What makes an unsteady graph unsteady or why oil is oil? - page 3
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
There are methods that are superior, mathematically superior, to Fourier-only predictions.
What are those methods?
Честно говоря никак. Я пока даже не думал над этим. Но, придётся подумать, спасибо за совет.
То, что Фурье не работает совсем - это не так. Работает, но по моей системе получаются большие просадки (это раз) и через 5-10 профитных сделок появляются 1-2 сильно убыточные (это два), изменение соотношений SL\ТР тут не помогают.
ST/TP and signal-to-noise ratio are two different things. As long as we are talking about the prediction. How we use it is a matter of tenths. The main thing is to be sure that you do it correctly.... mathematically correctly...and you can't do it better, that's the main thing.... and then say it doesn't work."Doesn't work" of course depends on the person. For me, it doesn't work well. For others it works well. I will look into the problem.
What can you advise about the signal/noise? Something needs to be added to the Fourier, but I can't figure out what. A filter, but what kind?
Можно я вам задам тоже 1 вопрос. что бы понять на сколько глубоко Вы исследовали Фурье ?Для прогноза Вы выбирали не все составляющие спектра а некоторые (выбирать все бессмыслено), если так то отимальной процедурой является байс, но на практике он не применим чаще всего из-за больших требований к априрорным данным. чаше всего доходят до отношения правдоподобия. И там есть только 1 неизвестный параметр. отношение сигнал/шум. В связи с этим мой вопрос какой величины у Вас отношение сигнал/шум и как вы его нашли ?
If you select the whole spectrum, just take the appropriate value of the function - it is periodic. There is no need to predict anything. And you don't need to decompose anything, unless the corresponding value is missing.
ZS This is rhetorical ;).If you do not take the whole spectrum, you are working with method error. Sometimes you need to discard some harmonics, sometimes others, but it's all a fitting, no better than a tester fitting, although it looks more scientific. Then there are reports of non-stationarity of the spectrum ;)....
Good luck.
Don't play the nastradamus game. You can only know exactly where price will go if you are, for example, an MM or something close to it. And then there is the option of clairvoyance, but that's another story.
А что за методы?
there are many methods
Look for example Box J., Jenkins G. - Time Series Analysis, Forecasting and Management. vol.1 (4 Mb)(djvu).djvu I wanted to attach the file but it is too big to fit
Don't play the nastradamus game. You can only know exactly where price will go if you are, for example, an MM or something close to it. Well, there is also the variant of clairvoyance, but that's another story.
>> what exactly are you interested in?Если Вы выбираете весь спектр, то просто возьмите соответствующее значение функции - она периодическая. Ничего прогнозировать не нужно. И раскладывать ничего не нужно, разве, что соответствующее значение пропущено.
ЗЫ Это риторитчески ;).Если Вы берете не весь спектр, то работаете с погрешностью метода. Иногда нужно выбросить одни гармоники, иногда другие, но это все подгонка, ничем не лучше подгонки на тестере, хотя выглядит более наукообразно. Потом появляются сообщения о нестационарности спектра ;).... зато есть чем заняться...
Удачи.
I agree with almost everything. If we accept the hypothesis that there is no noise, then I agree with absolutely everything. But if we assume that there is noise in quotes Big or small is another question, but there is. then the filtration procedure (removing the noise components of the speck) is reasonable. The main thing in theory is to determine the signal to noise ratio. Knowing it you can set the threshold for further variations, threshold by ideal observer or maximum likelihood...
Z.I. I'm just often killed by statements this .... insert any word - it does not work. You ask why. You ask why. The answer is I have lots of lots and no profit ((( and the person does not understand that such a statement is nonsense. If Fourier did not work, then sorry, you would have no copiers now. no televisions, no cell phones.... look around you. These items are there ? so it's Fourier working, only you're applying it wrong...
there are many methods
Look for example Box J., Jenkins G. - Time Series Analysis, Forecasting and Control. vol.1 (4 Mb)(djvu).djvu I wanted to attach a file but it is too big to fit
Thanks, I'll read it, here are the links, it seems to be it, only sometimes there are glitches:
http://eqworld.ipmnet.ru/ru/library/books/BoksDzhenkins_vyp1_1974ru.djvu
>> http://eqworld.ipmnet.ru/ru/library/books/BoksDzhenkins_vyp2_1974ru.djvu