What is High Frequency Trading?

Four terms that you hear over and over again in this industry:

1. High Frequency Trading
2. Low Latency Trading
3. Algorithmic Trading
4. Black Box Trading

A lot of traders (and publications) use these terms interchangeably - but these terms refer to different (though related) topics.

1. High Frequency Trading: Placing a lot of orders in a short time period. Usually this term implies short-lived positions of a few minutes to less than a second (sometimes less than a millisecond). High frequency trading is all about making a lot of small profits, not buying one instrument and holding it for 2 years before cashing in. High frequency traders (or, rather, their systems) buy and sell quickly and frequently. This means that high freqency trading can make a lot of money very quickly, but it also means that without proper risk controls, high frequency trading can lose a lot of money very quickly.
2. Algorithmic Trading: Trading that leverages computing technology for analysis, support, assistance, computation, market entry, or anything else that requires some level of automation. I tend to think of algorithmic trading in three categories:
1. Order Management: algorithms are employed to manage orders that humans have placed - typically this involves technology that masks the intentions of the user (such as reserve orders) or minimizes the impact of the user (such as iceberging orders). This type of algorithmic trading technology is usually the simplest of the three, though much more sophisticated models may get pretty complicated.
2. Supporting Intelligence: algorithms are employed to provide analytical information about the market for order entry, replacement of orders, to achieve better fill rates, or to see a clearer picture of the market. This is a common type of algorithm - it's a hybrid model where the computer and human are working together (like a cyborg!). A lot of commercial software falls into this category.
3. Automation: algorithms are employed to analyze the market, place orders, close positions, analyze risk, and any activities that do not require human intervention (or very little human intervention).
3. Black Box Trading: Automated trading with little oversight from humans, as well as little to no insight into the inner workings of the system (at least, from the trader's point of view). Since most strategies require significant oversight and constant re-working, they fall more into a "grey box" category. They're not black box, but they're not quite white box systems, either. White box systems are usually more into the category of complex event processing - they're very transparent and require a lot of interaction from the trader - the trader is regularly updating parameters to the strategy, as well as the strategy's logic.
4. Low Latency Trading: Usually high frequency trading enabled by low latency connections. Low is such a relative term that it makes it hard to pin down exactly what "low latency" means. A decade ago, low latency might be a few milliseconds. Today it might mean a few microseconds. This term also may pertain to different areas of technology: systems, networking, processing, analytics, etc. The drive for lower and lower latency is what pushes traders and firms into grid computing, hardware accelerated analysis, utilizing GPUs, and other creative ways to squeeze every last bit of performance out of the hardware. There's money to be made here, but it's very much an arms race and requires continual optimization and spending lots of money.