seeing has he said he trades on the 5, 15 and 30 min charts I cannot see how that’s the case…
Classic example of what I was noticing though… Thanks for letting me know I am making the right choice.
A few days ago
Demo
Overlapped with the vitrite
Eurusd gbpusd pairs
Scale to fix not checked
The 1 hour chart looked messy
The 30 minute looked messy
The 15 minute overlapped quite nicely
The 5 minute overlapped better, and there was already a decent separation forming so I stuck to the 5 minute
I waited for 30 pip separation made the trades
A couple hours later they touched and even crossed
Exited with 27 pips profit
So can you go through all the timeframes until you see one with a separation forming and then just trade that?
yes… i don’t use it on anything lower than 10. but any time frame works
yes, I traded the wild move yesterday, finished today with 0.18 % profit 
What I finished doing is, after the 200 usd deposit, to rebalance the trade with the move of both sides.
I had 0.01 minilots of each one, for AUDCHF (sell) and NZDCHF (buy)
NC had about 100 pip loss, while AC had 50 pip profit, in the same time. The “beta” was unbalanced in the beginning, because the starting equity was so small to balance the trade. But after the increase in equity, I was able to rebalance the trade, I sold 0.01 mini lots for NZDCHF and bought 0.02 mini lots on AUDCHF. After that, the price of both continued down, but AC profit eventually exceeded the NC loss and closed the trade for a profit. The “beta” value for my trade was more than 1 before I rebalanced the trade. After rebalancing, the resulting beta was about 0.2, pretty near to neutral beta.
I calculate the beta using the ATR for the first entry and after that I use a weighted value of 60 % of the ATR and 40% of the current difference in moves.
If the AC ATR is let say 100 PIPS and the NC ATR is 150 PIPS, and the moves was AC only 50 pips and NC 100 pips I get the beta this way:
BetaAC = (100/(100+150))0.6+(50/(100+50)).4=0.373
BetaNC = (150/(100+150))0.6+(100/(100+50)).4=0.626
And I calculate the lots of 1-Beta, so, if I’m going to assign 1 full lot, I assing 1-0.373 of AC and 1-0.626 on NC.
The starting beta for my trade was 0.5 for each pair (0.01/(0.01+0.01)), that was different from 0.373 and 0.626
The ending beta was 0.03/0.05 for AC (0.6) and 0.02/0.05 for NC (0.4) that was pretty near to the expected betas.
After that, the trade was more balanced and finished with profit
NOTE: I don’t know if it is right to call it “beta”, or I should call it “factor” or another name… please teach me hehehe
I’m trading with real money, pretty happy with the system, after adapted to myself
The basic theory is the same: statistical arbitrage and return to the mean
I was studying the t-student distribution instead the normal distribution to calculate the mean and how many standard deviations should be expected. Still don’t understand it very well, but the Kelton’s rule of 20 pips is good enough for 5m time frame.
The [B]“TW_ATR_Channels”[/B] indicator on a EUR/GBP candlestick or bars chart is a great way to find the accurate spot for separation entry points IMHO. That is for trading the EUR/USD and GBP/USD pairs.
I would have never been in the losing trade that I am in had I used it.
Crickey, where did that lot come from?
Apologies if you don’t like what is being discussed on this thread, and if you know everything already then well done, good for you. Some people don’t know everything already and use this public forum as a place for discussion.
You keep mentioning how this thread should only be used to discuss Keltons “original” method, and it should not be used for anything else. I have just read the first post in this thread (again), and it doesn’t say anything about overlaying charts, or measuring spreads in pips, or anything about precisely how the strategy should be executed. It is just a post on a principle of trading a strategy. The rest of the thread has been a discussion as to how to go about it. If you don’t feel it is appropriate to discuss the various ways about it, the maths, the trades and the theories, then don’t. You can obviously feel free to contribute exactly as you wish to.
I on the other hand will continue to discuss different approaches, styles and ideas, all around the same strategy/principle. I do this in the pursuit of knowledge as I believe one can learn something new every day. I don’t even need to shout about it, I just do it. So, I will continue to discuss different approaches, styles and ideas. I will also continue to talk about the maths and I will also explain it, as opposed to just mentioning something and then say it is far too complicated for anyone else to understand…
This is an excellent thread to discuss all aspects of pairs trading. Hopefully, it will continue to be an excellent thread to discuss all aspects of pairs trading…
Today was a very slow day for me… do you found some useful trades this Thursday?
There is no set amount. Whatever works best for your system will do. I prefer to look at 2+ weeks of data so whatever time frame your chart is on, this can be like 250 bars on a H1.
. just chimming in… after some research of Kelton’s origonal strategy ([B]Thanks Again Kelton if your silently following[/B]), I have just closed my first position in my live acct. (winner! :D) There is one thing that I do differently though, the charting on my brokerage’s trading platform (I don’t have MetaTrader, not yet atleast, maybe later) is not conducive to overlaying charts. So I use free online charts to do my research before deciding which pairs to trade. As it has been pointed out several times on this thread; the [B]trick[/B] is chosing the [B]time frame that fits your style[/B], (I like 1H and 15M) and more importantely, [B]scaling the overlayed charts to accurately reflect the spread in correlation [/B]between the pairs you are trading. Because I’m limited in my ability to manipulate the scale in charts that I view, I simply change the periods displayed (i.e. 15M chart of pair abc/123 [B]725 periods[/B] and pair xyz/789 [B]also 725 periods[/B])on a given chart until i get something that is ballpark (I have had success whith out perfect overlays) then i do something similar to Timehoppers post on calculating standard dviation except I find 2 or 3 points of interest (convergence) on the overlayed charts (preferrably 3, where 1 point is located at bottom 1/3 of chart, 2nd in middle 1/3 and 3rd in upper 1/3) and subtract the difference between the two pairs and compare with diffrence of 2 or 3 points (divergence) where there is a spread of say 20 pips as Kelton had suggested. By comparing these points of interest and the said diffrence between the spreads then i have an idea if they are overlayed correctely. Again, this is ballpark and is what is working right now for me for I am sure as the market will evolve so will my approach to this strategy. Happy trading all. :53:
I should have been more clear. Normal data refers to data that conforms to the normal distribution (link previously posted), not to be confused with the term ‘normal’ as in ordinary. The statistics you quoted only apply to a normal distribution. Price data for the most part does not follow the normal distribution, so the statistics posted will tend to be overly optimistic. An example of a normal distribution is a sine wave. Market data doesn’t look like a sine wave very often.
The goal (at least in my view) in combining price series via pairs trading should be to normalize the data as much as possible while still maintaining trading opportunity. This is the point behind using sizes other than equal sizing to alter the shape of the spread curve to look more like a sine wave and less like regular price data (with periods of trend).
The implication of relying on statistics that don’t apply to the underlying data is that positions may be sized too large based on these assumptions, and risk might be too high. Your money management scheme seems prudent, but there are others reading these messages who might draw the wrong conclusion for position sizing from 0.3% (3rd deviation) or 5% (2nd deviation). I ran a quick study in Excel and it appears with a 15,000 bar length (about 2 weeks) and 3 standard deviations on 1 minute data that the percentage excursion outside the 3rd deviation is closer to 0.7% on data from 3/23/12 to 4/9/12 (note this doesn’t include the most recent data) and 35% at the 2nd standard deviation, which is significantly higher than 5% deviation were this data normal. Some may think 95% within 2 deviations sounds good but they ought to know that the reality is closer to 35% excursion beyond 2 deviations and might be higher if the study is applied to more data. This could be useful information when developing a robust and realistic money management strategy.
Interestingly, I use a similar amount; 200 bars on the 1H chart. I was going to look into a longer period, maybe 400ish, to see how it is different, but haven’t had the time to do that yet.
I had no trades with this strategy yesterday, but I don’t think it was a particularly slow day, just a day without any trades…
That is very interesting. 35%, really, wow that is high. I think that is an important fact to be aware of. I was going to do a similar exercise, to see how often price moved x standard deviations, over a period of time (and for longer periods it would obviously require a longer timeframe). Knowing this information would obviously be very helpful in allowing entries at points where statistically price is more likely to revert, rather than keep moving apart.
Today all my open trades were closed for profit. I had no losing trade since I started trading this way! That is AMAZING! and no, this is not scam, spam, or whatever ejjejej, I’m really happy, and I hope this strategy works on the long term.
And for all the gurus here (well, I think there are some people here that has more knowledge on statistics than me)
If I calculate the beta with covariance/variance using one of the pairs as the point of comparison? Let say for EU/GU
betaGU=cov(GU,EU)/var(EU)
betaEU=cov(EU,EU)/var(EU) (this should be 1, I think)
Great result, well done.
yes, the beta of the EU against the EU would be 1.
Great to hear that YET ANOTHER person has not lost. Keep up the good work. and PLEASE DON’T CHANGE A THING AND REACH YOUR DREAMS.
Nice to have a SIMPLE AND RELIABLE system with a very high SUCCESS RATE huh?
If you were tuning this strategy, what do you prefer? More trades per day or less risk per trade?
Let say that you have 1.1 trades per day with a max risk of 170 pips or 0.5 trades per day with a risk of 110 pips.
What do you chose?
Yeah!
I think that you said that if this system does not win the system of the month, something is wrong. I agree with you hehehehe.