Where is the line between fitting and actual patterns? - page 19
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
Figar0:
... А почему ООS должен быть обязательно после периода обучения??
... Далее, сам период обучения, почему он должен быть непосредственно до текущего момента?
...I, for example, take a "synthetic" training period, made up of pieces of movements that I expect in the future or just pieces covering all kinds of behaviours.
The answer to these two questions is trivial - it is more convenient and customary, there is no other justification. Optimized chunk - moved, the process is logical and formalized.
And your variant requires, first of all, a classification of "all possible variants of behavior" and their weighted choice, and secondly, a convenient repeatable methodology and software for it.
That is I can recognize the constructiveness of such an approach but cannot follow it. And even a small mistake in docking data can lead to the opposite result.
lasso:
Then the question is: What are the fundamental differences between the OB(sample) and your "classic OOS"?
For the particularly gifted: The difference is that on a training sample any fool can adjust the results, while on OOS you can only squeeze a profit out of something worthwhile.
And your variant requires firstly a classification of "all possible variants of behaviour" and their weighted selection, and secondly - a convenient repeatable methodology and software for it.
That is, I can recognize the constructiveness of such an approach, but I cannot follow it. And even a small error in docking data can lead to the opposite result.
Oh hi)
I'm used to not complicating some things. It can all be simplified to the point of impossible.
Look. Suppose we have an EA that trades on watches. We take the last two weekly candlesticks and according to their pattern insert the filter and counter in the EA, and attach it all to the beginning of the start, i.e. allow the EA to trade if we have a "semblance" of the last two weekly candlesticks and the counter is even, and at the end of the week we minimize the trade. If the counter is odd, we trade with similarity of candlesticks. Here it is the expected "variety of movements" and here it is an MS spread over a period of training) All this within the terminal tester and with minimal changes of the EA.
Of course, this is a very simplified example, but anyone with intelligence can follow it. Try something similar at your leisure. It makes sense.
You're thinking right. It's only a little way to our understanding.
Your "classic OOS" is an illusion.
Actually there are only two periods (rays) and a point separating them on the timeline.
- Past and future
- Sample and OOS, OB and TV.
We can virtually move the point (or the "current moment") to the left (the past) as we want. To the right (to the future) we cannot, as there is no data.
If you don't mince words, in general, do you agree?
............................
Then the question is: What are the fundamental differences between OB(sample) and your "classic OOS"?
The fact that your TS didn't see OOS? That answer I know.
................
It's still the same sample plot from which a piece has been "cut off" and called by a big name -- OOS.
What's not to shed light on this answer you know? It's kind of muddy, the more you say the less I understand..... )
No offence, an old joke. Two traders meet each other, let them programmers:
1: - Who is a fool?
2: A fool is someone who says that other people do not understand anything! Got it?
1: - No, I don't get it...
****
That's me too, I don't understand a word you're saying)
No offense, ........
That's me too, I don't understand a word of your reasoning)
I agree.
I've long noticed that I can't get my "conceptual" thoughts across to people. )) Just don't make fun of it.
But there are those who do understand at once. (I know two for sure.)
For the very gifted: The difference is that on a training sample any fool can adjust the results, while on OOS you can only squeeze a profit out of something worthwhile.
A robust TS is supposed to be profitable without optimization
She doesn't owe anyone anything. And the market owes nobody anything. For this reason, ideas about robust grapes which yield profits without any conditions may arise in the morbid imagination of those who are unfamiliar with the volatility of financial instruments.
To put it mildly, "perpetual motion" ideas, please: do not suggest that.
In theory a robust TS should show profit without optimization, optimization should only slightly increase the profitability. If profitability and drawdown strongly depend on the variation of parameters, it will lose at the first "jump".
most likely you just need to know a limited set of robust parameters, but the limits of this multidimensional set will always change with the market - because the market is not a sine wave...
In fact the sets do not have to overlap
For the very gifted: The difference is that on a training sample any fool can adjust the results, while on OOS you can only squeeze a profit out of something worthwhile.
Then the question: What are the fundamental differences between OB(sample) and your "classic OOS"?
The fact that your TS didn't see OOS? That answer I know.
Read the answers carefully and you will increase the number of especially gifted people in our team. ;-))
.................................
Our differences are only in approach, methods.....
I'll try to explain it again later....
I will try to explain it again later....
Rather, you just need to know a limited set of robust parameters, but the scope of this multidimensional set will always change with the market - as the market is not a sine wave of some sort...
In fact, the sets do not have to overlap at all