Product Overview

Adaptive Modeler creates agent-based models for generating price forecasts and trading signals for real world securities such as stocks, ETFs, forex currencies or commodities. Model creation and evolution is based on historical quotes and user specified parameters. Custom quote intervals are supported ranging from 1 second to daily or longer, limited only by the available quote data and processing speed.

Model creation and evolution

Models are initialized automatically based on the user’s parameters and consist of a population of agents and a Virtual Market. Each agent represents a virtual investor with its own assets and trading strategy. After initialization, the model starts evolving based on the historical quotes of the security.

After every imported quote, all agents evaluate their trading rule and can place a buy or sell order on the Virtual Market. The clearing price is then calculated and all executable orders are executed. The clearing price is taken as the bar-ahead forecast and if necessary a new trading signal is given. Trading signals are based on the forecast and the user’s trading preferences. Agents with poor performance are being replaced by new agents whose trading rules are created through crossover and mutation of trading rules of agents with good performance.

Model evolution never ends. When all historical quotes have been processed, the model waits for new quotes to become available and then evolves further. The model thus evolves in parallel with the real world market and every historical price is used only once for "testing" the trading rules (as in the real world and without the risk of overfitting historical data).

Real-time model visualization

Adaptive Modeler provides an extensive set of output data and visualization tools including live charts to observe the evolution and behavior of the model over time and the quality of previous forecasts and trading signals. For instance, the Forecast Directional Accuracy (FDA) is an indicator that counts the percentage of bars for which the forecasted price change was in the right direction.  

To get a deeper understanding of what is happening inside the agent-based model, various histograms are available showing distributions of agent values such as their wealth, returns, position, age and others.

Additionally, the agent population can be visualized multi-dimensionally in scatter plots of multiple agent values to identify relationships between different values and to gain further insight into the particular dynamics of a model. All these visualization tools are updated in real-time.

Simulating trading

Adaptive Modeler contains a Trading Simulator to simulate trading based on the trading signals. This makes it possible to see what returns would have been made when the suggested trades were actually executed. Trading can be simulated according to user customizable trading parameters such as enabling/disabling short positions and expected spread and slippage.

Performance analysis

Extensive performance analysis is available to study risk and return measures of the Trading Simulator such as Value at Risk, Alpha, Sharpe ratio and risk-adjusted return. Performance can be calculated per year, quarter, month or shorter periods and individual period returns and statistics are reported as well. Also it is possible to project the likely range of future trading returns using Historical Simulation or Monte Carlo Simulation. 

The table below shows the performance a model of the S&P500 index from 1950 using daily quotes. (This is the sample model that can be downloaded from the download page).

Contact Us
© 2003 - 2008 Altreva. All rights reserved.

Hosted by Bluehost.com