Metrics reference

Every backtest reports a standard set of performance metrics, computed over the mark-to-market equity curve (ats/workers/metrics.py). The same metrics appear in the run detail view, the runs log, and the research ladder.

Annualization (why the numbers are comparable across timeframes)

Ratios like Sharpe scale with the square root of the number of periods per year, so a 1-minute strategy and a daily strategy must be annualized differently. ClearEdge picks the right periods_per_year automatically from the bar spec and asset class:

• Daily equities/ETFs: 252 trading days/year.
• Daily crypto: 365 days/year (crypto trades every day).
• Intraday: the daily figure × the number of bars per session (derived from the cadence).

So a Sharpe of 1.0 means the same thing whether the strategy trades daily or every minute.

Return & risk

Metric	Meaning
starting_cash	The capital the backtest began with.
final_equity	Mark-to-market equity at the end of the run.
total_return_pct	Total return over the whole period, final_equity / starting_cash − 1.
cagr_pct	Compound annual growth rate, annualized by calendar time between the first and last bar.
sharpe	Annualized Sharpe ratio (mean return / stdev of returns, × √periods_per_year).
sortino	Like Sharpe but penalizes only downside deviation (root-mean-square of negative returns).
calmar	CAGR ÷ max drawdown — return per unit of worst-case pain.
volatility_pct	Annualized standard deviation of returns, in percent.
max_drawdown_pct	The largest peak-to-trough decline of the equity curve.
exposure_pct	Share of bars with an open position — how much of the time capital was actually deployed.

Trade quality

Computed over realized (closed) round-trips.

Metric	Meaning
win_rate_pct	Share of closed trades that were profitable.
profit_factor	Gross profits ÷ gross losses. > 1 means the winners outweigh the losers.
avg_win_loss_ratio	Average win size ÷ average loss size.
expectancy_per_trade	Average realized P&L per closed trade.
avg_holding_days	Average time a position is held.
trades_per_month	Trade frequency.
total_fees	Total commissions paid over the run.

Benchmark (the bar that matters)

In addition to the above, a single-instrument backtest reports how the strategy did against simply buying and holding the instrument over the same window:

Metric	Meaning
benchmark_return_pct	The buy-and-hold return of the instrument over the period.
excess_return_pct	The strategy’s total_return_pct minus the benchmark — its return above buy-and-hold.
orders	The number of orders the strategy submitted.

A strategy that doesn’t beat buy-and-hold isn’t earning its complexity — this is judged on every run and is a first-class output, not an afterthought. See the gauntlet.

Where else metrics appear

• The runs log shows the headline metrics per run.
• The research ladder (/api/research/confidence) rolls the gauntlet stages into a 0–100 readiness score — see the research ladder.
• The research copilot reads a compact summary of these metrics so it can discuss your results; see Studio.