Please tell us your opinion. - page 3
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
No, it's a simple combination of two trend indicators with a constant refill on the trend. i'm just a follower of this strategy
OK, how do you define overtraining?
No, it's a simple combination of two trend indicators with a constant refill on the trend. i'm just a follower of this strategy
Ok, so how is overtraining defined?
Usually from the selected trend change rates that affect the SL/TP levels. As long as I play around, I can't find any universal solutions.
Ulterior, even 0.1 lot is too aggressive MM in this case. Recommended MM: fixed 0.01 lot with initial $10K deposit (then drawdowns will not be so horrible). Or minimize the number of simultaneously opened trades (series lengths indicate that trades are opened in bunches and closed in the same way).
Management of number of orders and quality of entry will be the second step, as long as there are too many logical pits (no shorts)
Well, I'll try to give you a brief answer. Adaptive means that it retrains (changes) on every new bar. The dimensionality of the input vector means what? I did not write the network, but one good programmer did it for me, so I do not know the nuances of the network. It is clear from the avatar that it is rewritten in MQL. There is only one way out. Tests differ in different network parameters and different indices that are on the input and output. The decision thresholds are selected by the genoptimizer. I tried it on 15 min, not bad either. I cannot use minutes, like the bettor does. I cannot understand why. But on the other hand, do I really need those minutes?
I have only one question to solve - how to assess the functionality of the TS in the future?
Thanks for the answer. Too bad you are not aware of the network details. Good luck!
Thank you for your reply. I wish you were more aware of the details of the network. Good luck!
It's not about the details of the network, in my opinion. It's much more important and global than that. I will now try to explain.
Here we have made a TS. It does not matter what kind - with a neural network or conventional indicators or something else. By the way, when showing the statistics of two TS I hadn't mentioned what it is based on. It was only later, when I was asked, that I said it was an adaptive probabilistic network. So, we have good or not so good equity on training, we have good equity on OOS. Does this give a guarantee of the TS working in the future? Or what parameters should we use in the report to evaluate the TS functionality in the future? By the number of trades? By the profit? By equity fluidity on the training or OOS? By the maximum drawdown? By the ratio of profit and loss? Or other criteria? What do you think?
From my extremely modest experience, I will put profitability (profit factor) in the first place.
Then I look at the relative drawdown. And then the net profit.
And the number of trades, I think, should be at least 120/150 !
Otherwise the whole undertaking loses its meaning.
I often see post-optimization tests on the forum with 30 or even fewer deals. "Comrades do not understand ..." that this is a typical fitting! During optimization the tester often goes through trillions of variants, and of course among these trillions of variants there are hundreds, in which even obviously losing Eckpert with 3-4 dozens of deals will be the "Holy Grail"!
But when optimizing on several dozens of deals we have some (even a small) guarantee that at least a dozen of next deals, outside the optimization period, will be profitable in total...
From my extremely modest experience I will put the profitability parameter (profit factor) in the first place.
Then I look at the relative drawdown. The number of trades, I believe, should be no less than 120/150 !
Otherwise the whole undertaking loses its meaning.
I often see post-optimization tests on the Forum with 30 or even fewer deals. "What the comrades don't understand is that it is a typical fitting! During optimization the tester often goes through trillions of variants, and of course among these trillions of variants there are hundreds, in which even clearly losing echpert at 3-4 dozens of deals will be the "Holy Grail"!
Is this all on the OOS? Or on the training ground?
This is in the training area, i.e. the area where optimisation is going on...
I didn't quite express myself correctly above - "...I see tests on the forum after optimisation with 30 or even less trades...".
Here we are talking specifically about the optimization period.
This is in the training section, i.e. in the optimisation section...
Well, in my experience, the bigger the profit in the optimisation zone, the higher the probability of adjustment. Although I have not analysed the profit with the number of trades at the same time....