I heard people are using SAS as an analytic software to create a strategy/backtesting. But I was wondering if anyone actually use SAS (Statistical Analysis System) as an automated trading system. Thanks
SAS is not any platform you'd want to make a system on. SAS is for linear regression and statistical analysis and involves static EOD datasets. It's not really anything to use on an automated basis, but maybe as a reference for fundamental strategies. At Fidelity, I gather a 100,000 element dataset on 2000 symbols with 50 items and regress them on the log of the market cap with the log of revenue and other items. This is the only use I've seen for both SAS and STATA. I know of no real time basis or programmable macros you could use to make a system in SAS automated.
But people are using Matlab/ R for live trading and backtesting . Does anyone use SAS as the way matlab webnar shows ? http://www.mathworks.co.kr/webex/recordings/Algtrad_092507/index.html
the only SAS i have seen in the real worl was used to create a volatility surface. its awful and was rewritten
From a fellow SAS programmer/statistician, I wouldn't reccomend using SAS for this kind of thing. Sure it has some advatages in regards to databasing, proc sql, Proc ARIMA, Proc Reg, and things of that nature but there are too many other packages out their that are better for automated trading than SAS. I started out trying to do automated trading in SAS, did alot of programming from a statistical approach, and came to find out that curve fitting (using automatically diagnosing arima modeling routines I wrote) was not the best way to go about automated trading. I have a strong background in Time Series so I thought this would be the best place to start for me...but after serveral months of work decided to go a different direction. You'll find tons of posts on "why" this is the case on the forums so I won't go into it here but I would reccomend looking elsewhere. I personally use openquant (C#) for my automated trading and program in SAS when it is needed at work. I do use R for some automated trading tasks such as pulling data from IB on occasion, ect. However, this is just my 2 cents from trying to go down the SAS road.
Some of the biggest quant firms on the street use SAS for backtesting and automation. We are not in the league with the biggest, but we force all of our new quants to learn SAS day 1 and all of our systems are written in SAS. the statistics are fast and error free and when you are developing heavy quant models, you would be surprised how many PhD's mess up the statistics with a simple error so when we bring someone in we recode their idea in SAS and it is definitely a joykill. Also it is blazing fast.
matlab is crap and they have been touting their trading modules for over ten years. but it's all junk - they are hard headed and impossible to work with much like cqg and tt but worse.