Please tell us your opinion. - page 7
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
And the statistics is not all good, in my opinion. A lot of trades in 4 months are partial. The loss is more than half of the profit. A profitable trade is almost equal to a losing one. Expectation is small.....
Well, this is just a test of how predictable (repeatable) the market is, it's not a ready-made TS. We take the last few bars (we can call it a primitive price pattern), and on the history we gather statistics of the next bar up/down, then this statistics is stupidly "translated" into probability, and then we just open/close. That's the whole expert, that's why it is so frequent. But I'm more surprised why this simple primitive approach works so "well" in early 2008 and so badly in early 2007. I think in the context of the topic question there is a lot to think about. That's it, there's no talk of TC yet.
There were no optimisations at all, what I have posted is the first pass of the EA. I have now started optimization of three available parameters, it works slowly (every bar goes through historical data again, I should have saved them in a file, but I did it urgently). Therefore I will show you the results later, but again there are some surprises...
Well, this is just a test of how predictable (repeatable) the market is, it's not a ready-made TS. We take the last few bars (we can call it a primitive price pattern), and on the history we gather statistics of the next bar up/down, then this statistics is stupidly "converted" to probability, and then we just open/close. That's the whole expert, that's why it is so frequent. But I'm more surprised why this simple primitive approach works so "well" in early 2008 and so badly in early 2007. I think in the context of the topic question there is a lot to think about. That's it, there's no talk of TC yet.
There were no optimisations at all, what I have posted is the first pass of the EA. I have now started optimization of three available parameters, it works slowly (every bar goes through historical data again, I should have saved them in a file, but I did it urgently). So I'll post the results later, but again there are some surprises...
Interesting, yes.
But my Expert Advisor, I can say it works well in early 2007. But my personal opinion is that it doesn't have to work at the beginning of 2007. The market does change over time. As they say, thank goodness it works now.)))))))))))))))))))
Let's consider the optimisation finished, at least all the variants that were considered have been passed, there will only be repetitions from now on. The optimization report is attached. The results are quite amusing - there are simply no cases where the probabilistic Expert Advisor fails ! With any combination of input parameters there is a profit, no matter how small.
This is your test period for EURUSD Strategy Tester from the stats on the first page. I hope my little research will help you make conclusions about the robustness of your Expert Advisor in the future.
On what period was the optimisation carried out?
And is it possible to do a forward test (OOS) for the sake of experimentation?
And on the same period of 2007 from January till May the same optimization does not find any profitable variants, or rather it finds 2, but very ridiculous, pennies at all) Probably now is the time for TS and using probabilities of all colours, but will it last long? I wish I knew... Maybe that is the answer to your original question.
And the forward one, the first test was the forward one. And in general, the expert is probably still poking around for research purposes.
Well I don't seem to have it that way. Here are the same TCs, not even overtrained, just for the period from January to May 2007. Worse, in my opinion, but still good. And if I train them earlier (the end of 2006), I think it will be Ok.
I've been watching people wondering why the system works for a while, then doesn't work and then works again. I've tried to solve the problem not globally, of course, but locally. Within a week or two. In this case we have a system that works for the first week but doesn't work for the second one. We have another system that works for the first week but is not working for the second one. Then we make a choice. The point is that one TS has one inputs and the other one has another one. Then we calculate the correlation of their inputs to the required claw and use the TS with the higher correlation. Ie these inputs act on the pair at the moment. Maybe poorly explained, but I think the essence is understood... I hope so anyway
(You're measuring?)
Optimised, I think, for April this year. Maybe March, it was a long time ago, can't remember. The lot seems to be constant from some point. And do you know the conclusion? The real state of affairs is only shown by the real account.
Measuring up?)
Optimised, I think, for April this year. Maybe March, it was a long time ago, can't remember. The lot seems to have been constant since some time ago. And do you know what the conclusion is? Only the actual account shows the real state of affairs.
Measuring up?)
Optimised, I think, for April this year. Maybe March, it was a long time ago, can't remember. The lot seems to have been constant since some time ago. And do you know what the conclusion is? Only the actual account shows the real state of affairs.
checkpoint test.... - no longer a grail.