pairs trading clarification

[finalsolution]

bob:

Thanks for all the help. I can see now by looking at your chart that you basically set a starting point, and calculate the spread based off that starting point, never moving the starting date.

For instance, starting on day 1, you will continue calculating the spread against day 1 for a very long time. You do not have "spread period" where you resync the spread to zero.

I looked through your trades as well, very impressive!

My concern is using your methodology it is hard to signal a trade via a violation of the spread bands. You can't look at the chart and say, since 30 days ago the spread +2 standard deviations greater than it should be.

But I can see that your approach certainly works very well. I guess I am trying to fit my model to the one described in Professional Stock Trading.

The spead band calculation btw is--

spread band = k * (num standard deviations) * (hv(stock a,30) + hv(stock b, 30)) * (1 - R)

where k is a constant to annualize the result (sqrt(1/365))
hv = historical volatility
r = correlation coefficient


Thanks!
fs

 

[seisan]

Quote from Bruto Blukowski:

Maybe this will whet your appetite for skipping pair trading and moving onto the real bigtime stuff.


http://www.siliconvalley.com/mld/siliconvalley/6858310.htm

Brut,

A great read about Salem Abraham, and his Abraham Trading!

The Mounties will enjoy comparing notes with that story!...
...as will some futures traders!

It looks like Salem's got a few ET'ers workin' for him, to boot...


As for the rest of us - we need to straighten out those T-1 lines,
and call Dell - real fast!



... i wonder if canadian, texas, has its own mounties...

 

[djclif]

(Seisan, you make me laugh.)

Final Solution,

Sorry for not replying to you sooner, since you started a new post I did not know about your question. I generally only track the general pair trading forum. From the looks of what I am seeing in your posts, and please feel free to correct me if I am wrong, you are looking to trade spreads using standard deviation analysis, or 'spread bands' based on historical volatility, over a two to twenty day period.

You are probably interested in two things: one is a way of tracking the performance of you pair, and the other is a way of tracking entry/exit points based on your 'spread bands'. The reality of the situation is that a spread is defined by the following formula when it comes to calculating profits:

If you trade one share of Wells Fargo (WFC) for every two shares of Key Bank (KEY), and the prices are 53.23, and 26.07 then your spread is: 0.5*53.23-26.07=0.54. Say you happened to have 300 shares of WFC short and 600 shares of KEY long. When the spread number (currently 0.54) goes down by a penny you would make $6 (We refer to this setup as being short a spread, as we are short the lead stock WFC). There is no 'correct' way to determine what ratio you should trade a pair with, generally it is a balance of the volatility of each stock and the beta, or capital, of each stock. When dealing with shorter term trade where you are never rebalancing your capital this is probably the easiest way of reporting spread prices.

Recording your spreads in this fashion will also allow you to create your 'spread bands'. If you have a charting package, you can just use Bollinger Bands on the spread number. If you don't you can calculate it yourself. One thing to remember is to keep your 'spread bands' analysis to the appropriate time from you are trading. If you will be in trades 2 to 20 days then you should probably use a daily standard deviation analysis. The formula that you posted is a standard variance formula, just applied to spreads. If you use actual spread information and then let something like excel do a standard deviation from a moving average of your spread number then the observed volatility and the observed covariance are already taken into account in a plain standard deviation calculation.

I hope all of this helps, if you have any more questions let me know.

Cheers,

Darren