Issues with algo wheels remain in both the equities and futures space, despite growing popularity of the software, according to market participants on a panel discussion at the FIA’s IDX in London this week.
“It is a complex problem,” said Sanji Shivalingam, global head of algorithms and analytics at UBS. “What we do see is more and more firms are going down that approach, especially now after Mifid went live. What is complicating matters is that each firm is taking a slightly different approach with what they mean by: ‘are you meeting my benchmark.’ Some firms can gain benchmarks and do fantastically looking at the gains of a benchmark. If you’re playing the game according to the rules, you probably don’t. Again, it comes down to the questions you are asking and then at the lower end to the scale the mathematical side of it.
“The problem that we face in equities is as I say, you’ve got 10 brokers on your panel as part of the wheel and as you say nine and ten become deprioritized to some extent,” he said.
An algo wheel allows traders to automate choices on optimal brokers or algorithms, by helping to review both quickly and unbiasedly.
In the futures market it is more difficult to do accurate side by side comparisons of performance, particularly with futures transaction cost analysis (TCA)s being harder to measure than equities TCA, according to Gordon Ball, EMEA head of electronic execution and algorithmic trading, Citigroup Global Markets.
“There is a lot more of a disparity of contract liquidity. For example, if one broker ends up with the equity contracts and the equity futures, another broker ends up with the commodities futures contracts. Not surprising the commodities future broker is going to look worse on that same benchmark,” said Ball.
“So, it is hard unless on the buy-side you are really scientific about how you fairly split flow across. I think what we’ve found is brokers will try it. Buy brokers will get 20%, but then there won’t be that follow through to say, ‘well, maybe we ought to change the 20% for each of the five to bias one versus another.’
“That’s not to say nobody does it. There are a couple of clients who are really good at it, but I think the larger group of clients that we talk to want to do it, but either haven’t got enough flow to help make the stats work for them, or haven’t got the rigorous testing to be able to draw results from using it,” he said.
Responding to a question on what technology the panel was looking to invest in the future, Shivalingam said there is an arms race in equities to leverage tools for benchmarks.
“It is becoming more and more about benchmarks. Quants don’t really care about the algorithm that is behind it, they just want to know, ‘are you meeting my benchmark, and if they are not doing that then I’m going to go to the next guy,’” said Shivalingam.
“While the data service proposition is important, if I look forward especially as Mifid increases the scope of performance and the end buy-side client having fiduciary obligations to their investors, I think where we are seeing it at least in equities is geared more and more towards performance against benchmarks, less so on ‘can you offer me yet another variant of an ISR pro.’ Basically it is becoming benchmark to benchmark,” he said.