Algorithms: The Problem With Choice

Choice is good. I'm a beer connoisseur, so I do not simply walk into a bodega and grab a six-pack of Budweiser. Instead, I shop at places that cater to beer snobs, such as myself, to find something new and interesting that I've never had before.

The risk in this is sometimes I buy a brew that is absolutely rotten. The worst I can remember is a beer called Crazy Ed's Cave Creek Chili Beer that had an actual chili pepper floating in it. It made me want to puke and wash my mouth out with kerosene.

But that's the worst-case scenario. For trading shops toying around with new algorithms, the worst case can result in money lost and, as seen with Knight Capital, the destruction of a reputation that took years to craft.

Earlier this week, my colleague Jake Thomases wrote a story on this subject based on a panel discussion at TradeTech West in San Francisco.

Panelist Kurt Eckert, principal at Chicago prop trading firm Wolverine Trading, says he is inundated by vendors pitching their algos—so there's plenty of choice—but the challenge is figuring out what they do once they are live. To help in this process, Wolverine created a solution that tests each algo across all lit and dark markets using the same set of metrics.

The problem that many buy-side firms face when picking new algorithms is that not only are there so many choices, but that many of those choices are customizable, thus setting up a near-limitless number of options. Furthermore, once selected, the algo must be tested at least on an annual basis. This is all happening while the lifecycle of an algo is decreasing.

According to Stephen Temes, the former CEO of Lincoln Capital, and Simon Garland, chief strategist at Kx Systems, that lifecycle can last from a matter of months to as few as a couple weeks.

Clearly, for vendors, it's a delicate balance to provide an acceptable set of options without going overboard and frustrating potential clients with too many choices. I'd like to hear how you go about finding that equilibrium.

And I'd also be interested to hear from hedge funds about what they look for in a vendor pitch when it comes to selecting new algorithms. Send me an email at anthony.malakian@incisivemedia.com or give me a call at +1 646-490-3973.