Learning curve: Measuring up to make futures algos useful

Once the sole domain of equity traders, computer algorithms have now firmly entered the futures markets' mainstream. But, with global markets so complex, and data so big, regulators are naturally demanding greater transparency — even those that have not yet formalised the obligation towards best execution. An increasingly discerning buyside is also seeking this level of transparency in deciding where to send its order flow.

With all this at stake, it no longer makes sense to force over-worked futures traders to monitor the markets by eye. Watching many thousands of different data points and making decisions according to hard and fast rules is a job best left to computers, as any air traffic controller will tell you. The skill and senses of the human trader are best applied to the difficult orders, and to steadying the book when markets jitter or orders start to slip.

As has been proved in equities, algorithms have also become key differentiators for futures firms. While the theory of using algorithms is perfectly sensible, the proof is in the execution. As algos become ubiquitous, futures market participants will soon discover that simply dreaming up another super hero name for their latest algo will be if no use at all if its powers cannot be demonstrated clearly. So how can firms prove their worth in this arena?

An obvious starting point would be to lift up equities analytics systems and drop them onto the futures desk. It sounds simple enough, but this approach quickly becomes problematic. Comparing equities with futures is less like apples and oranges and more like apples and elephants.

For starters, an equity is the same no matter where it’s traded, and the where is a foundation stone of the analytical outcome. Futures contracts are owned by the venue they’re traded on, so the where quickly becomes irrelevant.

The more global nature of futures trading is another difference, together with the fact that futures represent a wide variety of asset classes, each with their own distinct subtleties.

Stripping the where out of an equities analytics engine and focusing only on the when misses all these nuances and fails to capture the behaviours specific to each venue.

Trying to reflect these retroactively is so fiendishly complicated that it makes more sense to throw the whole thing out and start from scratch.

But such is the demand for greater transparency around algo execution, that in-house analytics are springing up here and there in an attempt to plug the gap. This is not the solution either. The cost of building these, then retrofitting each new algo as it comes out, and then constantly adjusting for new parameters that are added to existing algos, makes it an extremely costly and never ending exercise.

For the very immediate future the cost may be bearable, but for every day the markets grow in complexity, and algos grow more numerous, this approach will becomes less feasible for anyone who hasn’t made a business out of building analytics engines.

In-house developed solutions also provide no defence against accusations of a firm marking its own homework. While in-house analytics may be, and probably are, completely agnostic, there is always room for suspicion wherever maker and marker are in the same bed.

A third approach is to bury the collective corporate head in the sand and hope nobody asks for objective proof about how good the algos really are. As the markets evolve and analytics become part of normality, this is fast ceasing to be an option.

With the use of algos in futures markets only going in one direction, what can be done to provide the kind of transparency the industry needs?

There are many benefits to getting analytics right. The first is the proof that the algo really is doing what it's designed to do. Really good analytics also allow traders to drop orders in and out of algos according to market conditions, letting the algos do the legwork and using human nuance to keep tweaking the book, taking advantage of opportunities as they arise and avoiding pot holes as they become apparent.

This on-the-fly capability can be built into a real time algorithmic consulting service — still a new and attractive proposition for many prospective clients. Taking this a step further, analytics can be boxed up and given to the buyside allowing them to see their orders as they unfold. Any firm allowing its customers to analyse and track the performance of its algos with a third-party system is demonstrating a level of confidence that many clients would find hard to resist.

Finally, good analytics can kick off a virtuous feedback loop allowing for ever better algo development. As the old adage goes, you can’t manage what you can’t measure. Measuring algos in test and live markets, and using this as the basis for continuous improvement, is a powerful prospect indeed in a world where most have no yardsticks at all.

Developing good analytics is, however, far from easy. Understanding the execution progress and participation rate is just the start, particularly when you have to support all the order types — including strategies — that futures markets throw at you.

You also need support for industry standard benchmarks such as Interval VWAP, TWAP and Arrival price. With a trader's view of the world differing depending on what asset class they are trading, that's a lot of benchmarks to monitor.

But that's not all. Understanding your fills in terms of passive versus aggressive trades and having real insight into why you underperformed or outperformed the market is also a crucial part of the assessment.

And, finally, this all has to be easy to visualise and interact with. Only then do you have the makings of an analytics capability that can genuinely provide the type of three dimensional transparency demanded from your customers, the regulators and your own compliance team.

The arms race in futures algos has already begun, and those not on the battlefield will quickly find themselves annexed into obscurity. As the battle rages on, intelligent algo analytics will be an important weapon for those with the foresight to see what it can become.