Monte Carlo & Robustness

What Monte Carlo & Robustness is for

Monte Carlo & Robustness asks how sensitive the selected Run Output is to the path, sample, costs, time concentration, and other retained uncertainties. It contains the Robust analysis that applies to this exact run.

Study-level selection evidence remains in Study Results. This page does not reconstruct a candidate search from one promoted result.

Confirm the analysis depth

Summary contains point estimates and the basic scope needed to explain what happened.

Robust includes Summary and adds applicable retained-data diagnostics. It never hides another automatic execution battery behind the label. If an analysis requires a new Strategy execution, it must be designed and queued as an explicit Study method.

Every component reports one of three states:

• Available — the named calculation has the required retained inputs.
• Not applicable — the method does not apply to this result or decision.
• Blocked — a required input, sample, alignment, or artifact is missing or invalid.

Read bootstrap and Monte Carlo evidence

The run can retain trade or return bootstrap intervals, trade-order permutations, Monte Carlo distributions, tail-loss summaries, and simulated equity paths.

Review:

1. what unit was resampled;
2. iteration count, seed, and interval level;
3. the observed result's position inside the distribution;
4. both profit and drawdown tails; and
5. any insufficient-sample or dependence warning.

These distributions describe sensitivity under the recorded method. They are not forecasts that future performance will remain inside the displayed range.

Review track-record strength

Where the sample and metric contract allow it, Robust can include:

• probabilistic and deflated Sharpe;
• minimum-track-record evidence;
• outlier dependency;
• null-model or matched-random-entry comparisons; and
• regime, session, side, or market decomposition.

Read each named component rather than relying on one blended score. The intended scorecard is a transparent list of component states, not an opaque numeric quality rating.

Test cost and time sensitivity

Robust can show:

• cost tolerance and break-even cost for the selected decision metric;
• rolling performance stability;
• concentration in a small number of periods;
• drawdown duration and recovery uncertainty; and
• unrecovered drawdowns as censored rather than falsely completed episodes.

If nonlinear sizing or execution means cost tolerance cannot be derived from retained data, the component remains Not applicable and points to a separately designed Study.

Read benchmark evidence

Benchmark-relative evidence appears only when the run binds an exact benchmark with compatible currency, session, calendar, and time alignment. Applicable measures can include alpha, beta, tracking error, information ratio, and upside or downside capture after their metric contracts are satisfied.

A missing benchmark is not a failed Strategy. It makes benchmark-relative conclusions unavailable.

Read Portfolio robustness

Portfolio Robust can additionally examine:

• dependence among aligned member return series;
• diversification and concentration;
• contribution to risk;
• concentration changes across retained scenarios;
• policy counterfactuals and stress scenarios; and
• member-level cost, time, and recovery sensitivity.

Changing membership, insufficient overlap, sparse history, or collinearity must remain visible in the component status.

Keep the conclusion bounded

Robust evidence cannot repair leakage, biased data, unsupported fidelity, selection bias from an unseen search, or future regime change. Use Validation for execution and data limitations, and open the parent Study Result for candidate-selection evidence.