Correlation between currency pairs

[options-george]

I am looking to compile data to give me correlations between the 10 different currency pairs that i currently trade. I want to do this for risk management reasons.

Ideally I am looking to arrive at data something like this (I have pulled numbers from thin air for this example):
EURUSD-EURGBP correlation 1H (for x periods) 0.8, 4H 0.7, D 0.73
GBPCHF-CHFUSD correlation 1H -0.7, 4H -0.78, D -0.79
etc etc

I may also looking at catching the correlation with the the ES/NQ/TF futures contracts.

My plan was to obtain data to allow me to calculate this as from a designated Day 1, and to then update at the start of each trading session. It will then allow me to calculate how much I can risk per trade if for example I am long GBPCHF, short USDJPY and want to go short CHFJPY given the correlation across the pairs and my set risk parameters for a single trade.

Before embarking on this work, I thought I should ask whether this data is actually available somewhere already (whether free or not) or whether on the forum here has already tried to do something like this. I have done some searching already but have not found anything yet.

Appreciate any comments in advance!

 

[options-george]

just having another Google search I found a super example of what I was meaning to produce - OANDA has put together table like this here

http://fxtrade.oanda.com/analysis/currency-correlation

Although they only do this for a few currency pairs - but this is along the lines of what I am looking for. Has anyone got leads on where such data can be acquired and/or purchased?

 

[options-george]

I can give you the answer for ES/NQ/TF (chuck in YM as well) - very, very, very correlated.

If you've got two positions on in different ones of those at the same time, you've basically just got double exposure.

Little bit less so on very low timeframes, some will be moving faster / further on certain days etc, but basically they move together.

Hi Jose - thanks for your post. Sorry I don't think I was clear enough in my email. I was meaning correlations between various currency pairs and the indices, rather than the correlations between the indices themselves.

 

[scose-no-doubt]

regress the % move per unit of time and check the r value. Obviously you'll get a more accurate picture the larger your sample size.

Don't say I never give you anything

 

[Aspen Trading Group]

I am looking to compile data to give me correlations between the 10 different currency pairs that i currently trade. I want to do this for risk management reasons.

Ideally I am looking to arrive at data something like this (I have pulled numbers from thin air for this example):
EURUSD-EURGBP correlation 1H (for x periods) 0.8, 4H 0.7, D 0.73
GBPCHF-CHFUSD correlation 1H -0.7, 4H -0.78, D -0.79
etc etc

I may also looking at catching the correlation with the the ES/NQ/TF futures contracts.

My plan was to obtain data to allow me to calculate this as from a designated Day 1, and to then update at the start of each trading session. It will then allow me to calculate how much I can risk per trade if for example I am long GBPCHF, short USDJPY and want to go short CHFJPY given the correlation across the pairs and my set risk parameters for a single trade.

Before embarking on this work, I thought I should ask whether this data is actually available somewhere already (whether free or not) or whether on the forum here has already tried to do something like this. I have done some searching already but have not found anything yet.

Appreciate any comments in advance!

I do not think there is any disagreement that various asset classes will impact and drive price action in FX pairs, i.e. S&P's vs EUR/USD - to cite a real basic correlation. Introducing correlations between various FX pairs/crosses though is not so clear cut and I believe at the end of the day, you will find that not much insight can be gained. Some work done by Peter Panholzer at DynexCorp in Geneva (top ranked FX fund) refutes the notion that price action in 1 pair can offer insights into the price behavior of another pair.

We have not done any exhaustive studies on this as has DynexCorp, but we would agree with his conclusion.

 

[scose-no-doubt]

correlation =/= causation

 

[options-george]

I do not think there is any disagreement that various asset classes will impact and drive price action in FX pairs, i.e. S&P's vs EUR/USD - to cite a real basic correlation. Introducing correlations between various FX pairs/crosses though is not so clear cut and I believe at the end of the day, you will find that not much insight can be gained. Some work done by Peter Panholzer at DynexCorp in Geneva (top ranked FX fund) refutes the notion that price action in 1 pair can offer insights into the price behavior of another pair.

We have not done any exhaustive studies on this as has DynexCorp, but we would agree with his conclusion.

Hi Aspen Trading Group - many thanks for your reply - I notice you are a head a trader at a fund? Thanks for taking time to post on this thread - much appreciated.

I guess I am trying to use the rate correlations for risk management purposes rather than as the foundation of a trading strategy. Do you consider that for risk management purposes using historical correlation data is good enough? With my current trading I will sometimes have more than one FX position open at a time - I am a more or less a swing trader when it comes to FX (I don't do that many trades - maybe 20 a month, though I do trade on a full-time basis). I am trying to get an understanding of how to best manage my position sizing for the situation where I have several open positions at a time, and figured that the correlation between the rates may be a good starting point for this.

As far as the correlation between fx rates and equities are concerned, this is of a lower concern for me. My equity trades are all daytrades and will vary from several minutes to at most several hours - and almost always the target and stop distances are far less than the equivalent target and stop distances for any fx trades that might be open at the same time.

 

[martymr]

I do not think there is any disagreement that various asset classes will impact and drive price action in FX pairs, i.e. S&P's vs EUR/USD - to cite a real basic correlation. Introducing correlations between various FX pairs/crosses though is not so clear cut and I believe at the end of the day, you will find that not much insight can be gained. Some work done by Peter Panholzer at DynexCorp in Geneva (top ranked FX fund) refutes the notion that price action in 1 pair can offer insights into the price behavior of another pair.

We have not done any exhaustive studies on this as has DynexCorp, but we would agree with his conclusion.

This is an interesting thread to my personally, since my trading method hovers in the correlation arena. My particular approach, however, is based on the natural cyclic flow of INDIVIDUAL currencies - not pairs. My experience with my own method tells me that correlation between currency pairs themselves is very much a "gray area". They do correlate RANDOMLY at certain moments based on their particular strength/weakness levels at that time, but then break away when certain levels of disparity are introduced.

The disparity itself has a cyclic behavior pattern of its own, which is what typically causes most traditional correlation strategies to 'break'. If you don't know what to look for and when, then your odds of sustaining a profitable strategy based on "commonly viewed" correlation are slim.

I can't divulge into too many details here, otherwise, I'd be jeopardizing the confidentiality of my own hedging methodology. But I can say with full confidence that "correlation trading" as most traders look at it is severely flawed and misunderstood.

Nonetheless, options-George, if you're able to find a way to make it work for you, then that's fantastic. All the more power to you. I'm merely speaking from my own personal experiences and discoveries.

I wish you well with your trading.

All the best,

Marty
-martymr

 

[Aspen Trading Group]

Hi Aspen Trading Group - many thanks for your reply - I notice you are a head a trader at a fund? Thanks for taking time to post on this thread - much appreciated.

I guess I am trying to use the rate correlations for risk management purposes rather than as the foundation of a trading strategy. Do you consider that for risk management purposes using historical correlation data is good enough? With my current trading I will sometimes have more than one FX position open at a time - I am a more or less a swing trader when it comes to FX (I don't do that many trades - maybe 20 a month, though I do trade on a full-time basis). I am trying to get an understanding of how to best manage my position sizing for the situation where I have several open positions at a time, and figured that the correlation between the rates may be a good starting point for this.

As far as the correlation between fx rates and equities are concerned, this is of a lower concern for me. My equity trades are all daytrades and will vary from several minutes to at most several hours - and almost always the target and stop distances are far less than the equivalent target and stop distances for any fx trades that might be open at the same time.

Hey options-George, solid question and my answer may not specifically address it, but let me give it a go.

From a correlation standpoint, we do not rely too heavily on historical correlations, we too are swing traders - 1-3 day duration, so our main focus in terms of correlation is what are the correlations right now - or over the last few days. Once we determine that we can then determine if we buy/sell a certain currency pair/cross do the other asset classes support or refute that idea. Until we have a solid foundation amongst the asset classes, we trend to ignore any trading set-ups that exist only from a technical perspective. This may touch upon your question in terms of risk management I suspect.

As an example, we are looking at a possible reversal in EUR/USD right now via a ending diagonal pattern (possible) but are ignoring the signal for now until we can get the S&P's to exhibit a pattern that suggests they too will go higher. Thus we consider this a 'risk management' tool as we survey the overall landscape first to make sure everything fits together.

If that was not clear or did not address the question - let me know.

 

[NVP]

This is an interesting thread to my personally, since my trading method hovers in the correlation arena. My particular approach, however, is based on the natural cyclic flow of INDIVIDUAL currencies - not pairs. My experience with my own method tells me that correlation between currency pairs themselves is very much a "gray area". They do correlate RANDOMLY at certain moments based on their particular strength/weakness levels at that time, but then break away when certain levels of disparity are introduced.

The disparity itself has a cyclic behavior pattern of its own, which is what typically causes most traditional correlation strategies to 'break'. If you don't know what to look for and when, then your odds of sustaining a profitable strategy based on "commonly viewed" correlation are slim.

I can't divulge into too many details here, otherwise, I'd be jeopardizing the confidentiality of my own hedging methodology. But I can say with full confidence that "correlation trading" as most traders look at it is severely flawed and misunderstood.Nonetheless, options-George, if you're able to find a way to make it work for you, then that's fantastic. All the more power to you. I'm merely speaking from my own personal experiences and discoveries.

I wish you well with your trading.

All the best,

Marty
-martymr

thank god for that ......more for the rest of us

N

 

[fayalac]

You can download the information with almost any broker and you can determine your correlation value with excell.
You can do it often for the 1 hr chart´s to have a dynamic value that can help you to decrease the risk, you could also, with enough input, try to run a regression to foresee the possible change in the correlation values.

 

[Mtweek]

More people lose more because they has little knowledge of trading.Demo account is not realistic.People must learn how to set up account and deposit $10, $20 and learn how to deal with emotion,etc.

 

[NVP]

More people lose more because they has little knowledge of trading.Demo account is not realistic.People must learn how to set up account and deposit $10, $20 and learn how to deal with emotion,etc.

good point...but off topic ?

N

 

[fayalac]

More people lose more because they has little knowledge of trading.Demo account is not realistic.People must learn how to set up account and deposit $10, $20 and learn how to deal with emotion,etc.

Besides off topic, when you have gained certain degree of experince you dan really identify the difference among a demo account and a mini account.
All i will tell you is that any account with a deposit of 10 or 20 is lost. It is useless to open an account with such amount....

 

[QMForex]

Most products such as TradeStation and Metastock have correlation functions built in, however for a specific currency pair each timeframe will have different correlations. Eventually they all revert to the mean however that does not stop large correlation swings in the opposite direction. For example EURGBP has limited correlation to GBPUSD at present, having said that most asset classes are correlating to one in this environment, hedge funds are having a hard time these days.

It is interesting to have this information for diversification and money management, however trading correlation is a very specialist subject.

 

[options-george]

Apologies for my delay in replying here - it was really great to get a number of replies on this thread!

Hey options-George, solid question and my answer may not specifically address it, but let me give it a go.

From a correlation standpoint, we do not rely too heavily on historical correlations, we too are swing traders - 1-3 day duration, so our main focus in terms of correlation is what are the correlations right now - or over the last few days. Once we determine that we can then determine if we buy/sell a certain currency pair/cross do the other asset classes support or refute that idea. Until we have a solid foundation amongst the asset classes, we trend to ignore any trading set-ups that exist only from a technical perspective. This may touch upon your question in terms of risk management I suspect.

As an example, we are looking at a possible reversal in EUR/USD right now via a ending diagonal pattern (possible) but are ignoring the signal for now until we can get the S&P's to exhibit a pattern that suggests they too will go higher. Thus we consider this a 'risk management' tool as we survey the overall landscape first to make sure everything fits together.

If that was not clear or did not address the question - let me know.

Thanks again for this - yes it makes sense to some degree. So if I get you correctly, you wouldn't trade the EURUSD setup unless the S&P supports the intended setup? I guess I judge the setups individually and then alter my position sizing if I consider there to be significant correlation. So slightly different but along the same lines. Thanks again for sharing your thoughts here!

You can download the information with almost any broker and you can determine your correlation value with excell.
You can do it often for the 1 hr chart´s to have a dynamic value that can help you to decrease the risk, you could also, with enough input, try to run a regression to foresee the possible change in the correlation values.

Most products such as TradeStation and Metastock have correlation functions built in, however for a specific currency pair each timeframe will have different correlations. Eventually they all revert to the mean however that does not stop large correlation swings in the opposite direction. For example EURGBP has limited correlation to GBPUSD at present, having said that most asset classes are correlating to one in this environment, hedge funds are having a hard time these days.

It is interesting to have this information for diversification and money management, however trading correlation is a very specialist subject.

Yes I had thought about getting the info and doing it myself. My concern is how much of my time this would take if I was doing this every morning (the correlation of the last few days would impact heavily on my view of the inter-pair correlations). So I was hoping that there might be a subscription-type service out there. I will check back on the other sites to see how often the data is updated.

I agree that it's good to consider correlation for risk mgmt purposes, but I am definitely not planning to design a correlation-based trading strategy :cheesy:

 

[BILLV]

My concern is how much of my time this would take if I was doing this every morning (the correlation of the last few days would impact heavily on my view of the inter-pair correlations). So I was hoping that there might be a subscription-type service out there. I will check back on the other sites to see how often the data is updated

Hi George
I don't understand why you need to look at correlation daily.
Once you've picked your favorite pairs you then put some rules in place to avoid trades which have a high probability of going in the same direction

 

[fayalac]

By the otgher hand George....., adding to what BILLV has just said, markets are very dinamic. To know an updated correlation degree each 15 minutes can be as useless as a total lack of information.
Anyway, i donot remember now, i guess it was fxstreet or dailyfx, one of those sites gives you that information, for some pairs, updated daily - i guess daily but i am not sure -.
Focus on the pairs you choose.....

 

[options-george]

Hi fayalac and BILLV - at the moment I complete an assessment of how each currency is doing at the start of each day (I review the 4H and 1H chart) - I track USD, GBP, EUR, CHF and JPY - giving me a total of 10 pairs. I would want to incorporate the correlation assessment at the same time.

I have noticed that the 1H picture can change quite drastically from one day to the next - and being conservative, I would want to consider the possibility that this could also happen in regard to the correlation - though I'd consider that much more remote. Once I start tracking the correlation I might find that doing it once a week is enough.

I am not trying to put together some rough rules, but instead am looking to make some detailed calculations regarding position sizing of simultaneous trades based on the correlation across the pairs.

 

[Blaiserboy]

Mataf.net has some good info re correlations