ASTRIVERA Research is independently developed quantitative research infrastructure combining strategy research, historical backtesting, statistical modelling, machine learning, automated experimentation and research asset management in a professional working environment.
At its core is an evolving library of strategy assets based on distinct economic mechanisms, trading rules and research methods. It maintains strategy definitions, parameter configurations, research records and historical versions. Strategies can be studied individually or evaluated together under explicit combination rules, allowing findings to accumulate as searchable, reproducible and extensible research assets.
The research and backtesting architecture brings strategy configuration, experiment execution and analysis into a single workbench. Researchers can choose markets, study periods, strategy parameters, trading costs and comparison benchmarks directly, then run historical backtests, parameter optimisation, strategy comparisons and portfolio studies. Existing strategies can be rerun through standard workflows without manual edits to underlying code for routine experiments.
The platform supports statistical modelling, time-series analysis, feature engineering and the training and evaluation of machine-learning models. Predictive, signal, risk and portfolio models are managed within a consistent experimentation framework. Feature selection, parameter search, model comparison and automated training are used to investigate strategy behaviour and model suitability across data characteristics, trading mechanisms and market conditions.
Research evaluation considers statistical validity alongside real economic outcomes. In-sample and out-of-sample separation, rolling validation, robustness tests, trading costs, liquidity and strategy capacity contribute to systematic assessment. The research workbench displays equity curves, returns, drawdown, risk-adjusted performance, turnover, risk exposures and benchmark comparisons, with interactive analysis across strategies, parameters and models.
Automated research extends the scope of strategy discovery. It organises strategy exploration, candidate generation, experiment design, initial screening and deeper validation around defined research objectives. Configurable study tasks and model experiments combine broader exploration with systematic evaluation, improving research throughput and the ability to identify promising methods.
Research outputs are managed as enduring assets. Data versions, strategy definitions, model configurations, parameter choices, experiment processes and evaluation results remain traceably connected. Positive, negative and unresolved results are retained with their applicability, supporting evidence and historical conclusions for later reproduction, comparison and refinement.
Validated strategies and models can be packaged as candidates with defined versions, runtime dependencies and verification records, then submitted through a controlled handoff for independent assessment and application within the trading platform.
Cross-platform research environment. Supports macOS, Windows and Linux through an independent, portable architecture for local and private deployment, supporting strategy backtesting, machine-learning model training, experiment management and research asset maintenance.
Core Capabilities
- Source Strategy Assets and Research Management
- Interactive Backtesting and Performance Analysis
- Statistical and Time-Series Research
- Machine Learning and Model Training
- Feature Engineering and Factor Research
- Automated Strategy Discovery and Experiment Management
- Parameter Search and Model Optimisation
- Multi-Strategy Portfolio Analysis
- Out-of-Sample Testing and Robustness Evaluation
- Research Lineage, Versioning and Candidate Handoff