Charting Backtest Results Honestly

Two researchers chart the identical backtest of the identical signal. One's chart shows a smooth, confident climb to strong returns. The other's shows a bumpy, unremarkable line with a visible multi-year flat stretch in the middle. Nothing about the underlying numbers differs. What differs is the y-axis scale, the start date, and whether costs are included before or after the chart is drawn. A chart is not a neutral rendering of a result — it's an argument, and the same honest data supports very different-looking arguments depending on choices most readers never think to ask about.

The choices that quietly change the story

Linear versus log scale on cumulative returns. A strategy that compounds steadily looks like it's accelerating out of control on a linear axis and looks like a straight line — its true, constant rate of growth — on a log axis. Linear scale flatters a long backtest and makes early, smaller-dollar periods look uneventful by comparison; log scale is the honest choice whenever the chart spans a period where the strategy's size, or the market's level, changed meaningfully.

Where the chart starts. Starting a cumulative return chart right after a bad stretch, rather than including it, is the single most common way a chart quietly overstates a strategy's history. The fix is mechanical: the chart's start date should be the start of the available backtest, full stop, never chosen after looking at where the line happens to look best from.

Gross versus net of costs, and whether that's labeled. A gross-returns chart next to a costs table in a footnote reads, to most viewers, as a net-returns chart — nobody reads footnotes at the speed they read a chart. If costs are material, the chart itself should show both lines, or show net only, never gross alone with costs disclosed elsewhere.

Drawdown, shown or omitted. A cumulative-return line alone hides how painful the ride was to hold. A separate drawdown panel underneath — the running peak-to-trough decline — is what actually tells a reader whether they could have sat through the strategy live, and its absence from a presentation is itself a signal worth noticing.

Smoothing. A weekly or monthly resample of a noisier daily series can be a legitimate way to make a genuine trend visible through short-term noise — or a way to make a genuinely mediocre signal look smoother and more trustworthy than it is. The test is whether the same smoothing would have been applied if the raw series had looked worse.

A comparison that makes the point

Choice	Flattering version	Honest default
Scale	Linear	Log
Start date	After the worst early stretch	Full available history
Costs	Gross, footnoted	Net, shown on the chart itself
Drawdown	Omitted	Shown in a panel below
Resampling	Monthly (smooths noise)	Match the trading frequency

None of the "flattering" column choices is a lie on its own — each is defensible in isolation, with a plausible reason attached. Stacked together, they can turn a mediocre, choppy, cost-eaten signal into a chart that looks like a clean, ascending line. The tell is usually that no single choice looks unreasonable when asked about individually; the manipulation lives in the combination.