Statistics, optimisation and "lucky coin" .... - page 5
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
Yeah. That's right.)
Volodya the spider is Stanley Drakkenmiller
and C-4 is George Soros
but it's a sucrut, don't tell anyone about it
You got it all wrong.
My real name is Sidorov. Just Sidorov.
You ain't got it right.
My real name is Sidorov. Just Sidorov.
Come on, Stanley! Everyone already knows...
What's wrong with Sidorov? Why so much adulation for foreigners?
He wants Soros, Dranenmiller, you know.
The branch question is getting robust systems, or how to distinguish regularity from fitting.
There are two main directions.
1. logical. Any earning system is a frontrunner of mass trades of other market participants. I.e. it is necessary to understand why traders buy or sell at certain points in time and/or at certain prices. As in any zero sum game - you need to understand your opponent's strategy to win.
2. Statistical. The more trades in a test, the more likely the results reflect reality (but the later we start trading it ;)). Of course, the more degrees of freedom there are when optimising, the easier it is to fit the system. But we should not look at individual runs, but at the change of the target when the optimized parameter changes. The form of this dependence shows well whether this parameter is robust or it is curvafit. This means that the system can be broken down into parts and tested separately for robustness. A good system has a minimum of parts and they are all robust))
There are other statistical methods which increase the probability that the estimates on the story were relevant to reality, rather than a fit. Consequently, there is a chance that the pattern will continue, or not disappear immediately, allowing you to make money. What matters here is when to disengage the system from trading. Again this is based on the analysis of the latest history with 1 and 2.
For starters, I don't quite understand the reason for your intense excitement....
From what I wrote in the first post, you didn't understand anything (perhaps I wrote too much, well, sorry...). I'm not going to retell it, but I will explain for you personally, that my idea is that the price (as a certain set of numbers, presented in the form of both actual levels and indicators) used in the optimization process is no better than the same set of numbers, obtained from the LFO. Since any TS built on random matching on history (RNG coin) does not give "confidence" in its performance in the future, the TS built on some detected relation of prices to history also does not give such confidence....
If you are willing/able to identify the demagogy/logical fallacy in the above statement, it would be extremely interesting....
There, again you are making equality between the price and the GCF:
i.e. price, in your opinion, is the CLO, but it is not. No one has yet proved the identity between GCP and price. And to conclude that the two are identical on the basis that they are similar is a logical fallacy.
The branch question is getting robust systems, or how to distinguish regularity from fitting.
There are two main directions.
1. logical. Any earning system is a frontrunner of mass trades of other market participants. I.e. it is necessary to understand why traders buy or sell at certain points in time and/or at certain prices. As in any zero sum game - you need to understand your opponent's strategy to win.
2. Statistical. The more trades in a test, the more likely the results reflect reality (but the later we start trading it ;)). Of course, the more degrees of freedom there are when optimising, the easier it is to fit the system. But we should not look at individual runs, but at the change of the target when the optimized parameter changes. The form of this dependence shows well whether this parameter is robust or it is curvafit. This means that the system can be broken down into parts and tested separately for robustness. A good system has a minimum of parts and they are all robust))
There are other statistical methods which increase the probability that the estimates on the story were relevant to reality, rather than a fit. Consequently, there is a probability that the pattern will continue, or not disappear immediately, allowing you to make money. What matters here is when to disengage the system from trading. Again this is based on the analysis of the latest history with 1 and 2.
A little correction to.... Branch question: the value of statistics and the value of TC optimization in trading. I question the absolutization of obtained statistical data in determining the robustness of TS, because I tried to show that "good" statistics can be obtained in any way for any "Trading Idea" when increasing the parameters to be changed.
As for the two directions: on the first one I agree almost completely (deliberately I don't want to discuss it now, because...
On the second, there are questions:
1). = system can be broken down into parts and tested separately for robustness. A good system has a minimum of parts and they are all robust) = Fitting is able to show good statistical performance both as a whole and on any segment\sections....
2). = There is a possibility that the pattern will continue, or disappear not immediately, which will allow for an earning = The statement has caused bewilderment... In my mind, a pattern cannot start or stop, otherwise it would no longer be a pattern..... Regularity either exists objectively or does not exist at all... Let me give you an analogy: there is a playing cube... On every fourth roll, a "5" falls, and this has been going on for 200 rolls in a row with absolute accuracy... Is it a pattern? How long can one use this detected phenomenon for betting, and when should one stop playing?
I'll give you an analogy: there's a dice...
On every fourth roll, a "5" appears, and this has been going on for 200 consecutive rolls with absolute accuracy... Is it a pattern?
How long you can use this phenomenon for betting, and when you should stop playing?
That analogy alone should be enough to sober the minds of many frequenters of this forum.
But they don't even want to think about it.
There are questions about the second one:
1). = the system can be broken down into parts and tested separately for robustness. A good system has a minimum of parts and all of them are robust) = Fitting is able to show good statistical performance as a whole, as well as on any segment\ of segments....
It is not about dividing up the story and comparing the results, but comparing the results when the parameter being optimised is changed. Read more at
There are questions about the second one:
2). = there is a possibility that the pattern will continue, or will not disappear immediately, allowing you to earn = The statement is perplexing... In my mind, a pattern cannot start or stop, otherwise it would no longer be a pattern..... A regularity either exists objectively or does not exist at all... Let me give you an analogy: there is a playing cube... On every fourth roll, a "5" falls and this has been going on for 200 rolls in a row with absolute precision... Is it a pattern? How long can you use this detected phenomenon for betting, and when should you stop playing?
Regularities here are not like in physics, which are immutable. They are almost all temporary. And the reason is not only that the more a pattern exists, the more people see it and use it (which causes it to disappear), but also that the rules and conditions of the game (market microstructure) periodically change, which is the basis of many "patterns".
As for your example, you can quite accurately calculate the DI for it, which is the probability of one or another outcome depending on the number of throws. The more throws, the narrower the DI for the same probability of hitting it
That's exactly the mistake - it's not logical, rather it's philosophical. Who says that your system will make money? Nobody, there is never 100% certainty, just as there is no certainty that we will even be alive tomorrow. As I recall, the Tibetan Book of the Dead says that there are only two facts we know for sure about death: that it will inevitably happen and that we can never guess in advance when it will happen. Nevertheless, there is no reason to deny an afterlife. It's the same with market patterns: absolutely every one of them will stop working at some point, but that doesn't mean that the coupon clipping opportunities will end there. In short, pray, fast, listen to Radonezh radio, and everything will be chikipuka.
I.e., you've finally come to realise the futility of finding a market pattern.... :)
Without any banter: there is a certain generally accepted idea, which is discussed on various forums, it is talked about at lectures for young traders, books are written, it is applied in practical activities, and everything like that....... I.e., it is believed that every competent trader should use this idea, and the most interesting thing is that this idea is really quite popular... I have tried to build some kind of logical construct that questions this idea. I'm not trying to sell anything, nor am I looking for support for solutions to any problems... :) I am interested in the same logical arguments of the adherents of this idea in its defence. Unfortunately, apart from almost religious indignation, there is very little constructive evidence (and if there is, there are very many logical inconsistencies...). It turns out that the application of an idea is not based on its comprehension, but on its "general acceptance"?
2). there's a dice... On every fourth roll a "5" falls, and this has been happening for 200 consecutive rolls with absolute accuracy... How long you can use such revealed phenomenon for betting and when you should stop playing?
You can calculate by formulating the corresponding problem about the breakdown.
The criterion for stopping is approximately of the following form: -6.204558 * win / n > 4.59512 / n - 4.422849, where n is the number of bets, win is the number of wins (criterion significance 0.01, power 0.99).
And it basically does not give you an opportunity to earn. But you can stop to drain the deposit in time. Surprise! :)