What is 100% true is that having superior information is a key for successful high-frequency trading. From my every-day experience quality of data you receive from market data providers and execution speed are critical (we are talking here about milliseconds). You do not need to do any fundamental analisys - at this "frequency" it is senseless . It is like particles theory has nothing to do with Newton's physics.
Actually I would not make such an analogy. Understanding the value of what you trade is always worth something. Regardless of frequency. An arbitrageur or market maker can follow the order flow closely, but if he gets stuck on one leg or with residual inventory because of a bad news, he's writing a loss. It's not fundamental information that matters as much as differential information. A flow trader doesn't have to analyse financial statements, but he should be aware of rumors and news - as are his peers - if he wants to understand the meaning of his order flow. That is also why I lack faith in automated trading. Machines are the last ones to fully know about the bad news. People should learn their corporates and economics before getting their foot on the markets, else they might end up being someone else's dinner.
I would say that it is possible to use simple "mechanistic" approach to using of fundamental market information (and it is exactly what we are doing). A machine (or simple put an algorithm implementation) will scan a news feed and depends on its finding will act correspondingly to pre-defined "scenarios". Now, the question is who defined these scenarios, if the person is good in understanding of all aspects of potential development and correlation between news data and market behavior. From what I have seen it is not possible to be 100% correct but statistics shows that 80% of being on right side is quite reachable.
For 11 months. As we are acquiring experience we are adding different trading models. But foundation remains the same - low latency of data feeds and fast orders execution.
Please be reasonable. Our approach does not tolerate the risk of this magnitude. It is an arbitrage - the system takes position for a period of fraction of second and tries to unwind it in another market as soon as we got a fil or partial filll. If it fails to unwind it for profit for a reasonable period (usually 5 seconds), a stop order is issued. As a result losses are pretty limited. So as you can see for our model speed is everything. That is why we process market data in-house because other solution fail to meet necessary latency requirements. Thx, Dersu