!VAR HAGENDOORN

The Aesthetics of Financial Time Series

Years ago I was invited to contribute a chapter to a book by some artists whom I'd met at the Documenta. I don't know whether the book ever saw the light of day. Shortly after submitting my project I moved to London and I lost contact with the artists who edited the book.

I had decided to do something none of the other invited artists would or could come up with, a selection of financial time series based on aesthetic criteria. I don't recall what kind of graphs I'd printed out or what I wrote in the accompanying text. I may have it somewhere on disk, it's that long ago, except that I don't have a disk drive.

Anyway I was reminded of the project when I saw these graphs. They are not graphs of actual prices or volumes, but of bid-ask quotes, sometimes several thousands in a timeframe of a few seconds. They were created by Nanex, a firm that delivers real-time streaming market datafeeds, based on the data that runs through their systems.

In the wake of what has been termed the flash crash of 6 May 2010, during which major U.S. Market indices dropped about 7% in a matter of minutes before rebounding, some analysts at Nanex decided to dig deeper into the data around the time of the crash. They discovered some intriguing patterns in the bid and ask quotes submitted to NYSE, Nasdaq and BATS, another trading platform, which they attributed to high frequency trading algorithms. They decided to continue to monitor their systems for unusual patterns.

There's a lot of debate about the origin of these patterns. Are they emergent properties caused by multiple trading bots playing against each other? Are they the product of a single algorithm? In any case, it wasn't me if that's what you're thinking!

On the topic of quote stuffing, I find this a highly dubious practice and I agree with bloggers who claim that regulators are behind the curve. It appears that the game that is being played is to feed a large number of bid/ask quotes into a trading system within a fraction of a second without the intention of honouring them. But why would anyone want to do so? The analysts at Nanex suggest that it gives competing algorithms more data to process, which may result in a competitive advantage for the firm generating the quotes, because it can filter out its own quotes. In my view this has nothing to do with high frequency trading, which is "just" trading at micro intervals of a fraction of a second. I'm rather ambivalent about the merits of high frequency trading, but that's another story.

Anyway some of the patterns are quite striking.

Update: According to a report by the SEC/CFTC the "flash crash" was not caused by quote stuffing, but by a single large sell order that was completed in 20 minutes rather than the few hours it would normally take, causing various algorithmic trading programs to start selling as well.

Update 2: Nanex disagrees with the SEC/CFTC report.