Backtesting

What This Page Is For

Use this page to configure backtests, run queued jobs, and read the results without needing to inspect the database.

Backtest Settings

Open Backtesting > Backtest Settings before running new tests. Review:

• cash and commission assumptions;
• optimization metric;
• Buy & Hold benchmark mode;
• market-phase filters;
• Strategy Quality Score weights;
• approval rules and grade scale.

These settings affect future backtest runs and the approval status used by trading workflows.

How Backtesting Works

BEC runs strategy tests against historical OHLCV candles for each selected symbol, timeframe, and strategy.

Unless Use full history is enabled, the configured timeframe lookback and warm-up candles determine the requested range. Full-history runs start from the exchange's available market history; for OKX, BEC automatically clamps the request to the instrument listing time and paginates all available candles.

The backtest uses the current settings for:

• starting cash and commission;
• stop-loss, ATR trailing stop, and take-profit rules;
• market-phase filters;
• strategy parameters;
• approval rules.

After each run, BEC stores the result in the database and can generate an HTML report. The dashboard then uses those stored results to approve or reject candidates for trading.

Backtesting is queued. When you select rows in Backtesting Results and start a run, the dashboard adds jobs to the queue and the background jobs_runner executes them one by one. Use the queue panel to check progress and logs.

After a queued backtest succeeds, BEC automatically creates an inactive Positions candidate or refreshes its matching position row when all live-trading gates pass. The strategy must be approved and selected in Main Strategies, the timeframe schedule must be enabled, the symbol must still be a current top performer, and the current exchange-scoped backtest must have a positive return and pass the configured approval rules. The job log records whether the candidate was inserted, updated, or skipped.

This makes the candidate available to the next live cycle for its timeframe. It does not place an order immediately: entry signals, position limits, and all other live risk controls still apply.

Backtesting Results

Open Backtesting > Backtesting Results to filter, select, and queue backtests.

Use Load Top Performers to populate the symbol filter from the current market-phase ranking. Select the approved strategy and the enabled 1h, 4h, or 1d timeframes, then use Run Selected Backtests to queue the work.

Use the grid to compare:

• return and drawdown;
• trade count and win rate;
• profit factor, expectancy, SQN, and Kelly criterion;
• Strategy Quality Score and grade;
• trading approval status and rejection reasons.

Strategy Quality Score

The Strategy Quality Score is a 0-100 score designed to avoid ranking strategies by return alone.

It combines five components:

Component	Default Weight	What It Rewards
Return	20%	Total return, annual return, and performance versus Buy & Hold
Risk	25%	Lower drawdown, shorter drawdown duration, and better Calmar ratio
Risk-adjusted	20%	Better Sharpe and Sortino ratios
Trade quality	20%	Profit factor, expectancy, SQN, win rate, and a reasonable trade count
Robustness	15%	Healthy trade count, balanced exposure, low commission drag, less dependence on one winning trade, and controlled drawdown

You can change these weights in Backtesting > Backtest Settings > Strategy Quality Score. The weights must add up to 100%.

Penalties

After the weighted score is calculated, BEC can subtract penalties. Penalties are capped at 30 points.

Common penalty reasons include:

• too few trades;
• drawdown above the expected range;
• too little or too much market exposure;
• high commission drag;
• too much dependency on a single winning trade;
• weak performance versus Buy & Hold combined with high drawdown.

The final Quality Score is:

weighted component score - penalties

The final score is always kept between 0 and 100.

Quality Grade

The grade is a simpler label based on the final Quality Score:

Grade	Quality Score	Meaning
A	>= 85	Excellent
B	70 - 84.99	Strong
C	55 - 69.99	Acceptable / moderate
D	40 - 54.99	Weak
F	< 40	Rejectable

Approval rules can use the grade. For example, a default Quality_Grade_Min of C keeps Grade A, B, and C backtests and rejects D and F.

Backtest Reports

Select one row to render the HTML report. Reports include charts, statistics, configuration, trades, and risk/exit information.

The project also publishes an example report:

Open the example backtesting report

AI Strategy Analysis

AI analysis is available when OPENAI_API_KEY is configured. Use it as a review aid, not as a trading signal. It summarizes strengths, risks, recommended tests, and data quality notes.