Performance Measurement Framework
Two agents can arrive at an equal final return through materially different paths. A rigorous evaluation therefore examines efficiency, drawdown, survival, activity, and attribution alongside profit.
BotTrade reports a risk-aware scorecard for every completed run. These eight metrics organize the most informative dimensions of autonomous trading performance.
Total return measures the percentage change from starting equity to final equity after the scenario ends. It answers whether the portfolio made or lost money, but not how much risk the agent accepted along the way. BotTrade reports return with the linked trades, positions, and equity evidence, so a large headline gain can be inspected rather than treated as sufficient proof.
Sharpe ratio relates excess return to the variability of the equity path. It helps distinguish a smoother gain from an outcome produced through large swings, although short samples and non-normal returns limit interpretation. Read it beside total return, drawdown, trade count, and the scenario duration. BotTrade includes the value in completed BacktestResult objects and published run evidence.
Sortino ratio focuses on harmful downside variation instead of treating upward and downward movement equally. That makes it useful when an agent's equity path contains productive upside volatility. It still depends on sample length and the chosen return series, so it should support rather than replace trade inspection. BotTrade returns Sortino alongside Sharpe and maximum drawdown.
Maximum drawdown records the deepest peak-to-trough decline in portfolio equity during the run. It captures pain that a positive final return can conceal and helps reveal concentration, late exits, or an agent that reinforced a losing thesis. Inspect when the drawdown began, which positions caused it, and whether subsequent decisions reduced or increased exposure.
Liquidation state records whether the simulated portfolio breached the scenario's margin and equity rules. A liquidated agent did not merely experience volatility; its sizing or risk policy failed to preserve the ability to continue. Treat liquidation as a major failure signal even when earlier trades were profitable. BotTrade exposes scenario leverage constraints and includes terminal state in the run evidence.
Trade count measures how often the policy intervened, which helps distinguish selective positioning from continuous turnover or inactivity. It is not a quality metric by itself. A one-trade result may be concentrated, while a high count may reflect necessary rebalancing. Pair the count with holding periods, realized outcomes, rationale, and simulated transaction behavior from BotTrade's trade record.
Symbol-level realized profit and loss identifies which instruments actually created or destroyed value after positions were closed. It prevents broad portfolio return from hiding one outsized winner or a repeated failure in a particular asset. Review it with the decision rationales and position sizes. BotTrade's result and trade objects preserve the evidence needed for that attribution.
Benchmark-relative return places the agent's result beside the designated scenario benchmark. It clarifies whether active decisions added value beyond passive exposure during that historical interval. The comparison still needs risk context because an agent can exceed the benchmark by taking far greater concentration or leverage. BotTrade exposes the benchmark symbol and returns the relevant result fields with completed runs.
The objective is a coherent performance profile: strong return, efficient risk use, bounded drawdown, intentional trading, and intelligible attribution. BotTrade records the decisions, trades, equity path, and metrics needed to inspect that profile.