Concepts & glossary

The vocabulary used throughout ClearEdge and these docs, grouped by area. Deep dives are linked.

Data

• Catalog — a ParquetDataCatalog on local disk (./data/catalog). The single system of record for market bars; backtests read only from here. See the data catalog.
• Instrument — a tradeable symbol with a venue, e.g. SPY.XNAS (an equity/ETF) or BTCUSD.CRYPTO (crypto). The text before the dot is the ticker; after it, the venue.
• Asset class — equity or crypto. It changes fee models, fractional sizing, and the annualization factor (252 trading days vs 365).
• Bar / timeframe / bar spec — an OHLCV bar over a fixed interval. A bar spec names the cadence, e.g. 1-DAY-LAST or 1-MINUTE-LAST. Intraday bars come from Databento; higher timeframes can be resampled from a base cadence.
• Ingestion — pulling vendor data into the catalog. Sources: Tiingo (daily), Databento (intraday 1-minute trades → bars + VWAP), and Tiingo crypto. See ingestion.

Strategies

• Strategy — a Nautilus Strategy subclass plus a typed config, registered under a key. The same code runs in backtests and live. See the strategy contract.
• Registry — the in-process catalog of strategies (ats.strategies). Built-ins are registered at import; Studio strategies are loaded into it under a gen:<slug> key.
• Parameter / search space — a strategy’s tunable settings, and the ranges the optimizer may explore for each.

Validation — the gauntlet

The anti-overfit pipeline. Full detail in the gauntlet and the research ladder.

• Backtest — one historical simulation of a strategy on an instrument over a date range, producing metrics, an equity curve, fills, and round-trips.
• Run kind — what a BacktestRun represents: a plain backtest, or a rule_test, optimize, walk_forward, wfo, cross_sectional, pairs, portfolio, portfolio_search, book, or intraday_relval.
• Rule test — the entry-rule significance test: the strategy is run for real, then N times with its entry rule replaced by random entries at the same frequency (same exits, sizing, costs). If the real Sharpe doesn’t clear the 95th percentile of that random cloud, the entry rule adds little.
• Walk-forward — re-running the same parameters on N out-of-sample windows. WFO (walk-forward optimization) is the stronger form: parameters are re-optimized on each window using only the past, traded on the next, and the out-of-sample steps are stitched into one honest equity curve.
• Optimize — an Optuna (TPE) search over the search space, one backtest per trial, with a train/test split so the optimizer never sees the test window.
• Monte Carlo — resampling the trade sequence to estimate the distribution of outcomes (how lucky was this particular ordering?).
• Deflated Sharpe (DSR) — a Sharpe ratio discounted for how many variants were tried; a high DSR is confidence the edge is real, not the luckiest trial. A hard gate.
• Cost stress — re-scoring the strategy at, e.g., 2× the modeled trading costs.
• Buy-and-hold benchmark — every run reports the return of simply holding the instrument, and the strategy’s excess over it. Beating buy-and-hold is the bar that matters.
• Readiness score — a 0–100 score on the Research tab combining the gauntlet stages; a ⭐ marks a strategy that has cleared the hard gates.

Multi-instrument

See multi-instrument engines.

• Cross-sectional — rank a whole universe each rebalance and hold the top-K (momentum/reversion/ low-vol). Its gate is the rank test (does the ranking beat random selection?).
• Pair — trade the spread between two cointegrated instruments, dollar-neutral. The hedge ratio β can be fixed, rolling-OLS, or a Kalman state-space filter.
• Sleeve — one ingredient (a strategy on an instrument, or a pair) inside a multi-strategy book.
• Book — several sleeves run as real strategies on one shared cash account/engine — exactly as a live node deploys — reporting true per-bar capital deployment, not a blend.
• Portfolio search — searching subsets of validated sleeves × allocation methods for the best held-out Sharpe, with a deflated-Sharpe correction for the size of the search.
• Correlation / regime stress — how a book’s diversification holds up in named crises and under a forced correlation spike.

Trading

See paper and live trading.

• Paper trading — simulated execution: IB’s paper account (a DU… id) or a fully local crypto simulator fed by Kraken market data.
• Live trading — real money through IB (a U… account). Gated behind an explicit, audited approve-to-live step and a tighter risk profile.
• Session — one run of a paper/live node: its strategies, fills, telemetry, and lifecycle, recorded in the Sessions tab.
• Risk engine — Nautilus’s pre-trade RiskEngine: orders breaching the per-order notional or submit-rate limits are denied before they reach the broker.
• Kill switch — an operator control that halts a running session.
• Reconciliation — on connect, the node reconciles its view of orders/positions with the broker’s.
• Gateway health — a background monitor probes each IB Gateway and surfaces a health chip.

Platform

• Studio — turn a plain-English idea into a sandboxed strategy; see Studio.
• Research copilot — a per-strategy chat that reads that strategy’s results and proposes revisions.
• Ask the docs — the in-app Help assistant that answers technical questions from this documentation.
• BYO keys — bring-your-own API keys (Tiingo, Databento, Telegram, Anthropic, IB ids), stored in backend/.env; see the configuration reference.
• Audit log — an append-only record of every security-relevant action and order intent.