How Factors Are Judged

The current Success checks and SSS–F grade behind a Factor Card — evidence, not promises.

Every submitted factor is evaluated on server-bound market data. Read the two headline results separately:

• Success or Fail asks whether every required evidence check passed.
• Grade reports a cross-sectional Sharpe band.

The evaluation runtime owns these semantics. The Factor Card's recorded values, thresholds, Health evidence, and failure reasons are authoritative for the exact run.

Current Public Evidence Scope

The Factor Analysis skill currently reads product-safe In-Sample (IS) evidence. It does not claim OOS or ALL evidence. IS is historical evidence and does not predict future or simulated performance.

Success Or Fail

A cross-sectional factor succeeds only when all four current checks pass:

Check	Success condition	What it asks
Absolute Rank IC	> 0.01	Is the ranking relationship strong enough?
Autocorrelation, lag 1	>= 0.4	Is the signal sufficiently persistent from one bar to the next?
Cross-sectional Sharpe	> 0.8	Was risk-adjusted cross-sectional performance strong enough?
Factor Health	Passed	Was the factor output sufficiently complete and usable under the recorded Health basis?

The Rank IC and Sharpe checks use strict greater-than comparisons; autocorrelation includes the threshold. Unknown, unavailable, or non-finite required evidence does not pass.

Cost viability, turnover, drawdown, and other diagnostics remain important, but they do not independently determine Factor Success or Fail.

Health And Coverage

Health checks factor values within the active universe. A symbol's active span runs from its first valid value through its last valid value; cells outside that span do not count against active coverage.

The card can record Health metrics, their thresholds, the coverage basis, and the exact failed fields. Treat missing or null Health evidence as unknown, not as passed. When comparing two runs, compare Health directly only when their windows, active-universe definitions, missing-value handling, and thresholds match.

Grade

The grade bands use cross-sectional Sharpe:

Cross-sectional Sharpe	Grade
< 0.8	F
0.8 – < 1.2	D
1.2 – < 1.4	C
1.4 – < 1.6	B
1.6 – < 1.8	A
1.8 – < 2.0	S
2.0 – < 2.2	SS
>= 2.2	SSS

The grade is relayed evidence, not a promotion decision. It does not override a failed required check and does not say that a factor is ready for real-money trading.

A successful factor showed evidence under the tested conditions. A backtest is evidence about the past, not a guarantee of future returns.