Factor Card

The structured report for one factor result — Health, grade, evidence, risks, and what to test next.

A Factor Card is the server-persisted evidence record for one exact factor result. It turns a raw backtest into something a person or AI agent can review, question, and compare without depending on local files.

How To Read A Card

Read in this order:

Success / Fail
-> evidence quality and Health
-> grade and continuous metrics
-> risks and caveats
-> next controlled experiment

Success And Grade Are Different

Result	Meaning
Success / Fail	Whether every required factor evidence check passed.
SSS, SS, S, A, B, C, D, F	The cross-sectional Sharpe grade relayed by the evaluation runtime.

A factor can receive a non-F grade and still fail a required check. Read the recorded gate evidence before interpreting the grade. See How Factors Are Judged for the current semantics.

Common Card Fields

Field	Meaning
Success / Fail	Combined required-check result
grade and grade score	Relayed categorical and continuous rating evidence
factor idea and formula	What the signal is designed to capture
data and evaluation scope	Headers, bar size, horizon, and visible evidence window
Health	Coverage, missingness, output usability, and any failed Health fields
key metrics	Sharpe, Rank IC, autocorrelation, drawdown, turnover, and available diagnostics
assumptions and caveats	Conditions that may weaken the evidence
next experiment	A proposed controlled change, not an automatic submission

Missing or null evidence stays unavailable; it must not be converted to zero or treated as a pass.

Example

For a factor evaluated on daily bars with a seven-day forward horizon:

Field	Value	Plain English
Success	Fail	At least one required check did not pass
Grade	D	Cross-sectional Sharpe falls in the D band
Cross-sectional Sharpe	0.81	Passes the strict > 0.8 check
Absolute Rank IC	0.008	Fails the strict > 0.01 check
Autocorrelation, lag 1	0.62	Passes the >= 0.4 check
Health	Passed	The recorded output-quality check passed
Max drawdown	−32%	The worst observed peak-to-trough decline
Turnover	0.63	How much the implied portfolio changed
Cost viability	Failed	A diagnostic warning, not a Success gate

The concise reading is: Fail because absolute Rank IC missed its required threshold; D because Sharpe was 0.81; and high implementation risk if cost viability also failed. Those statements answer different questions.

Server Evidence And Charts

The Factor Analysis skill reads product-safe In-Sample (IS) evidence for one exact run. Depending on availability, this can include the Factor Card, Health and rating fields, factor profile, group NAV, daily returns, simulation NAV, simulation PnL, and inert job-linked source used only for explanation.

Do not describe this public analysis as OOS or ALL analysis. A future public contract may expose other windows, but the current analysis surface is IS-only.

Optional Result Bundle

Factor Analysis does not require a local archive. When Factor Mining exports a completed result in a writable host, the canonical local output is one verified ZIP:

Quandora result/factor/<factor_slug>.zip

The ZIP is not automatically extracted or reconstructed. Its runtime manifest is authoritative for the exact included, pending, and omitted items.

Bundle state	What it means
Available	The verified archive is ready.
Partial	The archive is readable and the manifest identifies pending or omitted optional items.
Pending / materializing	The archive is still preparing; request the same result again later.

A delayed bundle is not a reason to start another backtest.