Glossary

Plain-English definitions for the terms used across Quandora

Definitions are intentionally plain English. Ask your AI agent to explain any term using one exact run as an example.

Research Objects And Skills

Factor — A measurable market feature turned into a score. Quandora tests whether the score contains useful cross-sectional evidence.

Signal — The factor output: one value per market and bar that can be ranked across the active universe.

Task Card — The structured research work order: objective, category, allowed data headers, horizon, and research context. See Task Card.

plugin.py — The executable factor definition validated and run by Quandora. See plugin.py.

Factor Mining — Creates, validates, and backtests a factor. It can also browse caller-owned factor history and explicitly export a Result Bundle.

Factor Analysis — A read-only diagnosis of one exact factor result using server-persisted IS evidence. It does not require a local ZIP.

Factor Card — The structured factor result containing Success/Fail, Health, grade, metrics, caveats, and proposed experiments. See Factor Card.

Success / Fail — Whether absolute Rank IC, lag-1 autocorrelation, cross-sectional Sharpe, and Factor Health all passed their current recorded requirements.

Grade (SSS–F) — A cross-sectional Sharpe band, separate from Success/Fail.

Strategy Building — Selects eligible factors and submits a cross-sectional strategy configuration for backtesting.

Strategy Analysis — A read-only diagnosis of one exact strategy run using its canonical configuration, retained artifacts, and numerical chart evidence.

Official / Mine / Shared — Strategy factor sources. Official factors are read-only product inventory; Mine are the caller's eligible Strategy factors; Shared factors have been admitted to the caller's Strategy pool.

Paper source — A completed, owner-scoped, eligible Strategy run that can be selected for simulated Paper Trading.

Strategy Portfolio — A static set of independently allocated strategy sleeves that can be backtested and then run in Paper Trading.

Result Bundle — The verified ZIP exported for a completed Factor or Strategy result. Its runtime manifest records included, pending, and omitted items. Analysis does not depend on this ZIP.

Live Trading — Real-money execution. It is not part of the public Quandora product, and this documentation does not specify an implementation or control model for it.

Testing Terms

Backtest — Replaying history to evaluate a factor or strategy. It is evidence about the tested period, not a promise about the future.

In-Sample (IS) — Historical evidence in the window exposed by the current public Factor Analysis contract.

Out-of-sample (OOS) — A separate held-out historical window. The current public Factor Analysis skill does not claim OOS evidence.

ALL — A combined scope that includes IS. In Strategy Analysis, ALL must not be described as pure OOS.

Overfitting — When a rule captures historical noise rather than a mechanism that generalizes.

Forward horizon (fwd_period) — How far ahead factor scores are evaluated. Current public tasks use seven daily bars.

Data-header blindbox — The agent sees the allowed header names while Quandora binds changing market data server-side. See Our Data.

Factor Metrics

Sharpe ratio — Return per unit of variability. Current Factor Success requires cross-sectional Sharpe to be strictly greater than 0.8.

Rank IC — Correlation between factor-score rankings and forward-return rankings. Current Factor Success requires its absolute value to be strictly greater than 0.01.

Autocorrelation — Similarity between the signal and its previous-bar value. Current Factor Success requires lag-1 autocorrelation to be at least 0.4.

Health — Recorded checks for factor-output usability, including coverage and missingness under a declared basis. Unknown Health evidence does not pass.

ICIR — IC divided by its variability: a measure of consistency.

IC decay — How predictive evidence changes across forward horizons.

Turnover — How much the implied portfolio changes between rebalances.

Calmar — Annual return divided by maximum drawdown.

Hit rate — The share of observations or trades that were profitable under the stated definition.

Maximum drawdown — The worst peak-to-trough decline in the evaluated path.

Net vs gross — Gross performance is before modeled costs; net performance is after the applicable fee, turnover, and funding effects.

Cost viability — Diagnostic evidence about whether an edge survives modeled costs. It is not a Factor Success/Fail gate.

Paper Terms

Paper Trading — Simulated execution using an eligible strategy source. It does not place live-money trades.

Portfolio snapshot — The current Paper balance, PnL, assets, and open or partially open positions.

Closed position history — Completed net-position lifecycles. Open or partially open positions remain in the current portfolio snapshot.

Fill — A simulated execution record.

Funding — Simulated funding transfers recorded for a Paper run.

Stop — A terminal action for a Paper run. A stopped run cannot be resumed; starting again creates a new run.