Why screenshots are not a track record

Screenshots are not reliable track records because they are images of selected information rather than verifiable records. They do not securely link results to a specific identity, cannot prove completeness of the trade history, do not provide trustworthy timing, and can hide selection bias. They also cannot support metrics such as the deflated Sharpe ratio, which require the full return series and multiple testing context.

A track record is only useful if another person can check what it represents. A screenshot does not provide that checkability. It is an image of selected information, not a verifiable record.

Identity is the first problem. A reliable track record needs to be tied to a specific account, process, or research artifact in a way that can be inspected later. Sonar’s research to publishing workflow is built around preserving lineage from research outputs into published material so that results remain connected to their source context rather than detached as isolated claims. A screenshot breaks that chain. It does not establish who controlled the account, whether the image came from the claimed system, or whether it corresponds to the stated strategy at all. The image shows pixels, but not a verifiable identity link.

Completeness is the second problem. A screenshot cannot prove that it contains the full set of trades, orders, positions, or returns. It only shows the slice that was captured. By contrast, methods that evaluate strategy validity require broad access to the underlying record. Sonar’s backtest overfitting audit is designed to examine whether reported backtest performance is likely to be inflated by selection effects. That kind of audit depends on the underlying backtest evidence and its variants, not on a cropped image of a favorable result. If the full history is missing, there is no way to determine whether the screenshot excludes losing periods, omitted instruments, abandoned parameter choices, or other relevant observations.

Timing is the third problem. A screenshot may display a date or time, but an image alone does not make that timestamp trustworthy. It does not prove when the screenshot was taken, whether the displayed clock was accurate, whether the trades occurred in the shown order, or whether the image was captured after later edits or filtering. A reliable record needs time information that can be validated against the underlying data source. A screenshot does not provide that validation path.

Selection bias is the fourth problem. A screenshot can be chosen precisely because it looks persuasive. That is the mechanism of the bias. The viewer is shown one handpicked outcome without access to the full set of attempts, discarded variants, or unfavorable intervals. Sonar’s backtest overfitting audit exists to test whether apparent performance may reflect selection over many trials rather than robust signal. An image cannot support that test. It conceals the search process that produced the displayed result, so it cannot rule out cherry picking.

This matters especially for quantitative evaluation metrics that require the complete return path. Sonar’s glossary defines the deflated Sharpe ratio as a Sharpe ratio adjusted for non normal returns and multiple testing. That adjustment requires the return series and the context of how many trials were effectively considered. A screenshot does not contain a complete return series, and it does not reveal the multiple testing context. As a result, a deflated Sharpe ratio cannot be derived from a screenshot alone.

The basic issue is simple. A screenshot is a presentation artifact, not an audit artifact. It may illustrate a claim, but it cannot verify the claim. Without verifiable identity, complete history, trustworthy timing, and a way to detect selection bias, a screenshot cannot serve as a reliable track record.

Drafted with AI assistance from cited sources. Reviewed and approved by Sonar Sciences Quant & Research Team.