Simulation Tools Key as Managed Futures Funds, CTAs Optimize Execution

“Skate to where the puck is going to be, not where it has been,” Wayne Gretzky once told an interviewer. As the Great One described it, what cuts certain players a level above in a fluid game like hockey isn’t native instinct alone, so much as endless practice seeing the ice and, frankly, the hard work of getting to where a scoring opportunity will be before it reveals itself. And doing that over and over.

Gretzky’s advice is one of Robert Almgren’s favorite lines—but not because the co-founder of Quantitative Brokers (QB) is a hockey fan. Instead, he says a similar idea applies to the business of algorithmic futures execution: the more you see, the more you test, the more instinctual an algo-based strategy can become. Changing expectations at systematic hedge funds and commodity trading advisors (CTAs) have put increasing emphasis on an active and dogged approach toward execution, with recognition that in correlated markets like interest rates, the extra step is crucial, even if it is also more expensive.

Shops both big and small now argue it’s worth it, including AHL—Man Group’s managed futures fund—and Revolution Capital Management, a Broomfield, Colo.-based CTA. And much of the value, they say, derives from what comes before any trades are even made.

Style of Play
A pair of reasons explains why the algorithmic “palette” for futures is diversifying. The first is that the style of trading is in transition, somewhat belatedly, just as funds have begun offloading more of internal brokerage functions—including algo development—built in the run-up to 2008.

What the independent algo providers—many of them, including QB and Pragma, run by former quants like Almgren—do is put a premium on quality rather than price. Sure, the uptake isn’t universal, says Tom Haldes, Trading Technologies’ product head for buy-side technology, but heads are beginning to turn.

If we review a particular algo and it looks interesting we'll onboard it with a small amount of flow. We're always conscious to compare on a like-for-like basis, and take into account the brokerage charge explicitly. - Murray Steel, AHL

“There isn’t an objective, per-strategy total cost analysis among CTAs quite yet, given the amount of latitude that lead traders still have in terms of using gut-level intuition of what’s performing better, rather than objective TCA analytics to determine the optimal algorithm on an order-by-order basis,” he says. “That is rapidly changing, though.”

According to Michael Mundt, co-founder at $650 million Revolution, firms in the space have faced pressure in recent years from a stiff headwind combining the effect of low interest rates and scalping from high-frequency trading shops. Whereas veteran futures traders could once look at a screen for six hours, picking fills they recognize or like, today that method just isn’t efficient.

“There’s a selective memory on which trades work out better: You’ll always remember how well the one went, and forget the five that didn’t,” Mundt says. “The most successful CTAs mostly started in the 1980s before electronic trading, and stuck to their ways. Eventually, the world changed underneath them—HFT entered in force around 2009, and for a few years really cleaned up. Some CTAs are looking for inexpensive protection now via algos, which they can get from their FCMs, but for others it’s really, ‘Let’s rethink this whole thing around execution.’ At Revolution, we’ll usually trade about 30,000 contracts daily on average—which is a lot relative to our size. If we can measure $2 per contract in savings with a boutique partner that costs us 50 cents per lot, as long as it’s a net saving, it makes sense to do it.”​

AHL was ahead of the curve, too, as it has been doing exactly that for a few years now, says its global head of trading, Murray Steel. The $12 billion London-based systematic hedge fund will scan different contracts at certain times of the day, generating a potential trading situation about 2,500 times, and ultimately settling on around 800 trade order structures daily.

"The decision to have a trade executed by human traders or Vtrader [the firm's proprietary order routing platform], is based on historical transaction cost," Steel tells Waters. "We capture and review execution costs data for both channels; generally smaller orders are more efficiently executed by Vtrader. Ultimately human traders are the last line of defense, and it is important that they have oversight over all our trading.  Given the nature of systematic trading, there are times when we're long or short for weeks or months at a time, so it requires patience and discipline to build up enough sample data points to understand each algorithm’s effectiveness."

Steel continues: "If we review a particular algo and it looks interesting we'll onboard it with a small amount of flow. We're always conscious to compare on a like-for-like basis, and take into account the brokerage charge explicitly," Steel continues. "We do perform reviews on all algos internally using our own simulator, and are always keen to compare these results with that of the provider. If they cannot provide a simulator, it takes a lot longer to see if we believe their story, and most algorithm providers unfortunately fall down at the first hurdle, because they don't understand what they're trying to sell."

Breakaway Speed
Indeed, lack of suitable options is the second reason boutiques are rising, and it’s a surprising one—both in terms of how consistently it is opined among buy-side futures specialists, and at whom it’s often directed: the major sell-side firms.

In part, TT’s Haldes says, bank algos are usually designed as part of a bundled package with clearing and execution, rather than as a “standalone” value proposition. They are comparatively inexpensive, and will do the job they are designed for—order slicing, calendar spreads, and other conventional strategies used to limit market impact—but in most cases, even where concerning top sell-side names like Credit Suisse AES, UBS and JPMorgan, they will do little more.

“One very large East Coast CTA client is now trading 100 percent of futures using algorithms, so no longer manually trading at all,” Haldes says. “They find FCMs will have pretty good products for some situations, but in other cases, the bank algos fall a little short, especially if you’re trading several contracts across correlated markets—soybean meal with soybean oil or other agricultural commodities, for example.”

Mundt agrees, citing speed of delivery as the issue. “The banks offer algos across a broad spectrum of products, but because of that, they tend to be fairly generic,” he says. “Most of those algos are ported from equities where paying an extra penny for crossing isn’t a big deal, whereas in rates futures that spread can be quite large. And banks are generally big companies that move slower due to greater bureaucracy. The development timeline is so long, every time they take one step, a small team like QB’s takes two steps. Six months later, you still might be working through the first incarnation of an idea; meanwhile, QB has produced four iterations of it already.”

That was the backdrop three years ago when AHL took a meeting with Almgren, Steel says. "We have a huge focus on research with Oxford Man Institute and the wider world of academic finance, so we knew Rob from there," he explains." Being a cynical trader, I was initially dismissive about the possibility of them besting what we have in-house, but Rob wasn't what I would view as a normal salesman, and using their simulation capability we identified which markets they could compete on in our opinion, really testing them, and we now use their algos across several fixed income markets. In some markets they're showing improved performance as compared to our algos; in others, our algos outperform. Managing our trading desks, I'm not looking to reinvent the wheel by developing every algo in house. We believe it is more efficient to use a blend of in-house and external algos, as long as we can identify which algorithm may be optimal for any given trade, and ultimately our clients."

Snap-Shot
So what is it about futures—especially rates contracts—that demands next-level simulation? For one thing, everything moves at once, while venues will introduce little subtleties like pro-rata versus pro-FIFO (first in, first out) allocation, too. As Almgren puts it, correlation is only the most obvious layer in the contracts’ “multidimensionality.”

“For example, equity futures are ‘front-month,’ whereas Eurodollar contracts have a whole spread across maturities that cause pricing relationships,” Almgren explains. “So even if you’re looking at one future, say a 15-month Eurodollar, you need to look at the entire complex when deploying the algo. And by contrast to equities, where key earnings numbers are released after market close, in our markets, employment, payroll numbers, inflation, and auctions across all the major countries drive the dynamics, and that information is released at various points throughout the trading day. A two-year note auction might affect the far-out Eurodollar more than in the short-term; a preliminary UK GDP number could hit the three-year mid-curve, but have a lesser effect on the fourth- or fifth-year. A goods and durables number will do something else—it completely varies by event.”

Being able to gauge those events’ effects on the market allows QB to develop algos that can really hunt for slippage, Revolution’s Mundt says. “In a word, they develop ‘opportunistic’ algos as opposed to the banks’, which are more ‘scheduled,’ in that the trades get parceled out at a certain rate and things generally stick to that schedule, either according to volume-weighted average price (VWAP) or percent of volume. But take a percent of volume like 7 percent—every hundred trades, we’re doing seven whether it’s a good time or not. QB has a much different set of constraints. Their algo can be passive for many minutes, and if it likes a good price it’ll get aggressive. The behavior is punctuated and more efficient.”

Skating Legs
Shaping that behavior first requires some serious vetting, with the same matching engine logic as the venues—CME, Liffe, and Eurex—taking in real-time market data that can then merge with orders from the algo. That is the genesis of simulation, and Almgren says QB’s environment is designed to pinpoint micro-pricing bands and short-term price signals, rather than liquidity or impact, because those are the important features for the beefier rates markets it focuses on. “It’s a sweet spot we were going for in trying to build this the right way, without making it too complicated for our users,” he says of the simulator. “Just interweave orders in with the market’s, and see how they perform.”

In 2014, its role has taken on new complexity, as well. AHL, for example, is eyeing more advanced algos provided by some of its agency brokers for new applications.  “Since the introduction of electronic trading the development of our execution algos has allowed traders to move up the value chain and focus on more complex markets and strategies. Currently, we are evaluating a new generation of algos designed to handle multi-legged and cross-asset strategies.”

And indeed, complex strategies like legging are where the two streams of QB’s information—post-trade total cost analysis and in-depth simulation—come together. Today, the agency broker takes on projects like constructing a 20-leg strategy that specifies the target price on a synthetic instrument using customized ratios across the curve, usually combining passive and aggressive legs at an expressed tolerance. QB also recently examined nuances around settlement price sensitivity for commodities indexing for one of its global asset manager clients. It’s hard work deep in the weeds, where others are disinclined to go, but worth the effort on two simple premises: avoiding being gamed, and coming through the trade in the money, with a little bit of alpha. “Dissection is our mindset,” as QB co-founder Christian Hauff puts it.

Averages, Not Chimeras
The future question, then, isn’t demand, but whether agency brokerage can be replicated elsewhere successfully. TT, Haldes says, is working on modifications to its platform to pull in a wider variety of independently brokered algos, as well as bank algos, set for early adoption later this year. That is no coincidence.

“It’s certainly becoming a more crowded space, with even the third-tier banks looking to say ‘me, too’,” he says, pointing out that some FCMs are white-labeling third-party algos just to get in the game. “At the same time, business on the sell side is operating more tightly today. That will push more expertise out into the middle, which could also come from the end-users with specialization in niche contracts like natural gas, too. We’re anticipating more boutiques emerging from the grassroots. It’s inevitable.”

Part of QB’s unintended benefit, Mundt says, is that they are raising the bar for everyone, and forcing the banks to engage in meaningful development work on algorithm improvement. “They are at least trying to play catch-up now, but this won’t necessarily erase QB’s value-add, as they will likely move on to the next area of improvement well before everyone else does.

And indeed, as QB pushes into other electronic asset classes like cash treasuries, Almgren says imitation is the best form of flattery, especially in a space where genuine debate—backed by years of advancing technology—is perhaps the most valuable commodity going. After all, what fun is it anticipating where the puck will be, if everyone else on the ice is hopelessly flat-footed?

“It’s a common conversation we have,” he says. “A colleague or client will propose a modification to the algo, thinking it will make a particular bad trade better and we’ll say, ‘Sure, but it might make the good trades worse.’ You have to test it; otherwise you can find yourself chasing chimeras. We’ll interrogate why we crossed as a micro-price moved up, for example—in one instance it was wrong, but we can point to 50 times on the same day where it was the right thing to do. There’s tremendous value in that. With execution, the average is what matters. You live and die by the statistics, and you only get there by having the details others don’t.”

Salient Points

• Managed futures specialists are increasingly taking advantage of boutique agency brokers’ algorithms, citing their ability to be opportunistic and adjust to markets’ behavior, as well as faster speed to implementation and greater alpha realized through price slippage.
• Rates futures, particularly, are ripe for these applications given their correlation and the characteristics of the complexes within which they’re traded, and are well-serviced by Quantitative Brokers (QB), among other independent shops. Hedge fund AHL and CTA Revolution Capital Management are among QB’s users for rates.
• Another value-added feature at smaller shops like QB is their simulation environments, which mimic the matching engine logic of relevant futures exchange venues and can test new adjustments to algorithms with real-time market data before putting the algos into production.
• Sources expect a greater variety of such brokers to crop up in coming years, while sell-side FCMs, sensing greater competition, are also expected to mature their offerings and continue bundling futures algos with other execution and clearing services.