my bad, not seen previous post, sorry
scooby-doo BOJ JPY pairs trading robot
- #3,662
- Nov 30, 2009 2:23pm Nov 30, 2009 2:23pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
Where's my mate Steve or did you lot p!ss him off lastnight ?
Sometimes the day job gets in the way of programming.
when i compile the 24H 2.8 or the last Original 1.2, an alert come up:
Function "CalculateBasketTp" is not referenced and will be removed from exp-file
From a few posts back ... not very hard to find ...
http://www.forexfactory.com/showpost...postcount=3652
- #3,664
- Nov 30, 2009 2:51pm Nov 30, 2009 2:51pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
Looks like he was a bit tired, spending time here and teaching piano...man needs rest..looking at the way he is going he can compete with Carnot engine!
I am not tonight, either, but will try to keep the nastier side of my nature in check.
Having said that, for those too dim to read back a few posts the latest Origianal is in post 1. There is nothing new in it, but the CalculateTp function is removed. That is the function that a few are too lazy to read about in the last, um, 10 posts or so.
To those of you not too dim to read the last, um, 10 posts or so, you don't need the update.
I have made an alteration to the 24H version. As originally coded, the Correlation thingy would never have allowed trading. Scoobs, would you confirm or contradict the change I made, please?
Let's hope BOJ gets out of bed tonight/this morning/middle of the afternoon - whenever it is supposed to happen in whatever part of the world you happen to occupy. Wherever you live, I hope the sun is shining.
Unless you live in a part of the world where it should be dark, in which case sunshine would be confusing.
- #3,668
- Nov 30, 2009 3:08pm Nov 30, 2009 3:08pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
Origianal, LOL, I get it origi "anal". Good joke Steve
- #3,669
- Nov 30, 2009 3:16pm Nov 30, 2009 3:16pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
Whats wrong with the correlation thingy ?
What I do know is this:
- the basket only consists of 2 pairs.
- one of those is GJ
- as written, the correlation computation ignores gj
- so it only computes the ej correlation
- then divides the result by 2
- which gives an untradeable reading every time I look at it - i.e. <0.7
- when there were more than 2 pairs in the equation, the reading was often over the 0.7 threshold. As coded now, it never is, so I have commented out the if(TradePair[i] != Symbol()) conditional
- #3,670
- Nov 30, 2009 3:17pm Nov 30, 2009 3:17pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
- #3,672
- Nov 30, 2009 3:28pm Nov 30, 2009 3:28pm
- Joined Jul 2009 | Status: Trader | 2,158 Posts
Oy. Where did you get that photo of my trading patterns?
- #3,673
- Nov 30, 2009 3:34pm Nov 30, 2009 3:34pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
Yeah, my mistake, what I was trying to avoid was cor("GBPJPY","GBPJPY")
I think it should be this:
Group.Cor = 0;
for(int j = 0; j < ArraySize(Pairs); j++)
{
if(Pairs[j] != Symbol())
{
Group.Cor += Cor(Symbol(),Pairs[j]);
}
}
Group.Cor = Group.Cor / ArraySize(Pairs);
- #3,674
- Nov 30, 2009 3:36pm Nov 30, 2009 3:36pm
- Joined Jul 2009 | Status: Trader | 2,158 Posts
Isn't that the same code that lead to merely reading the ej correlation and dividing the result by 2, thus giving a reading that is always about 0.4xxxx?
Group.Cor = 0;
for(int j = 0; j < ArraySize(Pairs); j++)
{
if(Pairs[j] != Symbol())
{
Group.Cor += Cor(Symbol(),Pairs[j]);
}
}
Group.Cor = Group.Cor / (ArraySize(Pairs)-1);
- #3,675
- Nov 30, 2009 3:36pm Nov 30, 2009 3:36pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
It was like mine today, I lost GBP 87K on USDJPY.
- #3,676
- Nov 30, 2009 3:40pm Nov 30, 2009 3:40pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
But it should work ok.
If there are 5 pairs in the basket, then dividing the result of the looped call by 6 can still give a meaningful result. Not when one of the only two pairs is not included in the calculation, surely? That only leaves ej. What is ej correlating with?
- #3,677
- Nov 30, 2009 3:43pm Nov 30, 2009 3:43pm
- Joined Jul 2009 | Status: Trader | 2,158 Posts
Yes, but that is the same as cutting out the loop thingy and simply sending a corr(ej) call. Dividing the result by 2 will lead to a result too small to allow trading.
If there are 5 pairs in the basket, then dividing the result of the looped call can still give a meaningful result. Not when one of the only two pairs is not included in the calculation, surely? That only leaves ej. What is ej correlating with?
Group.Cor = 0;
for(int j = 0; j < ArraySize(Pairs); j++)
{
if(Pairs[j] != Symbol())
{
Group.Cor += Cor(Symbol(),Pairs[j]);
}
}
Group.Cor = Group.Cor / (ArraySize(Pairs)-1);
- #3,678
- Nov 30, 2009 3:55pm Nov 30, 2009 3:55pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
Here you go:
Group.Cor = 0;
for(int j = 0; j < ArraySize(Pairs); j++)
{
if(Pairs[j] != Symbol())
{
Group.Cor += Cor(Symbol(),Pairs[j]);
}
}
Group.Cor = Group.Cor / (ArraySize(Pairs)-1);
- #3,679
- Nov 30, 2009 3:58pm Nov 30, 2009 3:58pm
- Joined Jul 2009 | Status: Trader | 2,158 Posts
Cheers. I figured it should be that all along, but didn't have the nous to suggest it. Ah well, can't always be at my brightest and best. Hey-ho.
- #3,680
- Nov 30, 2009 3:59pm Nov 30, 2009 3:59pm
- | Commercial User | Joined Apr 2007 | 8,331 Posts
Cheer up Steve, there is money to be lost.