Going Home - a multi-pair auto-trader for Dreamliner's strategy, by Steve Hopwood | Page 6 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

building multi-pair multi-timeframe scanner 48 replies

multi-pair percentage scale charts 28 replies

Price info indicator needed - one pair's price x another pair 5 replies

Multi-pair trading or Few-Pair Trading. [poll] 15 replies

Need help modifying Multi Pair Stochastic indicator 0 replies

Attachments: Going Home - a multi-pair auto-trader for Dreamliner's strategy, by Steve Hopwood
Exit Attachments

Going Home - a multi-pair auto-trader for Dreamliner's strategy, by Steve Hopwood

  • #101
  • Feb 1, 2011 7:16pm Feb 1, 2011 7:16pm
Dig it !


Quoting the-game
I think We should prevent going against the trend.

So for example let's say EURUSD is overbought and GBPUSD is oversold and the Stoch Diff is now above 80.

Cool! So we go short EURUSD and long GBPUSD right?

No! We test the trend of both and it seems both pair's trend is UP.
(Which is no surprise as they are positively correlated!)

So what do we do? We just go long with GBPUSD until the Stoch Diff close level is met or until we are in profit.

What do you guys think?
  • #102
  • Edited 1:37am Feb 2, 2011 1:15am | Edited 1:37am
Quoting SteveHopwood
That is fantastic news.

I have noticed this: so far (early days yet) only the sell trade in the pair has closed at a profit. It could be that either: 1) we should only be taking the sell trade, or; 2) this is pure coincidence because the bot has been trading for such a short time or: 3) we should lower the stoch diff at which GV closes the trades.

I am going to try CloseTradeStoch = 40 to see if this helps. Would someone else care to try CloseTradeStoch = 30 for a while?

Regarding your analysis about the sell pair closing in profit , could it be anything to do with the trend of the moment as mentioned here by the game

Quoting the-game
I think We should prevent going against the trend.

So for example let's say EURUSD is overbought and GBPUSD is oversold and the Stoch Diff is now above 80.

Cool! So we go short EURUSD and long GBPUSD right?

No! We test the trend of both and it seems both pair's trend is UP.
(Which is no surprise as they are positively correlated!)

So what do we do? We just go long with GBPUSD until the Stoch Diff close level is met or until we are in profit.

What do you guys think?
Sounds like we should think about this
I am just about to change to trigger 90 for you and set a link to my system rather than a tedious pdf download
We should start a test centre up for some of these EA ,s , have a load of fx book links to a row of systems with the set file detailed in the general tab .People offer to test systems and they must adhere to the setup standard. That way we have people testing different s permutations rather than all the same

http://www.myfxbook.com/widget?id=81...=1&color=green

et voila , now set to trigger 90, close is still at 30 as requested
Ta for now
  • #103
  • Feb 2, 2011 1:26am Feb 2, 2011 1:26am
Quoting peterporter
Regarding your analysis about the sell pair closing in profit , could it be anything to do with the trend of the moment as mentioned here by the game

Sounds like we should think about this
I am coding this filter right now.

  • #104
  • Feb 2, 2011 1:59am Feb 2, 2011 1:59am
I have coded the trend filter - details in the Inputs section of post 1.

I also fixed a bug in the trade-closure code, so I recommend this re-download even if you do not want the trend thingy.

  • #105
  • Feb 2, 2011 2:14am Feb 2, 2011 2:14am
.
Attached Image (click to enlarge)
Click to Enlarge Name: feb02.jpg Size: 75 KB
  • #106
  • Feb 2, 2011 3:59am Feb 2, 2011 3:59am
I don't wish to detract people from their system here but what may help is if you look at the correlation for the timeframe you are trading and not always the daily which I think is the default. It doesnt make much sense to trade the M30 using Stoch but are looking at the daily correlation.

Anyway, give it a try and see if it makes some difference, which I think it will.

Cheers

Scoobs.
  • #107
  • Feb 2, 2011 4:09am Feb 2, 2011 4:09am
Quoting scooby-doo
I don't wish to detract people from their system here but what may help is if you look at the correlation for the timeframe you are trading and not always the daily which I think is the default. It doesnt make much sense to trade the M30 using Stoch but are looking at the daily correlation.

Anyway, give it a try and see if it makes some difference, which I think it will.

Cheers

Scoobs.
Users can do this by changing the StochTimeFrame input. You can enter the D1 time frame as 1440 (minutes), 240 for H4 etc.

  • #108
  • Feb 2, 2011 4:29am Feb 2, 2011 4:29am
Quoting smjones
I have written some indicators for this system, but I have not necessarily endorsed it, but I am a bit confused when it is said that it will not work, simply because if it did, it would have been arbitraged away and you saying that "you suspect" it won't work. I am very interesting in hearing or seeing hard data that shows it will not work. Otherwise there is nothing left to do, but to test it. If you...
The concept of testing is an interesting one. Who can tell me how long a test should last ? Its probably not a matter of length of time but for your system to see as many different market conditions as possible. However, there is usually a black swan waiting in the wings somewhere and at the end of the day, the viability of any system is how it deals with those situations imo. Other than news trading, I'm rapidly coming to the conclusion, that the viaibility of any system is based more around its money management skills than choosing its in points.

But you're right, I don't have any hard and fast data as such, just my experience that there are times when the correlation of two pairs re-aligns and those times, unless you have good MM in place, will tend to blow your profits.

Don't get me wrong, I like some of the logic behind the concept but as scoob has said, if it was easy, institutions would be in there before you and combined with the fact that correlations can re-align at unpredictable times, I do have some doubts unless good money management can be built into the algorithm.
He was looking for the card so high and wild hed never need to deal another
  • #109
  • Feb 2, 2011 4:41am Feb 2, 2011 4:41am
Quoting dr_who
The concept of testing is an interesting one. Who can tell me how long a test should last ? Its probably not a matter of length of time but for your system to see as many different market conditions as possible. However, there is usually a black swan waiting in the wings somewhere and at the end of the day, the viability of any system is how it deals with those situations imo. Other than news trading, I'm rapidly coming to the conclusion, that the viaibility of any system is based more around its money management skills than choosing its in points.

But...
Institutions do trade correlations but in a different way to what is being proposed here. example would be where there buy/sell USD against the dow or buy/sell GBP against the ftse. However, not only are they experienced but they also have access to allot more hedging instruments if things go wrong, like offseting with bonds etc.

But hey, keeping going with the testing. You will find that some currencies groups work better than others during different market times and this is something that you will need to become experienced about.

I think it has already been mentioned but be very careful trading EURUSD/USDCHF using this method. Yes, you can make some quick profits but when the USD sell off occurs (your be the last to know) then you will more than likely empty your account or get margin called if tight stops are not in place on your trades.

Scoobs.
  • #110
  • Edited 6:15am Feb 2, 2011 6:03am | Edited 6:15am
Here is 2 different orders. First one (EURCHF GBPAUD) has no trend. Second one (EURUSD GBPUSD) has both up trend. The last one had 15 Euro plus on the buy order but ended up with minus.
(The 7 big candles are one week. One candle/day. The small candles are H4 - M1)
Attached Image(s) (click to enlarge)
Click to Enlarge Name: EURCHF.jpg Size: 70 KB
Click to Enlarge Name: GBPAUD.jpg Size: 72 KB
Click to Enlarge Name: EURUSD.jpg Size: 41 KB
Click to Enlarge Name: GBPUSD.jpg Size: 62 KB
  • #111
  • Edited 7:45am Feb 2, 2011 6:22am | Edited 7:45am
Quoting scooby-doo
Institutions do trade correlations but in a different way to what is being proposed here. example would be where there buy/sell USD against the dow or buy/sell GBP against the ftse. However, not only are they experienced but they also have access to allot more hedging instruments if things go wrong, like offseting with bonds etc.

But hey, keeping going with the testing. You will find that some currencies groups work better than others during different market times and this is something that you will need to become experienced about.

I think it...
Scoobs, I have been thinking about this post since you put it up and also the one where you said that we ought to be checking the correlation on the same time frame as the stochastic.

Here is something that just occurred to me, and my question to you. Do you think that the shorter time frame correlations eventually snap back to the longer term correlations? The reason I ask, is if we were to have a 80% daily correlation on pair1 and pair2 and then we noticed we had a -80% correlation on 15 minute pair1 and pair2. My question, is the 15 minute likely, eventually to revert closer to the long term correlation?

If this is the case most of the time, then it seems to me there is an edge in approach using these different time frames of correlation..

Regards, Scott



Dr_Who,
Thank you for your detailed post in answer to my questions. I value your answer and will consider it.
EDIT:
After consideration, I have an interesting idea that you have caused me to think of. I will post it in a few minutes, when I gather my thoughts logically.
  • #112
  • Feb 2, 2011 7:12am Feb 2, 2011 7:12am
Quoting smjones
Scoobs, I have been thinking about this post since you put it up and also the one where you said that we ought to be checking the correlation on the same time frame as the stochastic.

Here is something that just occurred to me, and my question to you. Do you think that the shorter time frame correlations eventually snap back to the longer term correlations? The reason I ask, is if we were to have a 80% daily correlation on pair1 and pair2 and then we noticed we had a -80% correlation on 15 minute pair1 and pair2. My question, is the 15 minute...
Checking correlations at Mataf shows that corrs. between the same pairs vary considerably from one TF to another. This may argue for trading based on one TF only.
Vincit qui se vincit.
  • #113
  • Edited 7:42am Feb 2, 2011 7:17am | Edited 7:42am
Quoting Pippopotamus
Checking correlations at Mataf shows that corrs. between the same pairs vary considerably from one TF to another. This may argue for trading based on one TF only.
perhaps, however, the problem with Metcaf is that you cannot see the change over time of the correlations. So I am thinking the question I had is still valid. You only see a snap shot at metcaf at that moment in time.

What if the shorter time frame correlation is temporarily out of phase with the long term correlation. Does the short term correlation snap back to the long term? or is there no connection on correlations of different time frames?

I think I will write a correlations index indicator that tracks correlation over time and over multiple time frames and see if the idea has legs. And then create a feedback loop to run the two time frame index through a correlation calculation on each other. That should answer my question definitively.

uh oh, getting dangerously close to Quant stuff. Oh well, it could not be helped.
  • #114
  • Feb 2, 2011 7:50am Feb 2, 2011 7:50am
Quoting smjones
I think I will write a correlations index indicator that tracks correlation over time and over multiple time frames and see if the idea has legs. And then create a feedback loop to run the two time frame index through a correlation calculation on each other. That should answer my question definitively.
crawling back into my box .... Why do I suddenly feel so pathetically humble and unworthy, its all too much ....
  • #115
  • Feb 2, 2011 7:58am Feb 2, 2011 7:58am
M5-D1
Attached Image
  • #116
  • Feb 2, 2011 8:00am Feb 2, 2011 8:00am
Quoting smjones
perhaps, however, the problem with Metcaf is that you cannot see the change over time of the correlations. So I am thinking the question I had is still valid. You only see a snap shot at metcaf at that moment in time.

What if the shorter time frame correlation is temporarily out of phase with the long term correlation. Does the short term correlation snap back to the long term? or is there no connection on correlations of different time frames?

I think I will write a correlations index indicator that tracks correlation over time and over multiple...
with respect to outcome. This is correlating the correlations! Smj, your work never ceases to amaze me.
Vincit qui se vincit.
  • #117
  • Feb 2, 2011 8:01am Feb 2, 2011 8:01am
Quoting smjones

I think I will write a correlations index indicator that tracks correlation over time and over multiple time frames and see if the idea has legs.
Like one of TRO's multi-meters. Nice idea.....
He was looking for the card so high and wild hed never need to deal another
  • #118
  • Feb 2, 2011 8:05am Feb 2, 2011 8:05am
Quoting Pippopotamus
with respect to outcome. This is correlating the correlations! Smj, your work never ceases to amaze me.
I am trying not to hav e an expectation of the outcome so as not to taint the result. LOL
  • #119
  • Feb 2, 2011 8:11am Feb 2, 2011 8:11am
Quoting dr_who
Like one of TRO's multi-meters. Nice idea.....
Actually I think I will set it up linearly so that I can display the correlations over time and for any given time in the sequence. This ought to piss off my CPU. See you all later.
  • #120
  • Edited 10:23am Feb 2, 2011 8:41am | Edited 10:23am
Quoting E2u
M5-D1



Luckily we have Steves EA to capture the triggers

On a more serious note
Steve , when you are doing an enhancement , filter or whatever next bug fix would it be possible to also put in an option to close the corr pair when in profit as an option and whatever value in dosh the user would like
It annoys me to watch the yellow line for an hour or so and only to close at something silly like -0.71.
Only when you have time yeh , its not a showstopper
0 traders viewing now