GENERAL OVERVIEW:
What is the FX algo called?
Basket Algo, Goldman Sachs’ unique portfolio algo, for intelligent simultaneous execution of multiple currency pairs
What category does it fall into?
A very exclusive category (in FX) of portfolio algos, and the first algo of its kind industry-wide. Basket algos are common-place in equities, but surprisingly exclusive in FX.
What does it attempt to do?
The basket algo attempts to achieve three objectives. 1) It synchronizes execution up to a certain tolerance, to ensure that at any given time, a roughly equal percentage of each leg of the basket is complete. 2) It seeks netting opportunities both directly and synthetically. Netting takes place without incurring algo charges. 3) It controls the overall slippage of the basket, by measuring performance versus an initial weighted average mid. The entire basket pauses execution if the slippage target is breached.
STRUCTURE:
What is the algo’s software architecture?
Goldman Sachs employs highly custom low-latency components, globally distributed in colocation facilities, to give clients the highest chance of optimal execution.
Does it use proprietary modelling?
Yes! The proprietary modelling features in various aspects from the Dynamic Hybrid flagship technology that underpins the execution of each leg of the Basket algo to the mechanisms controlling the simultaneous execution of the legs.
Does it use technology such as AI or ML? If so, how?
No. We are firm believers in the reproducibility of results, so while there is a self-learning logic in Dynamic Hybrid, this is a deterministic real-time reaction to changing market conditions.The logic governing the “grandparent” basket level structure (if we think of the basket legs as “parent” level and the individual executions as “child orders”) is equally rules-based.
FUNCTIONAL ASPECTS:
Does the algo adapt automatically to prevailing market conditions and if so how?
Yes, it does. The so-called adaptive dynamics in Dynamic Hybrid ensure that each leg of the basket is sniffing out any changes in market spreads, opportunities for price-taking near mid, market icebergs etc. Adaptive dynamcs considerably reduces the need to supply the algo with any hard-coded parameters. The basket algo as a whole is then much like the conductor of an orchestra, ensuring a harmonious whole, but relying on the execution quality of each individual constituent.
Does it incorporate smart order routing?
Obviously!! FX is highly fragmented, so smart order routing is the name of the game. We go one step further and create bespoke pools of liquidity that our flagship Dynamic Hybrid and therefore also the Basket are able to tap into.
How does it minimise market footprint?
Much of this is the secret sauce to our recipe, but a few pointers are: intelligent internalisation, which means not only finding existing matches, but also generating matches in a skew-safe environment, and spending a lot of time staring at markouts by external venue. Often we discover that less is more when it comes to choice of venues externally. The basket algo is a special case, in that very few market participants attempt to measure the effect of correlation between currencies on their trades. So with traditional means of execution, a trade in one pair could be creating a footprint in another. Trading these together and managing the joint performance could be beneficial.
What liquidity seeking and access capabilities does it deploy?
Seek liquidity and you shall find it. Even when markets look quiet, like a primary market in an Asian NDF pair in the NY afternoon, there are then other liquidity sources that come to the rescue. The basket algo is at any given time seeking liquidity in multiple currency pairs, a dozen or more at times, while carefully netting any internal matches and observing all the client-specified rules on limit prices on each leg and overall slippage.
If one leg of a basket is much less liquid than others, then it would usually tend to slow the basket as a whole down, but our advanced features ensure that users may specifiy one or more basket legs to trail the others, or even to be completely detached from the synchronization rule, while still observing the slippage requirement and netting.
What operational risk management does it include?
Within its native Marquee environment, the basket algo’s operational risks are quite tightly controlled. For example, it’s possible to save down commonly used baskets, and then simply re-load them whnn needed, automatically scaling the constituents to a new total notional.
PARAMETERS & CONTROLS:
What client inputs are available in the algo?
A full laundry list would be dull, but the minimum input requirement is direction, size, settle date for each pair. A default basket has the slippage target set wide. This can be adjusted if needed. The coolest advanced options include the leg flavour modes discussed above in situations where clients have combined liquid pairs with illiquid EM or NDF pairs, and wish to have greater control on the interaction between the legs.
How much real-time feedback does it provide?
Each leg has its own real-time order monitor to observe progress, performance versus that leg’s risk transfer and other benchmarks and real-time updates on the venues accessed. Aggregate performance (e.g. live basket slippage) can be seen on the order ticket.
CAPABILITIES & USE:
What execution styles (e.g. passive/aggressive) does the algo support?
The legs are executed in a hybrid (i.e. opportunistic) style, and hence can be set to slower/normal/faster modes. They are part passive and part aggressive.
How can it be integrated/called with/by higher-level workflows?
The GS basket algo is now available via FIX, so that it is integrable within systematic or semi-automated execution strategies.
What is the optimal scenario for its use?
Trades that are suitable for a basket algo include portfolio rebalancing, execution of DXY, SGD-basket or other currency baskets. Other applications include non-nettable executions on the client side, from simply buying and selling the same currency pair to having to trade both the legs and the synthetic cross.
More generally though, the only time a basket algo is probably an unnecessary choice is if the constituents are totally uncorrelated!
Any other functionality worthy of note?
The Goldman Sachs FX basket algo is an extremely complex beast, without doubt one of the most technologically advanced algos in the industry.
Its ability to opportunistically source liquidity across multiple pairs while controlling aggregate performance is inconceivable in traditional voice execution. It is one of many reasons that clients choose Goldman Sachs’ Marquee platform for their execution needs.
