Yes absolutely. Thank you very much. I was aware that forward testing would be required but had forgot to include it in my previous post. I suppose each pair should be treated as an individual strategy and forward tested accordingly. It will be a hell of a lot of work for an amateur programmer, if that!
I did my own in Matlab, but there is an easier way. Find cointegrated pairs at some point in time, save those pairs, then if you have PTF, simply input the pairs you found... and run the backtest on those pairs only, your output file would be very accuate to what you would have done live. This simple solution is probably going to save you a months worth of work. In addition it will give you what your desire, a correct as if scenerio with out using previously unknown information. c'mon you should have thought of these things already already!
what exactly is the cointegration math? Anyone would post the steps of calculating cointegration and the significant?
It is not easy to describe a 'futuristic' cointegration formula if you want it to be constructed based on other 'known' associated variables from the future. However, just like we are able to view historical correlation figures up to this present moment, we may define a simple 'relative' cointegration ratio based on the tendency for spread charts to converge/diverge within a fixed time frame: For e.g, intraday chart http://finance.yahoo.com/echarts?s=GOOG+Interactive#chart5:symbol=goog;range=1d;compare=aapl+^ixic;indicator=volume;charttype=line;crosshair=on;ohlcvalues=0;logscale=on;source=undefined Count the no. of times the lines converge/touch intraday for each day of the last 5 days, add them up divided by 5 ....=C2 Count the no. of time the lines converge/touch intraday for yesterday. =C1 C1/C2 gives a feel of latest 'cointegration' for me ... just wish someone could code this C1/C2 figure on some website for quick reference...
@abhave: I work with "pairs" in the FX space as a portfolio manager and am happy with my results. My experience is that you need to approach the FX markets with a far looser style than equities. No co-integration calculations and I do not screen for things like RSI and correlation in the last x period. I follow 14 currencies and track round 100 baskets, generating around 80-120 positions per month and profitable trades between 70-90% every month (yes, 90%). The results are volatile between break-even and very profitable, and the open drawdowns can be larger than most are used to handling. However, if you have studied position sizing you should be familiar with the need for a certain risk appetite to maximize your profits. But FX is perfect for position sizing and access to leverage. Trades can happen all the time as I'm constantly screening for levels to be met. I use midnight EST to change the date. Also, I very rarely trade some of the larger crosses like EUR/USD, EUR/JPY, although USD/JPY, USD/CAD and GBP/USD sometimes have trades on. What's interesting is that if you track many baskets you end up with a portfolio of exposures which can sometimes be very skewed towards long commodities/risk (AUD, CAD, NOK) and short risk (USD, JPY, CHF) as an example. Thus I end up taking on some additional risk since this exposure is from correlated positions.
Very interesting. I am looking at the FX space intraday: Few questions: What software are you using to calculate your entry/exit levels? What time frame are you basing your statistical analysis on? Are you auto executing your trades or are they handled manually?
- I'm using PTF for the moment. (And would love something better!!) - I guess it's easily dessectible from my post about number of trades per months etc that I work in the daily timeframe. However, I watch levels intraday as well. - Manually, but working on auto-execution. With a bit more AUM I'd probably just buy a pro institutional order-platform.
Hi, Sorry if this is the wrong place to post this, but I came across this thread and I love it! I am looking for some advice, not with trading techniques etc, rather with the amount of return to expect. Quick background - recent graduate (MSc Risk Mangement), working for IB as a graduate and I am basically fed up with it simply because I'm not feeling challenged. Looking to give up a career in the City for trading. I've had this idea of pair trading for about 8 months now and have recently sold a small business (graphic design) which I was running on the side to help with my Uni finances so I've become very use to being my own boss and am finding it difficult in my current job. So...my question is, what kind of capital would I need to start up and what kind of % returns should I expect per week or per day? I mean, I have followed pairs and have a certain feel for what to expect, but theory and practice are two very different things and I was hoping some more seasoned traders would be kind enough to offer some pointers to a start up? Would you recommend pair trading to a career with an IB?