True high-frequency is very hard. Most of these guys got started in the 1999/2000 time frame and didn't have to deal with many of the start-up issues that new e

True high-frequency is very hard. Most of these guys got started in the 1999/2000 time frame and didn't have to deal with many of the start-up issues that new entrants face. For example, just consider the data you need to trade that quickly. Not only is the real-time feed expensive, but if you want to source the data from the execution venue then you have a lot of code to write. The amount of data is also quite large. For example, the US equities market executes approximately 45 million trades per day and there are about 700 million quotes per day, not including the non-top of book quote activity in the various ECNs. I collect between 2 and 4 GB of data per day, if you were to go direct to the execution venues you'd be looking at >50 GB per day, probably. So, anyway, the "true" high-frequency game is very tough today because we're already 10+ years into it. The markets have changed and the edge has gotten smaller, but there is still plently for a lowly individual automated trader to scratch away at. Instead of looking for millisecond opportunities, look for second or minute opportunities. Go where the big guys can't because there isn't enough capacity. Can you find an edge that, on average, keeps you in a trade 30 seconds, for example? There's still plenty of alpha left, just don't step onto their playground and expect to get onto the swing set. A few practical notes: * You need to look into unbundled or cost-plus commission structures. These fee structures charge a per share commission and pass through all fees and rebates from the executing venue. This is required to do any sort of size with reasonable cost. * Most "retail" brokers are not sufficient for any sort of high-volume algorithmic trading. Interactive Brokers is barely ok if you are in any way interested in limit order trading because they have fairly large cancel fees for direct routed orders. If you're model doesn't rely heavily on strictly offering liquidity or you're ok with letting IB route your order then IB is ok and offers an unbundled commission structure. Lightspeed Trading and Lime Brokerage are two that cater to active individual and institutions. * Data storage is a big deal. Effective storage of regular (evenly spaced) and irregular time series will require you to engineer something. There are commercial solutions, but you can't afford them. When you're dealing with high-volume intraday trading this is one of the first issues you'll face. How do you store, query, and manipulate data that includes 45 million new rows per day? The relational DBs fall apart pretty quickly and even if they didn't they won't give you the time series operations you need/want. * Data feeds are expensive, but required. Look at DTN NxCore. It is a full market feed that will give you the best you can get w/o going direct to the exchange. Some brokers will give you access to raw exchange feeds, but you'll need to engineer feed handlers and a ticker plant for them. This is a non-trivial task, but not impossible. Once you've done that, you'll need to figure out how to get all that lovely data off your co-located server and back to your home base for analysis. Network engineering will be required because your broker doesn't want you pushing 10-20GB per day through their network connection, so you'll need a circuit from an on-premise carrier. * The banks are players in HFT, but not the original or best. Most of the guys that started it are still independent. Look at GETCO, RGM Advisors, etc. * Flash orders, what most folks in this thread are refering to when they say the exchange gives the HFT firm a first look, are no more. That edge existed, and I'm sure HFT took advantage, but no HFT firm was built on flash orders. When they started flash orders didn't exist. It is possible to be a successful, independent, automated trader. It is even possible to do it on a purely intraday, high-volume basis. Don't get caught up in the hype of needing to be high-frequency or not. --- EDIT: The other thing I forgot to mention is that naked sponsored access is likely going away. This is a near-requirement for "true" HFT. Any future regulation won't affect any of the existing players because they're all grown up now and most have their own broker/dealers. Some form of sponsored access will likely still exist, but pre-trade risk checks will probably be required and will therefore still leave you're broker between you and the market.