i consider myself a quant cause people pay me to be one and thats what they call me "their quant" after billions of manhours behind a cray computer i can tell you most modern "not all" market theories break down over time. if you have a great system that has made money in the past two years. when tested on 200 years it fails. the main problem quants have is they dont test with enough data. so how do you get enough data, you make it thats what i did. then when you have billions of points of data your computer will not crunch it. so what do you do, you find a cray to use. hopefully you make enough money you can buy you one. then you have a problem where to put it, how to power it, maintain it you must learn. thank god its fast cause the down time is spent cleaning up code and getting ready for the next run. i laugh when i read that this magazine did a research on predictions, or this university did this and that. its like getting the boy scouts to build an indy car. great effort and noble i respect the boy scouts but they are way out of their league. really they have no clue, in fact i would say that if you dont have a beowulf or cray or a similar setup you are not even serious as a contender to be a quant. i am not some genius, i am a guy who has a passion to risk everything like money and my health to find the truth. i think i have found some truths but i am still very passionate about continuing to look. i have better stuff than anyone has ever seen yet i am not satisfied. this is what makes a quant, and there are very few of us, but make no mistake i am one. mb ps when i create unpublished theories i don't name them with fancy names - no they just get a number like Ls0007 etc.
the only truth is this, http://en.wikipedia.org/wiki/Linear_regression the noise can be less or great, there are more then enough markets in the world where the noise is very minimal, hundreds of stocks or some futures. if your entry is within the noise, (hopefully) the edge, the noise shouldn't hit your stop. if you diversify this principle more often then not most of your trades the stops wont get hit. the timeframe regression is applied can be noisy, its best to find the timeframe with the least amount of noise or variance.
Mark - those days are over....ground-breaking new multicore processors and chips along with this terrific piece of software provide incredible performance improvements for number-crunching applications. http://www.hpcwire.com/hpc/1560982.html http://www.rapidmind.net/News-Nov5-CPU-Launch.php As usual, the hardware (new chips) was fairly worthless without the requisite software to access and utilize their power. Rumor has it that NAG is going to create a multi-core version of their software library. With this capability, even monte carlo simulations might be possible in realtime on tick data streams.
i am aware of the chips we get them from intel from time to time to play around with. i just set them on the desk and look at them myself. we have a guy here who is all nuts building stuff that emits a gamma smell cause its so fast. i came to the conclusion some time ago "cause i am a hardware junkie" i can waste my life on the bleeding edge or i can waste my life making money. so i choose to stay behind with old technology because i dont have the time to screw with the new. its not that i dont like all the new stuff but i wait for a shake out period, i will never do any deep projects on anything but a cray and tel language cause i have to many years with it. i then spend my time converting from tel to other known languages to do the trading. newer is not better but i agree it is sexier! mb
Check out the claimed performance increase for just doing the Black-Scholes formula using their API: http://www.rapidmind.net/pdfs/ClabbyAnalytics-RapidMind.pdf Not sure what the application would be for doing the BS in realtime (scalping options ?), but this provides a sense of the power of this approach and technique.
hey wiz, speaking of computer power, remember the guy at NAG who mentioned the potential within the new videocards? I found that rather amazing! Cray computers?? sounds like someone watched '2001 a space odyssey' one too many times.
Cray unveils 'hybrid' computer - they still are way on top. http://www.chippewa.com/articles/2007/11/06/news/801x.txt A space odyssey 2001? oh well you just missed by a few years. James Adamson, Mark Brown, David Leestma, Richard Richards, and Brewster Shaw, aboard the space shuttle Columbia, begin a "secret" military mission on August 9, 1989 to launch a spy satellite. Mark Brown
You bet. Mark is missing the "point": the combo of the software I mentioned and the new chips is UNBEATABLE from a price-performance standpoint. Sure, the latest Cray will beat anything....but at what price ?
No you are missing the point. Who cares about price when you trading 20,000 contracts a day and spending millions on research every year. I think you just have no clue and I like it that way.