manifests itself in ordinary conversation with the Chinese: to what appears to us to be a perfectly direct and precise question on a minor topic, the Chinese thinker gives an unexpectedly lengthy answer." Apparently I am Chinese, for your question to me is not at all concrete, but conceptual. So I have a conceptual answer to your question - read the bible on statanalysis and follow it. Or are you asking me to expound that bible right here? I have no secret or magic recipe, I have read the same statanalysis books as you. I understand that all the right concepts sound simple, but are hard to follow. The Ten Commandments contain no more than a hundred words, the instructions for using a phone are dozens of pages long. I'll try to fit it into a hundred words. :)
I think that it is important to start with the consideration of the phenomenon that we are going to exploit. It is necessary to understand what phenomenon we are going to study and how this phenomenon must give us the desirable statistical advantage. We need to start from reality, from the property of the market, not from speculative assumptions or the desired behaviour of the market.
From this example I understand that you are talking about elementary testing of TS on available historical data. This variant is known to everyone, and everyone uses it. So do grail-writers that get significant "statistical advantages" of their systems, especially after optimization.- Well, I disagree: they do not take advantage of this variant. Firstly, in tester optimization it is not the market property and its multiple realizations that should be studied, but one single curve in a given area. The profanation comes at the moment when the overbidding of takes and stops creates an illusion of finding a statistical advantage. Therefore, I am in favor of disassociating myself from this type of "research" into a single implementation of a curve "probed" with takes and stops, which are not a property of the market. I've never seen, that it is written in books on statanalysis - flip a coin 100 times, write down the sequence, but don't count the realizations of heads and tails, and try to write on paper your sequence and the hundred letters O and P, ideally if you go through all possible options, then choose the option, which best corresponds to the sequence in the experiment with the coin, - and become proud, because you have the statanalysis result in your hands! I am categorically against this kind of "statanalysis". Therefore, I do not agree that everyone uses the well-known, read, book version. I am becoming more and more convinced that no, not everyone. Far from it.
Probably needless to say, the market is changing and just because TC makes money today doesn't mean it will continue tomorrow. In that sense, how do you assess the credibility of the statistical advantage you have discovered ? I don't mean this idea of yours, it's not a strategy after all, it's just an illustration. I mean the fundamental point that concerns all expert writers. Can you say something concrete about it? - That TS doesn't make money tomorrow is the result of banal fitting of the TS action sequence to a particular curve, and it has nothing to do with statanalysis. Statanalysis, on the other hand, gives an answer to how much you can trust the identified pattern, including how much you can trust fitting the answer to a single realization. I like numbers from 1000.
What about the sufficiency of the statistical data to get reliable results ? Is 1581 bar enough ? Is 30 years enough ? - Yes, it allows to draw the right conclusions at a glance. In the case of smaller numbers the statistics gives us the opportunity to calculate the confidence interval.
Having a statistical advantage is central to any TS. And I asked you my question because I don't know any other approaches besides basic history checking. I thought that since the man is so confident about it, perhaps he can offer something more substantial. - I completely agree with you, there is nothing but trivial statistics. I would be a dreamer, engage in self-delusion and join the ranks of graaleoptimisers if I suddenly suggested something more substantial. I am wary of the words "elementary history-tested", as any grail "history-tested" can be hidden behind them. But the statistical advantage that should be a central point of any TS should be searched only by statistical methods, not by profanity, as I gave an example above.
You guys are interesting :) . You're also referring to mate theory. You'd do well to read more carefully. You are given an example of PPP UUUU PPP UUU... the number of outcomes is 50/50 . The system is profitable and it is impossible not to see it. As you said one-way logical conclusions :-) class. If you can't see it then that's probably it - no logic at all.
Once again, be precise in your wording since you are referring to mate statistics.
These are completely different things. For the first example is given (the Grail in its purest form is given, and do not need 57 or 98% of positive outcomes), except that the second is not satisfied with 50/50 probability of transaction, because with this TS, the probability of "guessing" is 100%.
Nothing prevents, knowing that the next 3 trades should be unprofitable, to go the other way. I.e. we will have TS in which all 100% trades are profitable.
Z.U. Don't think you're smarter than everyone else. I do not remember how it sounds in Latin, but in translation. It is human nature to make mistakes.
The topic being developed here is whether MM in itself can lead to a profitable system. I contend that it cannot. You go in and say, give me PPP UUUUU (initially profitable) system and I will make it profitable. All I've said is that PPP UUU PPP UUU system cannot be derived from a 50/50 trade probability. Everything else is your speculation. Or do you think that the word "halva" (PPP UPP UPP UPP system) sweetens (simplifies) the process? You don't have a PPP UPP UPP UPP system, just an idea of how to make such a system profitable. Everyone knows how to make it profitable, so what?
Knowing the matheorem frees me from the idea of digging into the results of trades and analyzing them, because no MM will make a PPP UPP UPPP system out of them. It will make sense when the balance in the 50/50 probability of trades is disturbed. You should spend time searching for violation of this balance, and not for the joy of discovery that the PPP UPP UPPP system turns into a profitable one with an easy move. Again I ask, so what? Do you have such a system? Or thoughts on getting a PPP UUU PPP UUU system using MM?
I got into this thread to support the idea that adapting any system with fixed SL/TP is a simple but often wrong approach. OK, you can fix SL, and that's probably correct, but it's harder with TP, especially if the system is heavily trending. I also don't like optimizing by TP. It often turns out to be a fit.- It's always a fit. In this thread I just gave an example of the idea behind such fitting - we don't statanalyse market properties, we invent a law that our answers will coincide with the only specific market realisation - the curve over the period.
On the other hand having a "bad" signal, let's say not the best signal, and working towards maximizing profit and reducing drawdowns (through logic and filters on closures), you can squeeze out many times more than just finding a pattern, and using that pattern in determining TP. - Statanalysis tells me that this is the kind of market property that has a definite stat advantage, from which optimal take and stops are clearly calculated. Me, not statanalysis, but I can see that I can squeeze out many times more and I'm stifled by the sting that here a fixed take cut off profit too early. But my problem is that I don't know how to squeeze more out. Worse, statanalysis has already squeezed out the maximum, and I'm a faithful admirer of it, and I tend to think that as soon as I try to squeeze out more, I'll squeeze out less. What allows you to hope for more? Are you building a formula for peaks, trends, swings or the like?
I say "bad" signal as I know where to improve it, and how, but I don't do it until I'm done with the outputs. The outputs are the main thing, and the hardest part. It is easy enough to get a good signal, but what to do with it is a mystery to many people. Statistics have nothing to do with it. - Amazing! The thing is that a good signal is all for me, this is the answer when and how much money I will receive. My gut feeling is that you cannot explain to me what a good signal is. Or can you?
Simply, having outputs where you need them is easier to understand how to enter, so for me it is secondary. That's it, in a nutshell. - I see, I have a clearer idea of where I am not getting you - good signal, what is it?
.....
My objection was to the terminology used here (it can be misleading). What you just said is correct and I agree with it. But some people confuse the concept of 50/50 deal probability and the number of outcomes, say that 98% positive trades are good.
No MM will fix (not improve) the TS if the probability of profit in a trade is 50/50. In that theoretical TS that I cited, this is broken, there is a probability (my a priori knowledge of the deal is 100%). Although the number of outcomes is 50/50.
My objection was to the terminology used here (it can be misleading). What you just said is correct and I agree with it. But some confuse the concept of 50/50 deal probability and the number of outcomes, saying that 98% of positive trades is good.
No MM will fix (not improve) the TS if the probability of profit in a trade is 50/50. In that theoretical TS that I cited, this is broken, there is a probability (my a priori knowledge of the deal is 100%). Although the number of outcomes is 50/50.
Got it, I agree, we need to be even stricter. The outcomes outside the sense of probability didn't interest me at all, as PPPPPPPPPPPP... can just as easily be made unprofitable as profitable, despite the imbalance in the number of trades. And it is also because 98% of positive trades are almost always followed by 50/50 split, which is the subject of the discussion. Anyway, I wasn't thinking in any way about the quantitative balance of trades. I'm sorry if I misunderstood you.
Vita, the balance is the same, 50/50, in PPP UUUU. It is just that the distribution of the lengths of the series of results of the trades is very different from the Bernoulli scheme, since the trades are clearly dependent.
Obviously, that's what I've been talking about all along. On the contrary, I didn't think about the balance of P and U numbers as being of no interest and not a topic of discussion.
Which indicator, if it is not a secret?