Yoghurt systems and canned systems or The relationship between trading tactics and the reliability of historical test results - page 8
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
I would like to propose a discussion of the relationship between trading tactics and the reliability of the results of historical testing.
There is definitely a dependence between trading tactics and the results of historical testing. But it is not mutually reversible. In mathematics we often face situations when validity of the statement "If A, then B" does not necessarily imply validity of the reverse statement "If B, then A". It's the same story here. A good trading strategy (tactics rather refer to MM while trading is a strategy) will surely give good test results. But good test results do not at all guarantee good quality of the strategy. Therefore, the answer to the question about reliability of testing on history is rather negative. Although it is possible, of course, to extract some information about TS from test results. But, imho, it will not be of crucial importance.
When creating a trading system, we rely on some regularities depending on the input variables.
When creating TS, everyone relies on whatever comes into his/her mind. That is why the TS construction has such disappointing results. And everyone puts into the word "regularities" the meaning that his or her consciousness, education, notions and many other personal details allow. All this happens at the initial stage of initiation of a person into Forex and the search of his or her own "grail". As a result, these "regularities" to a large extent determine the task formulation and, therefore, lack of adequacy in ideas about sought regularities guarantees a complete failure (although even full adequacy does not guarantee success).
Here, for example, is a characteristic view of patterns:
I think the longevity of a system depends on the patterns on which it is built. The market is volatile and the patterns that worked before may not work now. There are all sorts of regularities in the market. Some last a long time, others do not reveal themselves much, some are not regularities at all, but wishful thinking. Therefore, to create a reliable system, you have to immediately eliminate signal systems based on short patterns. You have to look for long patterns and catch signals from them.
From my point of view, patterns are neither short nor long. They are simply not patterns. At best they can be called fluctuations. And the author of the above quote is going to build his TS on fluctuations, while sifting out those of the nach that constitute "short patterns". It is only unclear how to do this. Without knowing anything about where it came from, one can only know that the "pattern" is short at the end of it, i.e. post factum. For TC this is, of course, very valuable.
Imho, a pattern is a natural manifestation of the nature of a phenomenon. So a pattern can only disappear if the nature of the phenomenon has changed. If the nature of the phenomenon remains, but the conditions and circumstances of its existence change, then the pattern will remain, but it will behave differently in the new conditions. For those who know this regularity, this change of form will be quite natural. For everyone else it will be "Ahhhh, the market has changed".
Forex 20 years ago was a market of big players and the processes were developing in the same way. Now anyone with $100 in his pocket can play Forex. Moreover, computers and the Internet have completely changed the technology of the process. As a result, it has significantly changed and accelerated. And what about any known TC survived? None, it seems. This is because all the non-surviving TS were not even close to any knowledge of the patterns of forex. But the nature of forex hasn't changed, and if there was a lucky man who managed to get to the real patterns, his system still works. And I think that's a good enough reason not to make it public.
What do all these abstractions have to do with the question asked. Directly. If we are talking about actual patterns, no one here in this forum knows them. Therefore, the venture mentioned implies a very high risk. The author of the expert venture may be interested in such an investment. The one, who is really found, is more interested in silence and secrecy than in money.
When we create a trading system, we rely on some regularities, depending on the input variables. Often at the testing stage we optimise the system according to these parameters and when we get a good result, we rejoice.
If a pattern is known, it also dictates the "input variables". Such variables do not need to be optimised. Their current values are determined from the real data stream.
If it is not a pattern but only a model, it requires some parameterization like any interpolation. Optimisation on history is both the best and most deceptive way of determining these parameters. "The best" is when the behaviour of the model is similar to that of the market. "The most deceptive" is when the opposite is true. The problem is that even out of sample testing cannot tell how it really is. If even in some sections interpolation approximates the function quite well, it says nothing about other sections, even neighbouring ones. After all, the function itself is unknown to us - and that says it all.
Why are some systems yoghurt, others canned tuna, and is it possible to create a perpetuum mobileis, or at least approximate one?
It is possible. To do so, one has to research forex, discover its nature and find its patterns. If you are interested in it, by spending a lot of intellectual and creative effort, as well as time, you may (if you are lucky) become a creator of a unique TS.
If you are not interested and just want to get it all for your money, you will be disappointed along the way.
...It seemed to me that the question posed by the topicstarter was specifically about how to adequately assess TC's future behaviour...
No, I didn't. :)
It's just that the words "Statistics" and "Percentage" have been mentioned in vain more and more lately. :(
Searching for patterns is a pointless activity if we cannot substantiate with any degree of reliability that the TS built on the found pattern will behave acceptably in the future.
Well, I don't know... The future is not known in principle.
But there will never be a guarantee of the future. A year ago, who could say anything about oil today?
1. Pichimyu? )))
2. What kind of sample would be sufficient for you? In terms of number of trades, in terms of testing time.
1. Mischek answered well. Because the market is influenced by various external factors. They constantly change existing trends and patterns. It is impossible to find one formula of the market (read regularities) by which you may work profitably for 10 years.
2. Sampling is a very complex and not a simple issue. I select it only experimentally. I cannot put together those observations and feelings when I look for the right period of optimization or training. For example in my topic Please tell me your opinion. I raised this question. But never got an answer))). The TS trained for 3 months (01.10.2007-31.12.2007) worked successfully until the end of May. Further it also worked fine, it didn't plummet, but the deals became less, which consequently led to lower profits. Since the end of May the market has strongly changed. And since June we have another TS working, which has already been trained for 4 months, not 3. From my experience the amount of bars for optimization depending on TF is 300-3000. For example 750 bars for daily bars is 3 and a half years and for Hourly it is a little over a month. Do you feel the difference?
"The search for patterns" is an eternal and inexhaustible topic, for which no answer will ever be found (in the perpetuum mobile sense). And testing and evaluation of TC is quite a practical problem that can be solved for a given TC, even if the allegedly found patterns are not logically understood or completely unknown. It seemed to me that the topicstarter's question was asked precisely about how to adequately assess the behaviour of the TS in the future...
The search for patterns is meaningless if we cannot justify with any degree of reliability that the TS built on the found pattern will acceptably behave in the future.
Then I want to ask - is trading more art or mathematics? If it is mathematics then maybe we cannot ground it from the mathematical point of view, but if it is art, then maybe we can "feel" the market and understand what will work in the future by our "third" feeling?
Can you feel the difference?
Not really... But your approach is understandable... >> Thank you for your answer.
Not really...
Well, three years for days is OK, but a month or so for hourly bars is 'not enough'. And this whole 750-bar thing.
Then one has to ask - is trading more of an art or mathematics?
I think that trading is a craft, which can be varied by art. So, there is a certain learning process and also a place for flair.
I think that trading is a craft that can be diversified by art. That is, there is a certain learning curve, and there is also room for flair.
>> I agree.
>> I agree.
I suggest we formalise our reading.)