Fast Trading is not High Frequency Trading

Are't these denominations synonyms?

They seem so, but on the beginning of the algorithmic trading era, back in 2005-2007, the status quo choose to use the name HFT (or High Frequency Trading) for algorithms that operate faster than a human possibly could. On this article I'll show that what most consider to be "High Frequency Trading" is a misnomer and I'll introduce what appears to be a synonym, "Fast Trading Algorithms", in an attempt to clarify the denomination problem.

Nowadays, circa 2019, High Frequency Trading could better be called "As Fast as Possible Trading" -- AFAPT -- or "Maximum Frequency Trading" -- MFT -- which, not only sounds as cool as HFT, but is also more in sync with the HFT purpose, historically speaking. The reasons why the concept of "High Frequency Trading" is actually "Maximum Frequency Trading" are discussed in more details in the article What really is High Frequency Trading and who should feel threatened by it? -- by now, keep with me only in recognizing that there is a gap between "High Frequency" (or "Fast") and "Maximum Frequency" (or "As Fast as Possible"). We will now dig into those gaps, aiming at differentiating the purpose of the associated algorithms.

When automated trading was introduced (year ~2000) large investment banks and hedge funds went on an obvious race (which they were already doing) on who had the most lucrative pair: methodology and operation. The first algorithms they automated were those they were already familiar with and knew the results -- because they were operated manually -- and, while big companies could afford having humans doing the work machines should be doing, the first algorithms to be automated were the most simple ones -- some departments of those large companies used to create the methodologies and other departments, with cheaper workforce, were tasked to operate them. Back in 2000's, any algorithm automating those methodologies would bring advantages -- time based advantages -- over any human operated methodologies their competitors had. Trivial examples of such methodology are "market making" and "latency arbitrage".

So, the key factors one must keep in mind in really understanding what High Frequency Trading is all about, are, so far:

• Apart from some inevitable exceptions, they are not doing anything new. Most strategies were here for a long, long time and were already playing along with everybody else -- they were already executed "as fast as possible";
• For strategies where time based advantages do exist, competition favored the faster ones over the course of time. Nowadays this speed is enormous (using human standards) just because they evolved playing together on this environment for quite some time now;
• The involved strategies were (and still are) well known among players -- this is why they compete on time, not on the model;
• HFTs usually compete with other HFTs. We are talking in the thousands of operations per second range -- or even more. Usually they win or lose 1 cent for each instrument unit they negotiate. HFTs only care for other HFTs and vice-versa -- from whom they'd win/lose money at these speeds, after all?
• Although HFT might give or take 1 cent on specific transactions, on the long run, they won't interfere with your non HFT strategies, whether they are manual or algorithmic in nature -- note that HFTs do crash and they may disrupt markets -- but, utterly, they are the ones who pay for it;
• The simple strategies, with simple input triggers, continue as of today -- and the speed will continue rising. But, remember: There is a physical limitation involved here: for the same hardware and software, one cannot be "the fastest" and "the cleverest" at the same time.

Now we are at a point in history in which:

• Economists -- who create their custom investment strategies -- are documenting their methods and modelling their knowledge;
• More and more not so large companies can afford the infrastructure necessary to do automated trading -- not "as fast as possible" though;
• Cheap Super Computers allied with Computational Intelligence (or Artificial Intelligence) allow algorithms to make their decisions using much more data on a timely fashion. By "Cheap Super Computers" I mean special hardware tuned to run, at the same time, many threads of special software with no input/output bottlenecks -- fuzzy logic and neural networks + GPUs work like that;
• Most clever people no longer value working for large banks as they used in the past - they build their startup or investment fund instead.

The gap between High Frequency Trading and Fast Trading should be clear now:

• HFT Algorithms are all about implementing known strategies which depend on them being the first to react to be lucrative;
• Fast Trading Algorithms implement custom strategies (typically using much more data as input in the decision making process) and being less sensitive on timings -- after all, they are not competing directly with no one else. Fast Trading algorithms usually have a 1-5 second decision making window and complete the buy/sell cycle in 1-10 minute range.

Now, to conclude, these two kinds of strategies all have their space on the ecosystem: there are advantages both in being the first to react (which is what High Frequency Trading Algorithms do) and in being the only one to notice (which is what Fast Trading Algorithms should do).

If your "Fast Trading" algorithm depends on it being the first to react, you should conclude your strategy is known by many and you'd better consider selling it to a bank or put it in the Hall of Algo Trading Failures.

On the other hand, if you work for a bank and came up with an algorithm which perform slow because much analysis needs to be done, consider testing it in Ogre Robot and quitting your job :)