The diagram below depicts the development lifecycle of a typical trading strategy.
A hypothesis is formulated based on a number of possible sources, including journals, financial and economic literature, market observation and personal experience. Data mining can also be used to generate hypotheses post-facto, although this approach is usually frowned upon.
A prototype is developed and tested. Initial back-tests are performed over the in-sample period. The resulting statistics are analyzed and the hypothesis is either validated or rejected. Approximately 3 of every 4 initial hypotheses are discarded at this stage.
Development and Evaluation Phase
The system is tested out-of-sample to check for robustness. Parameters and holding periods are optimized in search of peak edge manifestation. The strategy is then subjected to internal peer review to obtain an objective assessment. Approximately two out of three systems are discarded at this stage.
Deployment and Maintenance Phase
The strategy is graded based on its performance statistics, trade count and equity curve. Position-sizing is determined and a walk-forward is performed. The strategy is reviewed on a quarterly basis and is upgraded, downgraded or retired depending on its continued robustness.