Using R for my stat.arb strategy .

[HedgeFudge]

Hi all ,

I am kinda new here . To introduce myself, I'm HedgeFudge from India . I've been trading for over 1.5 years in the Indian markets and currently working at a boutique IB . Cleared my CFA L1 last dec , will be appearing for my L2 next year . Anyways , ill stop beating around the bush now .

I donot have a programming or engineering background , but have been able to wrap my head around a the basic computational Finance and have made a few stat.arb stratergies which i've been trading for over 8months . I've been using Excel and trading only a few selected stocks , cause excel doesnt have the power to do many calculations for over 2k+ stocks . Hence , I am thinking of learning R . But before I do , I wanted to ask a few questions from experienced people who use R on daily basis (Or close to daily) .

Currently cause i work at an IB , I cannot concentrate on the markets all the time , so I use EOD data (End of day) . So I use a googles EOD data on excel , my question is can EOD data be uploaded to R and then can I run my codes ?

Secondly , can someone pin point me to an "already coded" stat.arb strategy ? (So , i can take bits and pieces from it , which are relevant to my strategy )

Thirdly , can research regarding quantitative aspects of 2 commodities/stocks/currencies , be done on it ? like correlation analysis , cointergration analysis etc..

Regards,
Fudge
P.s: Sorry for the noobish questions , but i have no coding idea . need to learn it from online materials

 

[Levi McClenny]

R can read CSV files, so if you can do it in excel you can do it in R. I personally use Yahoo because the data comes in CSV format already. Not sure how the data is given from Google. Never used it.

second question: if its profitable you'll likely not be able to find it anywhere.

third question: absolutely. If you can think of any sort of statistical analysis R can do it. It can absolutely handle correlations between 2,3, ... N datasets and whatever else you need it to. You'll hit the limitations of your computers memory before you hit Rs.

A good place to start learning R is Coursera. Johns Hopkins has data analysis modules on there that focus exclusively on R and teach it pretty thoroughly for beginners.

 

[HedgeFudge]

Thanks a ton Levi !!
Looks like choosing R was a good idea after all .
Regarding the strategy , I already have them, the reason I ask for other strategies already coded , is that I can just copy and paste a part of the codes from that , whichever is relavent to my strategy . would definetly reduce my workload .

Thanking ,
Fudge

 

[Martin Thomson]

Secondly , can someone pin point me to an "already coded" stat.arb strategy ? (So , i can take bits and pieces from it , which are relevant to my strategy )

Statistical arbitrage is a wide field. I guess, there may be 1000 strategies ...

 

[HedgeFudge]

Statistical arbitrage is a wide field. I guess, there may be 1000 strategies ...

Hi Martin ,

Yes , but currently i am using a very very basic stratergy of stat.arb . I am still trying to learn more , but due to the CFA and my job i hardly get the time .
If you dont mind pointing me in the direction where different types of stat.arb strategies and explanations of those are available .

Thanking ,
HedgeFudge

 

[Martin Thomson]

If you dont mind pointing me in the direction where different types of stat.arb strategies and explanations of those are available .

Well, that is not really possible. Basically you can build a statistical model to predict any financial quantity, that is used for asset selection, e.g. the P/E ratio for instance.

I think you have to decide in the first step, wether you want to predict the fair/equillibrium price of a stock or if want to estimate the future return of stock. Both approaches often lead to controverse investment advices.

Hand on the heart. Didn't you just read "statistical arbitrage" in the CFA book, respectively in its recaps, and now you want to know, what are the most common statistical arbitrage models?