Yoghurt systems and canned systems or The relationship between trading tactics and the reliability of historical test results - page 4
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
Do you agree with my message or not? (This is in reference to the need for your comment)
typed long - apparently it didn't go through.
For me the priority is testing on a long period of history and I don't accept short over-optimisations, because the market is extremely unstable now and one can get a 50-100 pp drawdown "out of the blue", and no indicator will be able to predict such situations.
I think the minimum run time for an EA should be 3 years. In 3 years a large number of all possible situations are simulated, which can prevent drawdowns in the future - but not for sure.
I typed long - apparently it didn't go through.
For me the priority is testing over a long period of history and I don't accept short overoptimizations. The market is very unstable now and you can catch a drawdown of 50-100 points "out of the blue" and there is no indicator that can predict such situations.
I believe that the minimum run time for an EA should be 3 years. The minimal run period is 3 years and a lot of different situations are simulated that may prevent from the drawdown in the future but it is not a fact.
I, for example, am not intimidated by drawdowns. It's a matter of MM. You don't just have to open on half a depo or the whole depo.
And about the variables - the fewer there are, the more you understand - what you are going to do, rather than just adjusting to the story.
>>I absolutely agree!
It's not about the signal. It's about how to understand whether this TS will work in the future or not. But either I don't understand your method or you explained it wrong somehow(((.
I'll try to explain in simpler way: we are looking for a certain pattern in the near future history (well, almost the present). We optimize it and get some parameters that work best now (i.e. in this immediate past). Then we "window" these parameters further on in the history and fix the results. Thus, we very easily determine the period of time that this pattern (with our now good parameters) exists. As a rule, the regularity emerges gently from the past and will gradually worsen in the future through the present. If you check it on history (the starting point, that is, the optimization shift into history, so that the future is relative to it), you can clearly see how the pattern appears and then disappears. It is possible to make money on disappearance. And the more stable the pattern was in the past, the longer and with fewer deviations it will last in the future.
Well the market is far from perfect, so describing an "ideal" pattern makes no sense.....
It's just for colours :)
I, for example, am not intimidated by drawdowns. It's a matter of MM. You don't just open on half or all of the depo.
To some extent I agree - but if the EA is optimized for a trend, and it will get into a sideways movement, then the stops will eat most of the deposit - i.e. the losses will be tens or hundreds of points. And vice versa.
I've been typing for a long time - apparently it didn't go through.
For me, the priority is testing over a long period of history and I don't tolerate short overoptimizations, because the market is very unstable now and you can get drawdowns of 50-100 points "out of the blue", and not even one indicator can predict such situations.
I believe that the minimum run time for an EA should be 3 years. The minimal run period is 3 years and a lot of different situations are simulated that may prevent from the drawdown in the future but it is not a fact.
Often deep optimisations result in very heterogeneous optimisations. See the balance graph. It is likely to grow well somewhere and dip or "flatten" somewhere :) The depth of optimization also needs to be "optimized".
These "reoptimizations" are necessary to detect the patterns. Further work can be done however you like, but how else can you find the patterns?
I'll try to explain in simpler terms: we are looking for a certain pattern in the immediate past (well, almost the present). We optimise it and get some parameters that work best now (i.e. in this immediate past). Then we "window" these parameters further on in the history and fix the results. Thus, we very easily determine the period of time that the given pattern (with our now good parameters) exists. As a rule, the regularity emerges gently from the past and will gradually decrease in the future through the present. If you check it on history (the starting point, that is, the optimization will shift into history, so that the future is relative to it), you can clearly see how the pattern appears and then disappears. It is possible to make money on disappearance. And the more stable the pattern was in the past, the longer and with fewer deviations it will last in the future.
I agree with you somewhere, of course, but it's still not certain that it will be like this.....(((
Often deep optimisations result in very heterogeneous optimisations. See the balance graph. It is likely to grow well somewhere and dip or "flatten" somewhere :) The depth of optimization also needs to be "optimized".
These "overoptimizations" are needed to identify patterns. Further work can be done in any way, but how else can you find regularities?
I also don't like "going deep" into history as there are many different patterns that may interfere with each other and it will be more difficult to find the ones we need..... And we need to work "here and now" and not in the year 2000, for example))))
Let me try to explain in simpler terms: we are looking for some kind of pattern in the immediate past (well, almost the present). We optimise it and get some parameters that work best now (i.e. in this immediate past). Then we "window" these parameters further on in the history and fix the results. Thus, we very easily determine the period of time that the given pattern (with our now good parameters) exists. As a rule, the regularity emerges gently from the past and will gradually decrease in the future through the present. If you check it on history (the starting point, that is, the optimization will shift into history, so that the future is relative to it), you can clearly see how the pattern appears and then disappears. It is possible to make money on disappearance. And the more stable the pattern was in the past, the longer and with fewer deviations it will last in the future.
It's just for colour :)
Well how can I tell you - a trend can essentially be taken as a pattern, and so can pullbacks on the terrnd - but how can you determine the duration of the trend, and whether the pullback is the start of a new opposite trend?