Data quality guard trading and DQG backtesting searches reflect a hard truth: backtest data quality trading problems create phantom edges that disappear in production.

What DQG-style checks ask

Are bars continuous where your indicators assume continuity?
Are timestamps aligned to the exchange session rules you think you use?
Are gaps handled honestly, not hidden by silent forward-fill abuse?

Why DQG matters before robustness metrics

If inputs are wrong, walk-forward efficiency and retention measure fiction. Fix price integrity check backtest issues before debating Sharpe (Look-ahead bias).

Practical habits

Log gap detection and refuse to optimize across large unknowns
Keep data revision in your lab notes (Why WFA differs each run)

Exchange microstructure and bar construction

Even "clean" OHLCV can mislead if your strategy depends on microstructure you do not model. Stale prints, auction halts, and session boundaries interact with indicators in ways a simple gap counter will not catch. Document your bar source and any resampling rules.

Corporate actions and symbol continuity

On equities, splits and dividends change price series. On crypto perpetuals, funding and index references matter for PnL. If your series stitches multiple contracts or symbols, state the mapping explicitly.

DQG as a gate, not a vibe check

Treat data checks like unit tests: they should fail loudly. Silent repair (heavy interpolation, aggressive fills) can salvage a chart while destroying statistical meaning. When in doubt, shrink the sample to a known-good window rather than "fixing" unknowns.