Global Liquidity Focuses on Building Buy-side Adaptive Algorithms
Building algorithms is no longer is a hallmark of Wall Street as smaller firms are building them on Main Streets across the U.S. Now trading software is being built in the suburbs – 12 Broad Street, Red Bank , N.J., to be exact.
In a small town, more known for its chic and trendy clothing boutiques and restaurants sits Global Liquidity Partners, a broker-neutral execution broker that serves both the buy- and sell-side. GLP finds itself serving the firms the larger bulge bracket and even second tier brokers are either unable or unwilling to serve – building algorithms to help everyone achieve their best execution goals.
The firm, founded in 2010 by chief executive officer Tim Lang, touts itself as the answer to combat HFT and their strategies, offering the buy and sell-side a custom-made, low-latency, broker-neutral, co-located trading system designed at the outset to execute just as fast as HFT and eliminate their oft-talked about speed advantage. The firm took its time building out, opening for business in 2011.
And now it is hitting its stride, after launching with a handful of staff and a stable of plain vanilla algorithms, it now also writes its own algorithms, beginning with basic VWAP and TWAP tools – all written by Lang and his team- and is now moving into the next frontier of designing and building so-called “adaptive” algos that can be altered to adjust performance on-the-fly when prices or volume shift. As Lang told Markets Media in an interview, these bespoke algorithms are the future.
“Our buy-side clients have asked us for customized solutions when it comes to algorithms,” Lang began. “I believe the next frontier in algorithmic development are adaptive and dynamic trading strategies.”
So how does Lang and GLP write an adaptive algorithm?
It all starts with a client visit, Lang said. He goes and spends the day siting on a buy-side trading desk- watching and observing.
“When I am there I ask them about the behavior they desire on a particular order. We program behavior,” he explained. “The algorithm or trading strategy can and does change based on time and price shifts or time and volume shifts. If a stock moves favorably, the client may want to be more aggressive. As an order runs the algo now needs to shift. Early algorithms were static. Never changing.”
And this is GLP’s buy-side focus -sitting with existing or extant clients and learning their behaviors. And then programming them.
“This is really where our attention has turned,” Lang said.
After evaluating what a trading desk or trader needs, Lang and his team sit down and write the algorithm. After a recent trading desk visit, Lang sat at his desk and in about two hours, created a new algorithm. He then sent it to his team of engineers in Chicago who code the software.
In this instance we had the algorithm on the trader’s desk in about two weeks,” Lang said. “This was an adaptive custom built strategy.”
Typically, a bulge bracket firm can create an algorithm in about two to three months.
“We can do almost anything the buy-side asks of us. If it can be explained in math as behavior, we can build it and put it on the desk. We are very good at this.”
Lang describes himself as a quant and mathematician – but his pedigree is all Wall Street. With more than 25 years in trading, Lane has been a managing director at Spear Leeds & Kellogg, overseeing off-floor trading functions. He has also worked at Sherwood Securities where he was head of listed trading, building its trading desk and overseeing the internalization of order flow from national Discount Brokers. Most recently, he served at Credit Suisse’s Swiss American Securities subsidiary and was head of internalized trading, making markets, facilitated block trades and served to advise the bulge bracket on market structure.
GLP under Lang has invested heavily in technology, as the firm sees itself as a technology company that is focused on efficient execution regardless of cost – something the CEO said is what the buy-side says is paramount.
Aside from its custom algorithms, its trading platform is built on a C++ event engine, providing traders with super-high output and throughput of messaging and processing power. The event engine, designed in-house, enables users to manage their central order book, execute across various algorithms and monitor the functioning of its smart order router.
Second, it users what Lang terms “A Net Trading Model” that focuses on best execution and can be customized and adapted right in-house. This model, according to Lang, takes multiple variables into consideration – make/take rates, market center latency rates, trade velocity, market center fulfillment rates, average fill size, order execution time and client benchmarks. In case the net trading model isn’t for a client, an agency-model is also available via GLP.
And in order to protect itself the firm manages several redundant network and hardware co-located market centers to ensure seamless operation.
GLP sends market orders to all exchanges, dark pools, ECNs as IOC’s in a broker neutral fashion. Traders get the fastest data feeds – with real-time data coming directly from exchange providers raw data protocols – no third party providers.
By doing all these things, coupled with its seasoned staff, most of whom have more than 15 years experience, GLP looks to extend its buy-side reach even further.
“Innovation happens in small places,” Lang told Markets Media. “The scarce resource is not the money to get things done – it is the right people.”
More on Trading: