Generally speaking for most of us the shorter the time frame the harder it is to make money, which makes sense, i mean it takes times for profits to grow, however if you have the capital and the technology "HFT" can certainly be done. I haven't traded manually in more than a year except to get out of stuff if the automated stuff occassionaly crashes. (nothing full proof). :-( There is little discussion of it on msg boards because the edges are more likely to be arb like than other trading methods that use longer time frames. Some examples from the past are things like opening only orders and hitting regionals and ecn's on listed then laying off the specialist or visa versa. Stuff like that was the bread and butter of some traders as recently as a year and a half to 2 years ago, now automated systems are set up to trade the open, arb the locks and crosses which are increasingly rare, as well as do a bunch of other stuff the most obvious of which is to fade volatility. (most HFT trading systems that I have seen that worked trade like market makers) What kind of other stuff? Well like someone else alluded to....your not gonna find it discussed much msg boards cause alot of the stuff, it can be replicated, and all it takes is a place with cheaper rates, faster code/connectivity and all of a sudden someone is running the posted strategy only they get the fills cause they are faster. The Penn-Lehman Project has spurred some intersting papers but the way they simulate the mkt leaves a lot to be desired. I haven't read any papers from this year if they did it again, but typically the "SOBI" type strategies tend to work the best...howver that is to be expected if you look at how they run the simulation (they used to use ECN data and they would base it off a stock like MSFT, that same strategy run on IBM (any listed) stock) would likely be worthless because the book is setup completely differently in listed stocks or even less liquid nasdaq stocks for that matter. Long story short though: Its not not hype, but the edges are mainly carved out of technology, speed, commission rates, etc and the general consensus is that the easy money has been made. Hell 4-5 years ago if daytraders could have automated the stuff we have running today, 50-100K a day would have been the norm, on a high frequency level the market is vastly more efficient than it was just a few years ago. When I started trading Instinet would quote 3-5 cents crosses on nasdaq stocks all the time but not display on INCA. You could literally soes or select net someone get filled and flip it to instinet and make 3 cents on 5-10k shares sometimes before the arb was gone. Then the boxes started to take over.
In my opinion, large time scale movements are rather random around a certain trend (mainly economical). The only quite well predictable movements are short term ones, where dynamics of orders placement and orders matching are the strongest. short term = 0-30 minutes long term = over one day
i think each ones personal definition of high frequency on this thread ranges from one to one thousand trades per day ...
lol...what's the matter you didn't like the academic papers stashed there? EricP has been doing this stuff for years. Maybe he will stop by and offer a thought or two...