### Genetic programming

Production of models is the process of searching for patterns which have a statistical probability of repetition, which gives rise to a mathematical expectation of profitability. This is also the process of "information extraction" from the market signal. The profitability of any model is a measure of its efficiency in performing these functions. Many of the models which make up the Quadrant portfolio have been developed by an automated search framework that implements a Genetic Programming methodology to find such profitable patterns. A full description of our implementation of this technology can be found here

This process allows us to run millions of backtests and find populations of fit chromosomes that are able to survive and profit in the market environment. These programs are tested across years of data and we employ a strict means of ensuring that these models are robust and able to survive across a range of market regimes and profit from combinations of market behaviour. This is a substantial collation of intellectual capital, and the fact that the intelligence required to generate profit from financial markets has been encapsulated into a collection of software agents, renders their scaled deployment a matter of the management of technology as a commodity resource.

### Portfolio Optimization

Through the application of a stochastic heuristic search algorithm, a variable weighting is applied to the assets allocated to each of the models according to defined risk constraints. This value is dynamically updated each week to select the top performing models for the portfolio. This creates an optimal portfolio effect which smoothes the returns generated by each individual model and also ensures that the models within the portfolio are suited to the current market environment.

The resulting performance metrics are superior to single model performance statistics.