Our Research Projects

What it does:

Our solutions for MiFID-compliant best execution in a distributed executing environment is based on broker software agents, which cooperate to split and route deal requests automatically to the best offering with lowest possible latency. Market maker software agents monitor and assess the behavior of their respective integration partners/liquidity providers and learn how they behave at deal requests.

Background information:

Quantsware developers took on this extensive project because institutional users require electronic integration with their executing dealers.

What it does:

Stigmergy is method of communication in which individual members of a group communicate with each other by modifying their local environment. In many insect species, this takes the form of the production of pheromone trails to indicate the location of important sites, for example, food sources. Due to the dynamic effects of having many distributed members combined with the transience of the interaction with the environment, optimization problems may be solved, such as locating the shortest path to a food source. This represents a form of swarm intelligence. This project aims to simulate this intelligence by using mobile robots capable of detecting volatile liquid (e.g. methanol, ethanol) trails laid down by fellow members of the colony.

Background information:

When we talk about machine learning, we talk about optimization. Swarm intelligence may be used retrieving optimal trading strategies. In the more abstract case of retrieving optimal trading strategies, the liquid trail may be replaced by numerical values.

What it does:

A network of multi-agents model the distributed Forex market and mimic the behavior of diverse market participants (banks, speculators, international trading companies, and central banks). The project objective was to model an understanding of market microstructures and changes to these, e.g. introduction of transaction tax. 

Background Info:

Foreign exchange is being traded in distributed environments rather than on central exchanges. While such infrastructure has created efficiencies, it has also led to unprecedented challenges.  The credit crunch 2008/9 has demonstrated that – thanks to increasingly networked markets – events in one marketplace, such as the subprime mortgage market in the USA, can have global ramifications. Consequently, an increasingly urgent question was whether and how such markets should be regulated to prevent or contain such crises.

What it does:

One research project involved settlement optimization and auto collateralization, i.e. algorithms that would reduce settlement costs by minimizing the need for cash and securities to conduct actual transactions. Appropriate algorithms identified transaction chains and idle times and link transactions in a way that only the delta after netting transactions would have to be settled. Auto-collateralization was understood as an intraday credit operation in central bank money for facilitating settlements of transactions. 

Background Info:

This project had the objective of designing a real-time settlement optimization concept for a central bank. While settlement optimization algorithms that are based on overnight batch runs were already in use on national levels, little was known about the true efficiency of these partially proprietary algorithms in a cross-border setting. In this context it was required to demonstrate what multi-agent systems could potentially contribute to a superior real-time solution.