Pairs Trading ??? any new inputs ?

[kkseal]

http://www.pairtrader.com/index.php

http://en.wikipedia.org/wiki/D._E._Shaw_&_Co.

Thanks CV Sir , for the links.

Liked the D.E. Shaw valuation arbitrage where they invest in a bankrupt company with valuable assets.
We also have a good semi-desi example in Lakshmi Mittal. This guy goes one step further - he buys undervalued assets and value-adds to them. If this can be done by leveraging your existing assets without too much of fresh capital infusion then it'd be the finest business model.

Microsoft is currently trying the same thing with *****, but they should have done this during the internet-bubble-burst period when they could have got it cheaper with 'valuable (& complimentary) assets'. Post Google, things have bounced back to the other extreme of the 'rubber band' and the mean has also risen considerably.


Regards,
Kalyan.

 

[kkseal]

Coming back to the pairs part, why does the 'spread' need to be calculated in the form of ratios or differentials? This is more of the ATR variety - more empirical than statistical.
Also, good for hedging but not much of a value-add imho.

Regards,
Kalyan.

 

[kkseal]

Sorry, missed out on one thing - the lagged response - that could be useful. But is this much advantage worth the addl effort? Valuation gap within sectors would do the same job. Technically there's RS, momentum, alpha to play around with.
Also this correlation, spread, differential etc. keeps changing. Take INFOSYS & ROLTA for example. Only about a year back when the two moved in conjunction sectorally, INFOSYS was the outperformer by far. Today ROLTA is outperforming.

Regards,
Kalyan.

 

[rajivkhatlawala]

Friends ,

I am mainly refering to the pairs trading among stocks and not Forex. I presume pair trading can be quite usefel especially in volatile conditions and so if any one has any 'non-traditional' ideas on the subject - do let me know

Thanks
CA Rajiv D Khatlawala

 

[ivanboesky]

Try cointegration...

 

[skarpio]

I read about cointegration and the models used. However, I was unable to get my hands on a paper or two that would define the two-step method or the Johansen procedure. Can someone provide a link to papers related to these methods?

One question I've had: what happens if the stock splits? Is this taken care of in these models?

 

[rajivkhatlawala]

Thanks ivanboesky

I think i shall look in to your suggested co-integration statistical analysis ...

CA Rajiv D Khatlawala

 

[CreditViolet]

Thanks ivanboesky

I think i shall look in to your suggested co-integration statistical analysis ...

CA Rajiv D Khatlawala

You still havent mentioned your big idea?

 

[rajivkhatlawala]

Creditviolet

I am checking to what extent pair trading can be useful in our markets .. i have tested correlation analysis and want to check whether any other approaches are available .

CA Rajiv D Khatlawala

 

[oxusmorouz]

If this input is anything to go by...

{RSC Line}
a:= (C/ValueWhen(1,cum(1) = 1,C) - 1)*100 ; {ROC of current close over 1st bar}
b:= Security("E:\New MS Data\NSE\S&P CNX Nifty",C); {nifty's close}
d:= (b/ValueWhen(1,cum(1) = 1,b) - 1)*100; {ROC of nifty's close over 1st bar}
f:= a - d; {cum value of a - d}
g:= A 10 day JMA (or a 5 day EMA) of f;
g

Buy condition: {Go long on stock and short nifty}
If RSC Line > 0 and previous value of RSC line < 0

Sell condition: {Short stock - long nifty}
If RSC line < 0 and previous value of RSC line > 0

Buy exit condition:
If cum value of price differential = 20% or sell condition

Short cover condition:
If cum value of price differential = - 20% or buy condition

If the price differential is greater than 20% or lesser than -20%, restate the 1st bar to current date and start afresh.

I've not been able to improvise on the above idea.

A very similar but simpler substitute would be:
1st bar close = "a"
Stay long if current close > a
Exit longs if current close < a

Objective: Minimizing negative equity whilst staying achieving ~ buy and hold returns during the given period.

However, unless the software allows creating a new data series for spread, and during testing, it is able to retrieve original price series, I doubt if good progress can be made in the stock spread trading front.