Where is the line between fitting and actual patterns? - page 31
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
TheXpert: Пляя, давайте уже лучше про ООС.
+1 ))
Jingo:
24.01.2011 00:31
1) results at the best optimum ( parameter selection is an individual art of the observer-trader)
2) real (profitability within 5-15% of sample time)
3) inevitable future.
Using OOS in real trading = losing profits.
I want to talk about plot #2 in your picture and the 3H in my chart.
My current situation is as follows:
1) Optimized on a sample of duration, e.g. 5 months.
2) Selected by certain criteria FN (by the way, in my implementation, it can be more than one set))
3) Next, the moment of truth.
and here is this section (3H) I have an extremely tightly fixed, for example, 1 month.
At the end of it and get the result of working out FN.
............................
I understand that rigid fixation is not right. But even with this state of affairs -- the results are not disappointing.
The plan is to refine the functionality.
But how to determine the stopping point (the end of the current PH)?
There are several options. I do not know what is right.
.................................
Who has any experience on this issue?
In short, why did I ask about types? For the first type of TCs it is impossible to optimise and at the same time to avoid fitting, and for the topic of this thread - there is no boundary, only pure fitting.
Only for the second type it is possible to optimize adequately without adjustments. I see no one showed any interest - I'm going to close the topic.
PS And, oh dear! To the first type are probably up to 90% of codobase codes.
In short, why did I ask about types? For the first type of TCs it is impossible to optimise and at the same time to avoid fitting, and for the topic of this thread - there is no boundary, only pure fitting.
Only for the second type it is possible to optimize adequately without adjustments. I see no one showed much interest and the topic is now closed.
PS And, oh horror! The first type probably includes up to 90% of codes from the codebase.
Andrey, your statements are too categorical to leave them without comment.
People are silent because they have to immerse themselves in the subject before saying something... And rightly so.
....................
I have the following attitude to the time of the trade:
- The results show the values of min, max, and average trade length, taking into account weekends and the number of weekend passes, separately for longs and shorts.
And if the average trade is large relative to the optimization period, it will attract attention when parsing the results.
In short, why did I ask about types? For the first type of TCs it is impossible to optimise and at the same time to avoid fitting, and for the topic of this thread - there is no boundary, only pure fitting.
Only for the second type it is possible to optimize adequately without adjustments. I see no one showed any interest, so I'm going to close the topic.
PS And, oh dear! To the first type are probably up to 90% of codobase codes.
I do not agree, neither with such division of TC into classes, nor with properties which you assign them. According to this division, for example all the TC fall to the first type(TC with unknown time of each trade in advance. To this type belong all TSs in which it is not known in advance when there will be the next entry signal into the trade (and whether it will be at all) and in some systems it is not known in advance when there will be an exit signal) and except for adjustment and as a consequence the curvature in the optimizer, they are not good for anything else? But that's not the case, is it, or am I misunderstanding something?
Yes, there are TCs more prone to bad fit, there are less, but the matter there is more in TC sensibility, "robustness" of the idea, informativeness of processed data, ability to process it with necessary degree of approximation, etc. But in any case much will depend on the learning process itself and the interpretation of its results.
And if you are so inclined to characterise the contents of the codebase, could you please give an example of a prominent representative of each class. I'd like to see if I got something wrong after all.
I want modestly to share some information. I should say at once that I am far from the programmer profession, but my curiosity eats me, as well as any of those present here with a selfish purpose. It is called differently, but I will say it is a bar placed by its range (HL) in the previous candle (HL), i.e. fading or correction of the current trend.The question - how often the next candle after the "placed" will be the opposite colour from the first. My answer is about 50/50 for the daley, with a deviation of about 0.3%... I took another time frame and got the same ratio: 45 034 lines (minutes), positive outcomes - 2224 with a possible 4471.Hourly for the year 408 / 812. Daly since 2001 - 164 / 337.
And again a question - what's going on? Is the whole system set up and is it really artificially generated? Or am I completely clueless about excel :o). I can send the "book" to someone, maybe check it out... I am personally shocked... I just wanted to stay out of the market and occupy myself with something...
What is the secret meaning of the information?
What is the secret meaning of information?
I guess so.
lasso: People are silent because you have to immerse yourself in the subject before you can say anything... And rightly so.
More like no than yes. People have simply fallen silent, waiting for the right moment to plunge with renewed vigor and ecstasy into the sin of fludorastia. That's, asked, davvecha, mushroom hunters delete their posts - no, and did not scratch - whether they do not consider themselves mushroom hunters, or consider themselves proud mushroom hunters, one of two. :)
My attitude to time of trade is as follows:
- The results show min, max, and average trade times, taking into account weekends and how many passes through the weekend, separately for longs and shorts.
If the average trade is large relative to the optimization period, this already attracts attention when analyzing the results.
Close, but still cold.
-----------------------------
TCs using NS are also divided into these two types. And it's not about NS. And further, I suggest that the terms learning/optimization/law search should be considered synonymous.
Let's try to grasp it further together, if I get some help and don't get pelted with rotten tomatoes.
Figar0: Yes, there are TCs more prone to bad fit, there are less, but it's more about the meaningfulness of TC, "robustness" of the idea, informativeness of processed data, ability to process it with the right degree of approximation, etc. But in any case, much will depend on the learning process itself and the interpretation of its results.
Does that sound very fuzzy, don't you think? I, at least, try to divide TCs somehow, clearly into two types, so that we can further communicate in a meaningful way. This is not a rebuke, but there are more questions than answers. Which TCs are more prone to "bad fit" and why? And other questions. And there are a lot of questions in general on this forum (the legendary forum of MTS developers, the place where the best minds of the planet gather - judging by the globality of the problems they solve, or at least try to) due to lack of clear and unambiguous interpretation of the terms, although often used, examples of misunderstanding each other are numerous and examples are in this thread.
Figar0 : And if you're so inclined to characterize the contents of codebase, could you give an example of outstanding representatives of each class. I'd like to see if I've got something wrong after all.
Sure.
1) The simplest and most common example of a TS of the first type, or as part of a TS - the intersection of MA.
2) All the TS that have in their logic the rules limiting the time of an individual trade. Example - at the opening of the session on Monday we buy, if <condition>, we close it at the end of the day. And similar ones that have clear rules for entering and exiting, and have a trading time limit, and where it is known in advance when there will be an entry (exit can be during the time limit of a trade).
To be continued. I see there is interest, at least in two of them - yes or no, it doesn't matter, the main thing is interest.