divergence trading is a different ball of wax. there are systems out there that focus on that sort of thing although that style is not as easy. IMHO
in short (he he ) you would look for pairs that have a negative correlation. that would not include eur/usd with gbp/usd
ok, thanks
daveM,
Iāve used tradestation in the past (pre-brokerage) for its system writing capability but outgrew its level of analytics and programming, so I designed my own testing environment and donāt use TradeStation for Forex. The link posted just happened to be at the top of the search results and seemed to relate to medisoftās question regarding beta.
Youāre welcome!
Hello,
This is my first time actually posting to babypips. I have been so enthralled by this conversation that I had to way in. I have been using this strategy on a demo account with mixed success. My main problem is the same as many others. Due to the positioning of the charts, it is difficult to see whether one currency pair has actually deviated from the other and if so how much. What at one moment may look like a thirty pip deviation may later turn out to be only three, if not completely opposite from the actual current deviation period. I understand the gentleman from Mexico is using ATR to position his charts. I was curious how many daily periods you use in determining the ATR you use.
Also, I know that indicators, at least most, are inherently lagging. I actually did find an indicator that seems to have all of the requirements to trade this strategy. It is, however, 400 dollars. I am in no way promoting this indicator but I wanted to post a link so that better minds than mine may be able to take something useful from it. fx anglo trader .com Because this was my first post I was unable to post the actual link. Go to the products stat arb and look at the 8 min video about the indicator. It will be labeled dynamic and fixed trigger levels.
Timehopper
If I remember correctly I opened the trade on the 1 minute chart with a difference of more than 40 pips, and then I later checked the 5 minute chart and there was only a difference of over 35 pips.
So you say no longer than over night, so maybe 7 or 8 hours is the average time a 1min or 5min trade should take?
Right now Iām working on Beta neutral position. I think it is more accurate that ATR method that only measures volatility. Iām also working on a method to rebalance the trade to maintain the beta neutral position, and only have the spread to deal with.
If Iām doing the things right, the position should be beta neutral vs USDX (or JPYX, or CHFX or whatever benchmark the trade is using), so the only thing that can move the price is the variation of GBP vs EUR because the USD part will be neutralized all the time.
Found this at other fx site a while back and forgot all about it. It measures the stoch difference between 2 pairs. I set it to all 1ās just to observe how quickly they diverge.
okā¦ first the time frame that you use to open the trades you MUST stick to that frame. changing the time frame changes the math thus throwing off the trade. If you do it on a one min donāt even bother to look at anything else. check the 1 min charts from the time you opened the trades until right now, or the time you closed the trade. I think you may have missed your opportunity to exit. also this last weekend was not really the best time to trade. (worldwide holiday on different days). Also try using a higher time frame. like the hourly. I did this morning and started the day off with a profit. Anyone else???
You also need to make sure both currency pairs are show at the same scale. MT4 by default re-scales everything. Check my previous post in this tread how to set the scale.
Hi All,
So, Iām in deepest France on my hols, doing a lot of reading and thinking about this thread and this strategy. Hold on, just being passed another glass of sancerreā¦ 
Apologiesā¦ but i found an internet connection so wanted to post about some conclusions I have come to whilst Iāve had all this thinking timeā¦
Beta
First off, my thoughts on Beta is that it is not relevant for stat arb with currencies. That tradestation article that was posted is a really good article, it explains it quite well I think. All the papers and information on the internet relating to beta and trading the correlation generally relate to equities or commodities. So, if you were trading ibm and microsoft, the nasdaq 100 would be used to generate the beta for both.
However I couldnāt find anything specifically relating to beta for currencies. So, applying some logic, we have to remember that currencies are unique in that each instrument (say EURUSD) is in fact a pair that you are already buying and selling. So if you buy EU and sell GU, you are buying/selling three currencies and not just two. So, for currencies, you canāt use the same principle to apply beta. Iāll explainā¦ If you have the EU and GU, you might think that you need to generate their beta from some other related benchmark, like the dollar index. However the DXY is weighted massively to the EUR, so, your calculated beta would be biased in favour of the EU.
Instead, we need to think what beta is trying to do. It is measuring the magnitude of one instrument against another (a base). For the currency pairs example, we want the magnitude of the EUR and the GBP against the USD. Well, the beta for each currency pair, IS the value of that currency pair.
For example, the beta of EU against the USD is approximately 1.3 and the beta of GU is approximately 1.6. That is, the EUR has moved 1.3 against the USD and the GBP has moved 1.6 against the USD. Those ARE the beta valuesā¦therefore, we donāt really need to calculate a value for beta as it is known alreadyā¦
That tradestation article talks about whether it is necessary to be beta netural. My conclusion is that because we are already USD neutral, and our beta is calculated against the USD, that means we are already beta neutralā¦
That was my thoughts about beta. The other thing I wanted to mention was how to actually trade this strategy. There appear to be lots of problems with people saying their charts move off, so when something is aligned, later it is not and it makes it difficult to trade. Also there is debate about positions sizes and also about why sometimes it works and sometime it doesnāt.
So Iāve been reading a lot about this, material that is generally about stat arb. One of the links I found is this one which I think summarises quite well what we are actually trying to tradeā¦
The Secret To Finding Profit In Pairs Trading
Iām going to refer to the chart about halfway down that talks about Ford and GM, and is a chart of neither symbol, but the ratio between them. You need to have a read that link before reading the rest of this postā¦
If we now apply the same principle to currencies we want a chart that displays the ratio of the currency pairs.
Here is a chart showing a number of things:
(see http://i.imgur.com/7MIlW.png for the full size png as babypips will rescale it and it might be hard to see)
The main bars are the EU, the lighter bars are the GU and the blue line is the EG. Ignore the red line, it is the 85EMA and not relevant here.
Underneath the main chart is my ratio indicator as explained in the article above. It shows the ratio between the EU and the GU in black. It also shows the mean (average) in blue and then 1 and 2 standard deviations from the mean (in orange). Ignore the circled area, Iāll come onto that later in the post.
Now the investopedia article talks about using the mean over 1 year. I think that is too long a period and I reduced the mean to 20 days (recalculated for the timeframe). That might not be correct, and a longer timeframe might be better, but that is what I have on my chart.
So, for everything except currencies, this ratio chart is what is needed. So, we can pull up highly correlated instruments such as pepsi and coca cola, or brent and WTI charts, and see how the ratio indicates how they diverge from each other. We need to look at the history and see how many standard deviations the ratio moves, before reverting to the mean. We then buy and sell the WTI and brent as appropriate when the ratio between them has moved the appropriate amount.
This screen shot however is not brent and wti but is EU/GU. From this screen shot, what you should see is that the ratio indicator (in black) is absolutely identical to the EG line on the main chart (blue). Of course it is, it has to be, because the ratio between two currencies (EU/GU) IS the third currency (EG). We know this relationship exists between the 3 currencies and it is a unique relationship for foreign exhange instruments.
The ratio indicator on my chart is showing bands that are 1 and 2 standard deviations away from the mean. Lets say my historical analysis tells me that this ratiofor these currencies typically moves 2 std devās and then reverts to the mean. Occasionally, it does move to 3 std devs, but mostly 2 and then back to the mean. So, when the ratio touches the upper 2nd std deviation, we sell and when it touches the lower 2nd dev, we buy. Simpleā¦
ok, here is the next leap. The ratio we know (and can see) is identical to the EG, but the bands are just showing standard deviations from the mean, so they are effectively just bollinger bands.
So for highly correlated currencies this situation is unique. All we need to do is pull up the cross currency chart (in this case the EG chart) and apply bollinger bands to the chart. We set the deviations in the BB parameters to whatever our historical analysis tells us, in this case, 2 std devās. Then we sell when the price hits the upper band and buy when the price hit the lower band.
Here is my EG chart with bollinger bands for the same period. It looks virtually identical to my correlation ratio indicator, as it shouldā¦
(again the direct link to the png is http://i.imgur.com/psZCD.png)
So we can see clearly where the price in the circled area hits the lower b band, and then reverts to the mean. It looks identical to the circled area in the ratio indicator on the previous chart. If we had traded it, it would have been about 40 pips.
ok, I know there are bollinger band trading methods and I am not just saying we should throw on a bollinger band onto every chart. What I am saying is, I am drawing the conclusion that for highly correlated currencies, statistically, this should be a high probability trade, and certainly an easier way to view and decide when to enter and exit a trade based upon this strategy.
Now the same chart shows where the trade didnāt work, price hits the bollinger band, but then keeps going. However looking at the ratio on the previous chart, the same thing is happening. This is just one of those situations where the ratio between the two currencies is widening past 2 standard deviations, which we know does happenā¦ This is also my I want a long period on the bollinger band, but not a year. If we have a longterm divergence, I want my bollinger bands to take it into consideration and move, so that more trading opportunities arise.
So, thatās my conclusions from thinking and reading over the last few days. Not sure when Iāll next get an internet connection, but I wanted to post that while it was fresh in my mindā¦What do you think?
Right, Iām off for another glass of wineā¦
PLEASE NO ONE TAKE OFFENSEā¦ I see that as soon as myself and a few others starting talking about the math behind the system this kinda took on a whole new direction; thus leaving some people behind. I hate to say this but there is a lot of over thinking going on. This system works, and works well. Like I have posted before, it takes the need of being handy with math out of the Equation. ( Pun intended ) we can talk about all the indicators and extra little things to get this system to be perfect , but then you will never get a trade in as all the extras will almost never line up in a perfect situation to get a entry. This is the main reason why systems that people buy almost never work for themā¦ NO ONE FOLLOW THE RULES of that system; they always try to improve it. I can assure you almost every trading system out there works if you adhere to the rules and never change them. Thatās how it became a system in the first place . ( think the Turtle traders) I do the math by hand before every trade. I plot the ATR the variation the standard deviation, the correlation the whole works on a blank piece of paper. ( takes me 20 or so mins.) and almost every time just for kicks I compare it with the Kelton overlay. [ copyright pending : ) ] Oddly enough the overlay method is almost always produces positive results. No indicator no support / resistance no ATR etc. if you add that then it is no longer the system that was keltons. whom I notice is missing from the conversation.
Hi,
Pairs trading is a trading strategy mainly applied to stocks (from the same sector actually). And it has nothing to do with risk free profit. Pairs trading cannot be applied in FX simply because there is a third pair involved in the equation. e.g in eur/usd & gbp/usd the 3rd pair is eur/gbp.
Try this in excel: Find the daily, lets say, percentage changes (%Ī) on EUR/USD & GBP/USD (close prices) for a period of some months and then create a time serie by deducting (%Ī ā¬/$) - (%Ī GBP/$).
Create another time serie for the daily %Ī of EUR/GBP.
(Pay attention in order to compare same dayās results from the 2 series.)
Finally, find the correlation between those 2 time series The result will be 1 or perfect correlation. This means that there is no arbitrage opportunity.
This is happened because there is the so called ātriangle arbitrageā.
Even in stocks statistical arbitrage is a complicate process which requires the implementation of advance maths, like cointegration.
P.S. Even if you like this way of trading you have to try to distinguish your analysis between testing and trading periods in order to use updated data.
Just saying
I hate to say this but I think you are slightly misinformed. Statistical arb can be and is used in fx trading, as it has been done for years. (also my account would disagree with you) to be technical you do create a synthetic position (eu / gu). Perhaps you are getting tied up in the term āpairsā if you view an position say e/u as one instrument then advanced statistics can and is applied. So Iām going to say this and just going to have to take the backlash that come with itā¦ here goesā¦ If people stop thinking about the synthetic (e/g) that is created they would have an easier time with this. THE POINT IS TO PROVIDE A MEASURE OF SAFETY AGAINST WIDE MARKET MOVES IN THE CURRENCY WHICH SHOWS UP TWICE. ie the USD if you use eur/usd and the gbp/usd ā¦that is why you can trade this with a completely naked chart. It wonāt matter if the dollar goes up or down you will make money. IT IS LIKE BEING BOTH LONG AND SHORT THE EUR/USD MINUS ABOUT 20 PERCENT IF THEY ARE +.80 CORRELATED. again i will state for the uninformed it works like thisā¦
take two currencies that have a high correlation against a third.
WAIT until they break correlation on whatever time frame you use. DONāT change time frames. EVER
SHORT the currency that is outperforming
LONG the underperforming currency
WAIT until they correlate again
EXIT and take profit.
BEWARE one side will be a losing trade the other will be in profit. but that is the point
PRO-TIPS :
keep you price per pip SMALL
if the pairs continue to diverge add more to the position if you want / can
WAIT. If the pairs have a history of high correlation they will correlate again.
CONS: small profit range which can be estimated before entry
may take a long time.
if you miss when they correlate again and fail to exit then you are in trouble. there is a way to get out but iām not telling. 
if your position is too large and the pairs continue to diverge it could drain your account. donāt bet big.
dont use stops. Goes against everything you have been taught. I know.
I can make a list of the only ways this system can fail if you want.
MATH IS FUN
cheers, and happy tradingā¦ now let the back lash begin in 3ā¦2ā¦1ā¦
I started trading this system this week. Has been profitable everytime I did it as stated by Kelton. I found out what happened to one of my trade sets that did not work out. I simply mixed up the pairs and bought the pair I should have sold and sold the pair I should have bought. Every other time it worked. However, something strange occurred on the trade I did this morning. I did it exactly as i should but when the lines came together and I closed the positions for a profit, the profit was way smaller than what it should have been. I used .05 lots on each position and had made around $7 before, this time $1 and some changeā¦???
CONGRATS ON YOUR TRADESā¦ keep it up. as far as the trade this morning if you can post more details maybe we can figure it out together. but my first thought is that the divergence may not have been as big as your other trades. but it was still a profit, so please donāt try to change the system. now Wash Rinse and Repeat
cheers
No backlash from me you are correct on all points,
in order to avoid the missing the crossover which happened to me a couple of times
I now trade the EG and put a limit on the tradeā¦no stops just limits if it continues to diverge I add to the position
and move my limit accordingly, I am basically scalping (taking small profits right away before they separate again)
I use the FXCM Trading station for the web which allows you to overlay pairs and then use the EG chart below only for ease of placing the trade, so basically using the GU/EU as real time live indicators, no lagging.
I use the 1 min chart zoomed out to 300 periods and just draw a 20 pip rectangular box on the screen, anytime the GU/EU diverge 20 pips (if GU is on top of EU) then you go long EG, you have to make sure you know which one is on top if you mess that up you will go the wrong direction, anyways the best way to check this out is to put all three on the screen at the same time. I have never lost doing it this way and a few of my limits hit in the middle of the night no way to protect yourself trading pairs against that happening. Math is fun but not needed in this form of trading. not much thinking at all once you have your charts set up.
Thanks for the response timehopper. I like this system so far. Trade smaller and make more is the way to go. I use 2 demo accounts from NordFX but trade on Hotforex. Maybe sometimes the correlation is different at times between each broker?
Atlee, great stuff with your results so far! How many trades have you made and over what time?
15 trades so far this week, some i get in with 10 pips divergence and then buy in more if they move against me, I am using a micro account buying one lot at a time I am up 7.34 so far this week. like i said this is not gonna get you a lot of pips at once and you have to be glued to your screen to get in, but after you are in just set your limit, for instance if it is 20 pips divergence I would set my limit at 5 pips from my buy in. That way they do not have to totally cross to take profit and then if they move in opposite direction you can get in again instead of riding it up and down.