Trading Behaviour
The Algoteq Algorithm Suite is based on a multi-tier model, where low-level trading tactics capture focussed, independent aspects of trading, and higher level strategies coordinate multiple tactics to achieve overall trading goals. A particular coherent combination of such components comprises an algorithm.
This approach supports abstracting both human and machine trading actions as clearly coded and individually testable units of trading, that are easily combined into both standard and bespoke algorithms. Behaviours common across multiple strategies, such as auction trading logic, opportunistic trading, access to dark liquidity etc, are maintained independently of any particular strategy or algorithm and can be integrated as appropriate.
Multi-tier interoperability of the trading logic components is possible because their output is produced in a generalised, market-independent form, that also enables a uniform approach to trading across multiple asset classes, integrating quantitative models and signals, optimising output instructions, anti-gaming, alpha-generation and smart routing.
To minimise complexity and better align the algorithmic trading to client needs, two high level parameters (named urgency and sensitivity) are provided for all algorithms, and used to control multiple internal parameters across the strategies and tactics. In broad terms, urgency specifies how happy the algorithm to engage the current market, and sensitivity how much is the algorithm influenced by the possible future market state. There are also multiple limits for safety and control.
Data and Analysis
Fully integrated into the Algorithm Suite, and closely coupled to the trading logic, is an independently maintained and managed reference collection of derived historical data, covering all significant features of each stock’s intraday and overall trading characteristics. This is used to provide the algorithms with information relating to their given stock, and guides appropriate parameter setting and default behaviour. It is also an important reference for intraday assessment of price movements and other order book dynamics, informing the adaptive trading behaviour and parameterisation aspects of the algorithms.
Market data is provided realtime, but many calculations and enhancements are made and provided to the algorithms (and to the user) via analytics. This includes additional fields, moving averages etc designed to improve the performance of the trading logic. Additionally, custom aggregation and synthetic venue support are integral to the implementation of smart routing.
Available Algorithms
The many included algorithms in the Algoteq Suite can be understood as belonging to one of a small number of general types: profile, passive, responsive, and liquidity seeking.
The profile algorithms are designed to achieve full order execution over a fixed period of time, with minimum market impact and targeting a time averaged benchmark price. The general method of execution is to divide the order into distinct (but overlapping) waves and execute each in accordance with a predetermined, but adaptive schedule – called the profile. The success of this approach depends on two factors: how well the schedule reflects trading volume with respect to the target benchmark, and how effectively is each wave executes in its prevailing market. The user is able to influence these factors through their particular choice of profile algorithm, and choice of parameters. Adaptability in the profiles is through a tilt mechanism, where, depending on achieved execution and market conditions, the profile can be skewed about its initial shape.
The Algoteq Algorithm Suite includes the following profile algorithms:
- VWAP, for targeting the volume weighted average price and trading a profile calculated from the historical intra-day volume distribution,
- TWAP, for a consistent rate of execution throughout the order period,
- IS, for trading a dynamically adjusting profile, based on a comparison of the current and predicted market price vs a specific benchmark price (such as arrival, previous close etc),
- POV, for trading a profile that dynamically adjusts itself based on current and predicted market volume,
- MC, when the price benchmark is the future close price, the profile adapts in order to trade as much as possible in the close, while minimising adverse impact. Trading pre-close is supported when required due to a large order quantity.
Passive algorithms generally deliver excellent performance against a VWAP benchmark by avoiding spread costs, but, because of this, the degree to which they will complete can not be guaranteed. The Algoteq Algorithm Suite includes PEG, for minimising market impact by carefully managing the passive quantity on market and response to price movements, and ICEBERG, for carefully controlling order visibility.
The responsive algorithms are simple utility algorithms, more akin to smart order types, and are algorithms that will only trade in response to a particular market event to occur. The standard suite includes SNIPER, TOUCH, STOP and LIMIT-TO-MARKET.
Finally, liquidity seeking algorithms are a more general, bespoke class of algorithm, that are created by combining and adapting passive and responsive aggressive tactics with custom analytics and dynamic take profit or stop loss behaviours, usually with particular reference to hidden liquidity, to produce sophisticated and focussed bespoke trading behaviour. These kind of algorithms often have an algorithm of algorithms structure, and generally have a particularly close relationship with advanced SOR capabilities.
Simulation
A particular strength of the Algoteq Algorithm Suite is the ability to simulate production and experimental algorithms, without alteration, against historical market data. Our tools enable developers, quants and traders to visualise and analyse the trading behaviour (in real time and post-trade, with aggregate and market specific views), and also interact with the algorithm and explore alternative choices and parameterisation. Additionally, the results of simultaneous runs of tens of thousands of simulated orders, using specialised, highly parallel hardware and associated software control, form the basis of our parameter response models, and trading logic refinement.
Summary
The Algoteq Algorithm Suite offers a focussed and customisable algorithmic framework, with clear and powerful abstractions of trading behaviour, multiple and varied algorithm implementations, stock specific time-dependent reference data, and associated supporting tools for simulation, analysis and visualisation. The Suite is the embodiment of a process that explicitly supports the cycle of development, evolution, visualisation and optimisation, and thus guarantees the on-going suitability and extension of its automated trading capability, across asset classes and even in the face of rapidly changing markets and client requirements.