How many trades per second are we talking about?
What kind of strategies are used in this time frame?
Can the small guy play the game?
Quantitative Finance Stack Exchange is a question and answer site for finance professionals and academics. It only takes a minute to sign up.Sign up to join this community
You could for example look at this research paper released by Deutsche Bank's Research group (mirror) just yesterday which defines both high-frequency and ultra-high-frequency trading.
In the paper it says
Typically, a high frequency trader would not hold a position open for more than a few seconds. Empirical evidence reveals that the average U.S. stock is held for 22 seconds.
And in a footnote it says
There even is a subcategory of high-frequency trading, Ultra-HFT, which is sensitive to a latency down to the microsecond. Here, co-location [of servers] is exceedingly significant, and shaving off further microseconds is of utmost importance.
And no, the small guy can't play for reason well-put in the paper, co-location probably being the single most important one.
A survey by FinAlternatives in 2009 concluded that "86% believe that the term “high-frequency trading” referred strictly to holding periods of only one day or less." (Aldridge 2009):
There are two problems with this survey for our present discussion: (1) the meaning of the term has been clarified significantly since that survey and (2) it surveyed a wide spectrum of people. This latter point isn't necessarily a problem, but it amounts to asking non-experts questions about a very esoteric field: you are bound to get a wide variance in the answers. High frequency trading is primarily a proprietary trading phenomenon (including when it is done by banks like Goldman Sachs), because its primary benefit is very high returns with low risk on a small amount of capital. Trading at high speeds is not a high-capacity strategy, so it is not as suitable to a hedge fund structure which is (a) fee-driven and (b) subject to client scrutiny.
Long story short: in order to have any meaning, HFT needs to mean something akin to latency arbitrage, in which case co-location and going as fast as possible are critical. This means different things depending on the asset class, strategy, etc. But firms like GETCO are making edits to their Linux distributions in order to get higher speeds: that's your competition.
My definition is not pretty, but it's practical: If you trade based on 5- or 10-minute bars, I call that high-frequency trading. If you trade based on tick-by-tick data, including bids and offers, I call that ultra-high frequency trading.
(Trading 1-minute bars is somewhere in between. Trading more slowly than 10-minute bars is "day trading".)
I make this distinction because the distribution of tick data is very different than the distribution of 5-minute bars. 5-minute bars behave like daily data or like the data from your Time Series Analysis class. In contrast, tick data is essentially an integer-valued process, occasionally moving by a tick or two but usually unchanged, generating data at incredible speed. Its distribution is very different and resembles a Poisson process -- nothing like a conventional time series.
The distinction also determines the required caliber of your network and computing gear. You need a very fast connection and a very fast computer to process tick data. A desktop computer with a broadband connection could process 5-minute bars.
Bottom line: trading tick data requires a fat pipe to the exchange, a blazingly fast computer, and a whole new data model. Trading 5-minute data is much more like conventional trading happening in real-time. That's a big difference.
I. Re: # of trades...
To give some idea of the data volumes, the Nasdaq TotalView ITCH feed, which is every event in every instrument traded on the Nasdaq, can have data rates of 20+ gigabytes/day with spikes of 3 megabytes/second or more. The individual messages average about 20 bytes each so this means handling 100,000-200,000 messages per second during high volume periods.
Doesn't speak to # of trades executed, but in terms of # of transactions analyzed by an HFT trading sytem in search of opportunistic trades...
II. re: "can the small guy play the game?"...
Can't recommend highly enough reading WK Selph's post on "How I Built a Startup HFT Firm", e.g.
He started an HFT startup with only one partner:
my focus will be mainly on the technical challenges I faced as the engineering half of a two person HFT startup, the business challenges, such as acquiring trading capital and negotiating ever-lower trading costs, were similarly massive.
On the never-ending challenge for lower cost-per-trade:
Another problem, especially acute in high frequency trading, is that the cost per trade directly impacts the profitability of a trading strategy so, for a given per-trade cost, strategies that would otherwise be profitable are not. Clearing firms will only provide a low per-trade cost to customers that execute many trades, but to execute many trades one needs a low per-trade cost. So in addition to the purely technical challenges, building a HFT firm from scratch means solving two chicken-and-egg problems, that of building a track record and that of negotiating a low per-trade cost, simultaneously.
HFT can be loosely defined as any strategy where your profitability is a function of latency.
The defining characteristic of "high-frequency" is not the number of trades, but instead it is the number of orders you place, and in particular how often you are changing those orders. The scratch rate (cancel/fill ratio) is often very high. For every 1,000 orders you place, you might get 5 fills. This is the single most defining criteria of whether someone is playing the HFT game.
The reason this is the defining criteria is because the type of edge being captured by the high-frequency players is fundamentally different than that of other market participants. Whether its micro-structure oriented or market making, the edge a HFT is trying to capture is typically very small and also very dependent on factors such as receiving a rebate, getting a fill on a specific exchange, or achieving the highest possible queue placement at a given price level.
Speed is important in almost all realms of intraday trading, but that doesn't mean you're a HFT. Being fast is important to intraday stat arb traders even if they hold their position for 4 hours. The reason is because their signal might hit 80% realization very quickly. The faster you can act the more edge you capture. That doesn't mean folks playing the stat arb game are HFT, they most certainly aren't. However, when viewed from the outside the layman often sees no difference.
That being said, any attempt to classify most firms will fail. Most do a mix of everything, but they all have their bread and butter.
To the question of whether a small guy can do it: yes you can. Can you do it on 25k? Probably not. Can you do it on 150-250k? Yes.
"Major exchanges now have “latencies” of around one millisecond (one thousandth of a second) or less. Exchanges and practitioners now routinely time stamp their messages to the millisecond."
This is so fast that effects from special relativity come into play!
"Rules that attempt to force uniform prices at one moment in time over geographically decentralized places must take into account the fact that information does not move faster than light. At the same moment in time, two geographically separated participants may observe two different “best” prices."
All quotes are from this mind-boggling paper from the UK Government’s Foresight Project:
Impact of special relativity on securities regulation by James J. Angel
At the highest possible.
The High in HFT refers to the frequency of data, not of trading decisions [which varies quite a bit]. Data is nearly always tic by tic; computers typically synchronized with costly clocks servers.
It might have originated from the Fourier transform, and would indirectly point to speed of data once again. HFT then would be a good complement to FFT and DFT :-)
There is a very god paper by A Menkveld analyzing the activity of a huge HFT player during 2 years:
The best is to read the full article. You will learn: