Ok, so I tested version 28 on 30M charts (to get results faster). The S/L problem appeared to be solved, but another problem came up. Namely, all the trades were not closing at a S/L as they were supposed to, but were closing themselves after a small loss or profit. I’m currently running a new test with FX junction attached this time. I will come back later and post a link to the account once I have some results.
The only thing i can think of is the current bug we known of in the EA is causing the wildly different results.
Perhaps and i am taking an educated stab here as we know the bug is not setting the SL correctly, perhaps a random SL algorithm is being used by the EA that we are yet to determine (which i think James will know better than me and probably knows if from initial results v28 has fixed), this would explain the wildly different results and why you are having a profitable EA and why i have a losing EA.
Unfortunately we can only reliably test the different brokers once we are certain the bugs are fixed, so until we are 110% certain the bugs are gone i think we should hold off testing.
Once James fixes the EA if we get the same results as above then we might want to see which brokers produce different results and we can look into that further, as a future thing.
Interesting thanks for that Greensky, also can you try it on the daily charts, i know this is slower testing but perhaps that new bug is just due to the fib lines being so close on the 30min chart ?
When you compared v27b to v28test how much better was v28test, i just have not had the time to upgrade yet 
My job for this evening.
In addition for those not aware v27b is an improvement over v27a, it is more accurate and for me at least the EA has turned from a loss to being profitable instantly.
I would suggest if possible if those testing upgrade to James v28test just so we can see if the existing bug was fixed and how bad this new bug is.
Hi sandy - hmmmm . . . . I have trouble seeing it as an EA problem because these are all originating from the same EA on the same machine. To me, that says the input to the EA surely has to be different - so it’s broker variance ??
Good point, i just wanted to rule out the EA being the problem before we went onto the broker side. Are all of your brokers giving different results or are some giving similar results ?
Its just if we can find some that the EA performs on in a similar manor maybe we can find out what it is on the broker side that is causing this?
I don’t know if you remember a while back there was a strange bug due to i think hotforex being inaccurate and James made the EA do the calculations for data manually rather than relying on the broker, perhaps it is something like that again ?
(sorry if im a bit vague, its late)
Unfortunately I am not sufficiently up on coding or understanding MT4 peculiarities but, I frequently see comments in the Hopwood forums, where there are many top coders, about brokers giving different results due to their own customizing and MT4 is by default refered to as CrapT4 and brokers as criminals. So I am highly suspicious of broker funnies. I personally think we have the SL problem in the EA but the variance in results is surely due to broker feeds.
The three (mainly) Australian brokers are giving better results. FXOpen as I recall is a New Zealand broker but likely white label for someone else, unknown to me. I don’t know the answers but I do know brokers pull some real funny stuff and MT4 seems to be the biggest offender for bad platform code.
What about if we are having both issues:
- SL issue in the EA
- Broker issues
also don’t forget the incorrect entries issue James found annd fixed in v27b, perhaps a combination of all three is confusing the hell out of us all, even when we get the SL issue fixed you could be right, some brokers could be better than others for this EA as for the reasons you stated earlier.
Ok, here’s my FXJunction demo:
The first few trades were on 30M with V28 test EA. Everything has since been changed to 24H charts. Like I said, I did not find any weird S/L levels, but the trades were not closing properly.
Greensky, in what way were they not closing properly?
The change(s) in v28xx is to recreate missing fibs from already open charts. on my chart, it would properly replace a deleted fib to match the existing open and stoploss of the open order. I figured that’s possible what was causing incorrect stoplosses, if the fib went missing.
The actual stoploss is a “hard stoploss”, but the progressive “fib” stops depend on the price passing the actual fib line, then the hard stop is moved. No fib, might mess up the progressive stops.
Not closing properly meant that the trades were closing with small losses, or very small gains - ie. .05-.3% losses when the risk was set to 2%. These would not correspond to any levels at which the trade should be closed according to original rules (ie. go to hard S/L, or put new stops once trade hits fib levels in your directions). There should be full S/L, or partial wins of varying sizes, not small, partial losses. You can see on the small number of trades on the FXJunction account - they all had small losses when the risk was set to 2%.
Ok, I’ll review the stoploss lot size calc
are you leaving the lot size at zero
if you have anything but zero in the lotsize, it will override the risk setting.
You have to put lotsize = 0, for risk calculation to apply.
What about if you don’t use risk do you get these problems what i would say is try risk setting using these two options:
Risk as "0"
Risk as “0.01”
Im curious to see if we can replicate it, going to have a play around later about this
I now understand what you mean.
Ok shown below the grey box is v27a ( i closed all the trades at 0.16 to start 27b.
The orange box is v27b
The red box is v28test, its odd so many trades closed with a small, i had set the breakeven pips to being 5 pips. What is interesting is the SL is not set plus the trades open and close within 1 min of them opening, so something not quite right here.
Greensky with v28test was this what you were getting? Closed trades without a set SL and closing at a small loss? Also anyone else getting this with v28test ?
Damn, that isn’t good. Based on this, I removed the v28_test_a from download. Will revert to v27b and look again for stoploss issues.
Usually, the problem is obvious, but not so much now.
Ok everyone some good news James found that he didn’t filter the trade by magicnumber and symbol on the trailing fib order mod function, so he is certain this should do the trick and fix it.
I did a quick test and it looks good, but of course if you guys can let James know that would be grand !
The new version is v27c, so just use this version from now on. As far as i am aware this should fix the two bugs that i as aware of. v27b fixed one and this should fix the main SL bug we saw.
As usual please report anything strange !
Thanks James for the hard work you have on this EA, everyone give him a hand :35:
Wicked_Wicks_ea_v27c.zip (5.84 KB)
HA, so it WAS a simple fix! I just came home from work - had a very crappy day - and this news literally made me laugh out loud. Excellent job jbentz
Good stuff James. As the Aussies say “Good on ya mate”
SandyBeach, thanks for posting the ea. Just a small clarification: I said it definitely could cause the problem. I didn’t say it definitely will FIX the problem. But I hope so. Language, always subject to interpretation.
What is important now is we know there was a problem and were getting there in fixing it, at least now were stumbling in the dark, just a matter of waiting for trades to see how it goes.
By the way for those interested i put v27c on my H4 and Daily chart for this EA if you want to compare.
did I understand correctly… “stumbling in the dark” is a good thing? Guess it’s better than not trying at all 
Ah, the English language. 
OK James, I have installed 27 c across my six platforms. Each has several */jpy pairs all previously with wrong SL figures. I was frantically hoping 27c would correct the open SL figure, but no. It remains at 1.577 on them all. It seems to be only the */jpy pairs, currently, that have faulty SL figures although I did note some others previously. I have a pending GJ showing correct SL.