DislikedUse a 1 and not a 0 then.
You are still going to have the problem of the changing history in the off chart pairs unless you lock them down. This is the fault of MT4. You have to lock down the history for the off chart pairs.Ignored
Hedge and Correlation Strategy
- #2,542
- Feb 10, 2011 5:05am Feb 10, 2011 5:05am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
- #2,545
- Feb 10, 2011 5:21am Feb 10, 2011 5:21am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
- #2,547
- Feb 10, 2011 5:22am Feb 10, 2011 5:22am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
- #2,548
- Feb 10, 2011 5:24am Feb 10, 2011 5:24am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
- #2,550
- Feb 10, 2011 5:47am Feb 10, 2011 5:47am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
- #2,552
- Edited 6:08am Feb 10, 2011 5:55am | Edited 6:08am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
This was also my initial thought.
I wanted to have a confirmation of this.
I knew the "accident" had to do with the Lot sizes and their signs, and therefore the direction of the simulated trades taken against the Stoch Diff and the way the strategy uses the stoch diff.
However this is still a very interesting path if the iStochastic issue can be nailed down reliably.
How do you "lock down" the pairs history ?
Then do you force MT4 to use this "locked down" history data ?
Sq
BTW here is the latest chart. the thing bounces all over the place, unfortunately.
Also note, that as far as I can tell the current bars is correct. This would mean that EAs running multi-currency strategies are probably accurate in real time (forward testing) I wonder if this is why MT4 did not allow multi-currency back testing in it's strategy tester.
and another that snapped back.
That's enough to prove my point I think.
This confirms what i was saying about the stochastic values that sometimes seemed to be taken from the next bar or even 2 bars in the future... !
EDIT:
i just read your answer above, and looked at your code;
the "run_once" solution is similar to what i wrote here below i guess.
We then just need to compute our own stochastic on the buffer.
I'll try that.
/EDIT
What solution can we find here ?
Record our own "close price history" into an extra buffer in the indicator,
initialize it in init() to get a "snapshop" of the close prices at the time we start the indicator,
then, at new bar open only, update only the last bar ([1]) of it with the last MT4 bar close price (iClose(pair,0,1) :but can we even trust that one ???) , and use that "reliable-close-buffer" to compute all trade functions like iStochastic(), iMA(), etc on our own ?...
What a pain...
Hopefully MT5 will have these bugs fixed...
However, that can still be coded in the equity backtester indicator because we are only using iStochastic() which is simple to recode on a buffer rather than on history data.
(iMA() has a iMAonArray() version so that one at least we don't need to code ourself...but we are not using it here anyway).
Sq
- #2,554
- Feb 10, 2011 6:15am Feb 10, 2011 6:15am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
- #2,556
- Feb 10, 2011 6:27am Feb 10, 2011 6:27am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
So now our only solution is to use a tick data source and rebuild our own history and use only that one...
Do you know if iClose(Pair,0,0) possibly give different results from MarketInfo(Pair,MODE_BID) ?
MarketInfo(Pair,MODE_BID) is probably the most reliable way.
Well, i can always code a quick indicator to Alert when it is not the case...
Sq
Now to fall a little deeper down the rabbit hole; the price we see on the chart as the current actual tradeable price, isn't. It is the past by microseconds, to as much several seconds, but never real time; another anomaly of MT4. For most retail traders, it does not much matter, but I can predict the future price, because MT4 shows it in the past.
- #2,558
- Feb 10, 2011 6:33am Feb 10, 2011 6:33am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
- #2,560
- Feb 10, 2011 6:52am Feb 10, 2011 6:52am
- Joined Mar 2006 | Status: THANK YOU MERLIN,TWEE and FF Team | 4,603 Posts
ok, thanks.
may i ask why i see two lines in the subwindow at your picture? a red and a blue, but only one indicator attached?
i tried to replicate this, but did not get it.
The reason there is only one title, is because both indicators have the same current data in the IndicatorName()
This is where that indicator is now out of phase. The reason that it appears to be correct before the yellow line is because they were both initialized at that point. So, which do you trust? I do not know. I think the locked one is at least easier to rely on. But, is it accurate?
