Strategy Robustness Lab — Rawstocks Tools

01 · Import

Load the realized trade history.

TradingView Strategy Tester exports are detected automatically. Generic CSVs can be mapped manually. The lab needs realized P/L; dates and direction unlock the strongest robustness tests.

Trade-history CSV

Drop a CSV hereor

Choose CSVNothing is uploaded. Parsing and calculations run locally in this browser tab.

Column mapping

Auto-detection handles common TradingView and broker exports. Adjust these fields only if the detected columns are wrong.

Realized P/L

Date / time

Direction

Row typeTradingView: Exit rows only.

Minimum slice trades

Rolling window size

02 · Sample

Start with the full record.

The full sample is the reference point. Robustness asks how often meaningful sub-samples retain a positive edge and how much their expectancy varies around that reference.

Trades—

Net P/L—

Win rate—

Expectancy—

Profit factor—

Date span—

03 · Robustness score

Consistency, not headline profit.

The score rewards profitable independent slices, stable rolling expectancy and low concentration. It is intentionally not a prediction of future returns or a replacement for out-of-sample testing.

Robustness score

—

Profitable time slices—Years + chronological halves.

Profitable rolling windows—Windows meeting the minimum-trade rule.

Positive random subsamples—Deterministic random subsamples of the historical record.

Top 10% trade contribution—Share of total positive net P/L attributable to the largest winners.

Expectancy dispersion—Rolling-window variability relative to full-sample expectancy.

04 · Time stability

Does the edge survive different periods?

Years and chronological halves are independent historical segments. Rolling windows reveal whether the edge persists locally or depends on isolated bursts.

Rolling expectancy

expectancy per trade through the sample

Year / period breakdown

only slices meeting minimum trade count

Period	Trades	Net	Exp.	PF	Status

05 · Structural stability

Where does the edge actually live?

Direction and weekday tables are descriptive. A weak slice can identify concentration, but this tool does not claim that deleting that slice would recreate a superior strategy.

Direction

Side	Trades	Net	Exp.	PF	Status

Weekday

Day	Trades	Net	Exp.	PF	Status

06 · Random subsamples

How often does a partial sample retain the edge?

The lab draws deterministic random subsets without replacement from the uploaded history. This is a stability diagnostic, not a Monte Carlo forecast. Use the Monte Carlo Lab for sequence-risk distributions.

50% sample · positive—

Share of 250 random half-samples with positive expectancy.

70% sample · positive—

Share of 250 random 70% samples with positive expectancy.

90% sample · positive—

Share of 250 random 90% samples with positive expectancy.

Subsample expectancy distribution

250 random 70% samples

Interpretation

Rawstocks AI

Strategy Analyst.

Claude can interpret robustness diagnostics together with any Trade Journal and Monte Carlo summaries already saved in this browser.

Robustness synthesis

The deterministic lab results remain the source of truth. Claude interprets the saved Performance Lab summaries and does not replace the calculations.

Context: current lab only

Raw trade files remain browser-local. Only compact summary analytics are sent when you ask the AI a question. Saved lab summaries are stored locally in this browser so the three Performance Lab tools can be synthesized together.

Ask about this strategy

Rawstocks AI Strategy AnalystAnalysis ready. Ask me to combine historical performance, robustness and simulated sequence risk.

Claude Sonnet 5 · deterministic metrics remain the source of truth

Methodology

What this test can and cannot establish.

Robustness is evidence about distribution of historical results. It is strongest when combined with clean data, realistic costs, genuine out-of-sample testing and live forward performance.

Time slicesIndependent historical segments

Years and halves ask whether profitability is concentrated in one regime or persists across materially different periods.

Rolling windowsLocal edge stability

Rolling expectancy and profit factor show how much performance varies through the chronology of the sample.

Random subsamplesDistribution check

Repeated partial samples test whether many different combinations of historical trades still exhibit a positive average outcome.

Critical limitation: this lab never treats arithmetic removal of trades as a simulation of a modified strategy. In a state-dependent system, blocking one trade can change later entries, cooldowns, position state and signal eligibility. A weak weekday, direction or period should therefore be treated as a research lead — not proof that deleting it would improve the strategy.

Robustness score: the score combines profitable time-slice percentage, profitable rolling-window percentage, positive random-subsample percentage, trade-concentration score and rolling-expectancy stability. It is a transparent heuristic designed for comparison and diagnosis, not a probability of future success.