What makes an unsteady graph unsteady or why oil is oil? - page 23

 
timbo >>:
Да, правильно. С дискретным процессом я конкретно протупил.

It happens.


A common mistake many beginners make is to try to extrapolate from the mathematical models for the original BP rather than the first differences.


1. The BP of any financial instrument is less stationary as compared to differences and therefore it is knowingly unsuitable for extrapolation

2. There is complete invariance between BP and differences, i.e. one can easily obtain differences from BP, and easily reconstruct BP from differences, including extrapolated parts - calculated future.

 
Reshetov >>:

Бывает.

I have blundered on about the "derivative/difference" moment for a discrete series. This does not invalidate my criticism of your "ideas". The stationarity of the first derivative does nothing to help in predicting BP price. A random walk may well have a stationary first difference, but you can't make money on it.

 
Reshetov >>:

1. ВР любого финансового инструмента менее стационарен по сравнению с разностями, а следовательно заведомо непригоден для экстраполяции

Wrong. You're confusing the warm and the soft. For example a process of the following form x(t) = b * t + e(t), where e(t) ~N(0,1) and b is a constant. The first difference of this process is a stationary process. The process itself is non-stationary, but nevertheless very remarkably extrapolable as far into the future as you like.
 
timbo >>:

Я протупил с моментом "производная/разность" для дискретного ряда. Это не отменяет моей критики в отношении твоих "идей". Стационарность первой призводной ничем не поможет в предсказании ВР цены. У случайного блуждания может быть вполне стационарная первая разность, но заработать на этом не получится.

I'm not going to enter into a debate and argue the obvious, because I don't care whether others make money from it or are desperately opposed to it. Because I'm perfectly happy making money from it myself.

timbo >>:
Wrong. You are confusing the warm and the soft. For example a process of the following form x(t) = b * t + e(t), where e(t) ~N(0,1) and b is a constant. The first difference of this process is a stationary process. The process itself is non-stationary, but nevertheless very remarkably extrapolable as far into the future as you like.

I am not confused about anything. If the first differences are stationary, then the initial BP is predictable. This is quite obvious. If you hold a different opinion, then try to prove otherwise. And we will admire your efforts.

Another thing is that the presence of variance is a probabilistic risk. I.e. in this case Markowitz, Sharpe and the like are right. At least as pioneers who opened their eyes to an obvious fact. Even more right is whoever derived the Sortino coefficient. Why? Explanation. The variance itself is not going anywhere. So to fight it is at the very least a waste of time, especially in non-stationary processes. Why fight it when there is a simpler way, which is prescribed in TRIZ textbooks? Reduce the variance in the loss-making region by increasing it in the profitable region. One unsteady process is transferred to another equally unsteady process, and the variance remains, but it is already working for us, not against us.

 
timbo >>:
Не верно. Ты путаешь тёплое с мягким. Например процесс следующего вида x(t) = b * t + e(t), где e(t) ~N(0,1), а b константа. Первая разность этого процесса - стационарный процесс. Сам процесс - нестационарный, но тем не менее очень замечательно экстраполируемый как угодно далеко в будущее.

The formula is inaccurate: x(t) = b * x(t-1) + e(t), where e(t) ~N(0,1) and b is a constant.
 
Reshetov >>:
Я не собираюсь вступать в дискуссии и доказывать с пеной у рта очевидное, потому как мне до лампочки, будут ли на этом зарабатывать другие или же начнут отчаянно возражать. Потому что меня вполне устраивает, что я сам на этом зарабатываю.

Making money on an I(1) financial instrument? I can only wish you good luck.
 
FOXXXi >>:

Зарабатываешь на I(1) финансового инструмента? Могу пожелать только удачи.
The last thing I am relying on is luck, as that would not be trading, it would be a gambling addiction. Another thing is that I have not yet managed to create a 100% fork, because there are many independent factors. But I managed to stabilize this matter, getting more stable results through portfolio investments, rather than on TA of single instruments.
 
Reshetov >>:
Меньше всего полагаюсь на удачу, т.к. это будет уже не трейдинг, а лудомания. Другое дело, что сваять 100% вилку пока не удалось, т.к. есть множество независимых факторов. Но зато удалось устаканить это дело, получая более стабильные результаты через портфельные инвестиции, нежели на ТА одиночных инструментов.

Do an AR(1) of the resulting process of your portfolio and everything will become clear.
 
FOXXXi >>:

В формуле неточность: x(t) = b * x(t-1) + e(t), где e(t) ~N(0,1), а b константа.
My formula is perfectly accurate. I described exactly the process I wanted to describe - a deterministic trend with some random fluctuations.
 
Reshetov >>:

Я ничего не путаю. Если первые разности стационарны, то исходный ВР прогнозируем. Это вполне очевидно. Если Вы придерживаетесь иного мнения, то попробуйте доказать обратное. А мы полюбуемся на Ваши потуги.

A process of the form x(i) = x(i-1) + e(i), where e(i) ~N(0,sigma^2). The first difference of this process is a stationary process. Now try to tell us how you plan to predict a random walk, "and we'll admire your efforts" (c).
Reason: