PortfolioBacktest
Helpers & Utilities · AiDotNet.Finance.Trading.Evaluation
Runs a policy through a portfolio environment deterministically and scores it. A “policy” is just
state → action, so this evaluates a scripted baseline and a trained RL agent
(s => agent.SelectAction(s, training:false)) the same way — the honest-eval seam for the greenfield
trading-agent research: hold out a chronological tail, run each candidate, and compare on these metrics.
Methods
| Method | Summary |
|---|---|
PortfolioBacktestResult Run(PortfolioManagerEnvironment<T> environment, Func<Vector<T>, Vector<T>> policy) |
Runs policy through environment for one full episode and returns its risk/cost-adjusted performance. |