Glossary

The sigc vocabulary.

The DSL keywords, runtime operators, and quant concepts you will meet across the docs and this site.

.sig file: A sigc strategy file with four blocks — data, params, signal, and portfolio. The unit the compiler parses, type-checks, and runs.
signal block: A named computation that emits a cross-sectional value per name per date. Signals reference operators and other signals and can be combined with explicit arithmetic.
portfolio block: Turns a signal into a tradable book: a weight scheme, rebalance cadence, transaction-cost model, date range, and optional benchmark.
operator: A vectorized primitive in the runtime — ret, lag, zscore, rank, winsor, ema, rsi, macd, atr, vwap, neutralize, and 100+ more.
content-addressed cache: A cache keyed by the blake3 hash of the IR and inputs, stored on sled. Identical inputs return identical outputs from the cache in microseconds.
IR (intermediate representation): The typed operator graph the compiler lowers a .sig file to before the runtime executes it.
daemon mode: Running the same binary as sigc daemon: a long-running process that owns the cache and serves compile/run requests over nng REQ/REP on port 7240.
nng: The nanomsg-next-generation messaging library. sigc uses its REQ/REP pattern for brokerless client-to-daemon RPC — no Redis, Kafka, or RabbitMQ.
long_short: A portfolio constructor: rank(signal).long_short(top, bottom, cap) goes long the top percentile and short the bottom, capping any single position.
scale_vol: A portfolio constructor that scales weights so realised volatility tracks an annualised target over a lookback window.
winsor: Winsorisation — clipping a signal at a percentile (e.g. p=0.01) to limit the influence of outliers before ranking.
zscore: Standardises a signal cross-sectionally to zero mean and unit variance, so factors are comparable before combination.
neutralize: Residualises a signal against a factor such as the market or a sector, removing unwanted exposure.
Sharpe ratio: A standard backtest metric: annualised excess return divided by annualised volatility. sigc run prints it by default.
Max Drawdown: The largest peak-to-trough decline of the strategy equity curve over the backtest window.
Turnover: How much of the book is traded per rebalance — a proxy for transaction-cost sensitivity.
pysigc: The Python binding to the sigc runtime, so analysts can stay in Jupyter while keeping the compiler’s guarantees on the signal computation.
sigc.yaml: The production posture file: circuit breakers, order rate limits, Prometheus metrics, Slack alerts, and scheduled jobs.

Ready to write your first strategy? Head to the quickstart or the full docs.