Read Backtest Results: Separate Returns, Risk and Trading

Your first task is to establish which research run produced the report. Check input identity and SHA, settings, period, software version, price basis, fees and observation frequency. Inspect exit status and logs to establish successful completion before proceeding to curves.

A matching filename or similar image does not establish matching research. In the reference course, different operations may bind different manifests despite selecting the same rows. Use each result's receipt and follow inputs through to outputs alongside the first-research preservation steps.

The equity curve shows the combined valuation of cash and holdings. By contrast, jQuantStats return analysis computes N−1 return intervals from ratios between adjacent equity observations. It excludes the interval from initial cash to the first observed equity value.

The source report's full-period return uses initial cash as its base, so its denominator differs from a summary based on the first equity observation. When figures differ, check periods and denominators first. Monthly and annual charts summarize observed windows, which may not cover complete calendar months or years.

Even when an equity curve shows substantial growth, inspect the underwater chart for declines from previous peaks. Maximum drawdown measures the deepest decline, while recovery time is separate information. See maximum drawdown and record depth and duration separately.

For rolling returns, volatility and Sharpe ratios, check the window and calculation frequency. Annualization with 252 periods applies only when the report explicitly declares daily sampling. Do not treat irregular observations or calendar days as trading sessions. Consult reading the Sharpe ratio as well.

Cumulative fees show how recorded charges accumulate. Do not subtract the same fees again from equity or trade P&L that already includes them. Record any omitted slippage or market impact as an excluded assumption.

The public dashboard's gross turnover includes both buying and selling. Comparing it under the same name as one-way turnover changes its meaning. Inspect cash and holding weights and the holding count, then use costs and turnover to keep units and denominators consistent.

An order is an instruction to trade, a fill records its execution, and a completed position cycle tracks a holding from opening to closing. Their counts are not interchangeable. Include skips when reconciling prior-day signals, next-observed-session opening prices, quantities, fees and cash movements against the price and execution chart.

FIFO fee-net closed-trade returns and realized P&L differ from the equity curve, which includes the valuation of positions still held. Do not explain the entire result using closed trades alone; inspect open holdings and cash. The public momentum example's 161 fills likewise do not mean 161 completed trades.

The public reference TOPIX is a price index excluding dividends. Do not call it a total-return index, and check exact alignment with strategy valuation dates. Do not fill missing dates with zero or a previous value to construct relative metrics. Different price bases or dividend treatment also constrain interpretation of the difference.

The reference regime chart is Off/unavailable with zero trace entries. Concentration and contribution charts likewise require supporting inputs. Do not interpret an unavailable panel as a favorable result; see the existing report-chart tutorial for chart types. Finish by recording returns, risk, trading and unresolved checks separately, together with the research-only status and pit_guarantee=false.

Source: the public GUI research courses. This article teaches research procedures; it does not recommend securities or trades or guarantee future returns. See the disclaimer.

Explore the risk and return learning path

• What Is the Sharpe Ratio? Why the Number Alone Is Not Enough
• CAGR: separate compound growth from the path of losses
• Volatility: align return frequency and standard-deviation conventions