Live Trading

Carry Strategy Logic Into A Real Broker Boundary

The live engine reuses the same strategy contract and event pipeline as backtests, then adds broker adapters, market-data streaming, reconciliation, throttling, risk enforcement, and auditability. This section is where research code meets operational reality.

Configure live trading Read readiness boundaries

Shared strategy contract The engine preserves the backtest strategy interface instead of inventing a second runtime model.

Operational guardrails Risk limits, throttles, reconciliation, and audit events are part of the live surface.

Broker boundaries are explicit Capabilities, validation level, and external restrictions are documented instead of implied.

What The Live Engine Is Responsible For¶

Market Data Intake

Connect to broker streams, normalize updates into engine events, and keep heartbeat expectations visible.

Strategy Dispatch

Run the same callback contract used by backtests so live logic remains behaviorally traceable.

Risk Enforcement

Apply live risk checks, order throttling, and reconciliation before calling the adapter layer production-ready.

Audit And Drift

Track execution reports, account state, and optional backtest-vs-live performance drift.

Entry Point¶

Primary source-build entry point:

./build/bin/regimeflow_live --config <path>

On Windows, use build\bin\regimeflow_live.exe --config <path>.

The live CLI loads .env automatically if present and merges environment variables with live.broker_config.

Operational Checklist

1. Configure Broker Inputs

Use Live Config to set the adapter surface, symbols, risk limits, and broker-specific options.

2. Validate Broker Boundaries

Use Brokers to confirm what is implemented, paper-validated, or externally restricted.

3. Check Readiness

Use Production Readiness before treating a configuration as deployable.

4. Review Resilience

Use Resilience to understand reconnect, reconciliation, and operational failure handling.

What The Live Engine Does¶

Connects to the broker adapter.
Subscribes to market data.
Runs strategy callbacks on incoming events.
Applies risk limits and order throttling.
Emits audit logs and system health events.
Optionally tracks live performance drift vs a backtest baseline.

Replay And Parity Capture¶

Live market updates now have a direct adapter into the shared backtest event model through live::to_engine_event(update). Set live.replay_journal_path to capture normalized live market-data events as JSONL through engine::ReplayJournalWriter; the same parser is available from regimeflow/engine/replay_journal.h for replay inspection and parity tests.

This is intentionally below the broker adapter layer: the journal records normalized engine events, not raw broker socket payloads. The live journal also captures order-manager lifecycle updates and risk/safety gate decisions as replay events, keeping replay artifacts portable across Alpaca, Binance, Interactive Brokers, and simulated feeds.

Live Performance Drift Tracking¶

Enable live drift tracking by setting metrics.live.enable: true. When enabled, the engine writes:

live_drift.csv for time-series snapshots.
live_performance.json for a summary.

You can also point the tracker at a backtest report JSON via metrics.live.baseline_report.

Execution Quality Tracking¶

LiveOrderManager now keeps an in-process ExecutionQualitySnapshot for broker-order outcomes. It tracks submitted orders, submit rejections, broker rejections, acknowledgements, partial fills, filled orders, cancellation reports, average acknowledgement latency, average fill latency, and limit/stop-reference slippage in basis points. When a submit-time quote is available, it also tracks effective spread cost versus the quote midpoint. Orders that carry queue_position and expected_queue_delay_ms metadata also get queue-model attribution, reporting realized fill latency versus the model estimate. Filled orders are rolled up by routing venue so promotion reviews can compare fill latency, quantity, slippage, and spread cost across venues. Use execution_quality() for dashboards, audit exports, and paper-live validation reports.

Prometheus Export¶

regimeflow/live/prometheus_exporter.h provides Prometheus text exposition helpers and a minimal HTTP scrape endpoint for live operations. Use dashboard_snapshot_to_prometheus(...) for account/health/dashboard gauges and live_metrics_to_prometheus(...) when you also want execution-quality counters, latency/slippage, queue-attribution, and spread-cost gauges in the same scrape payload. Set metrics.prometheus.enabled: true to let LiveTradingEngine serve the payload directly.

Read This Section With The Right Mental Model¶

Live execution is not just “backtest with sockets”.
Broker capabilities, IP restrictions, venue permissions, and account state remain external constraints.
Validation status matters. A working adapter does not automatically imply unrestricted broker access.
Operational truth comes from reconciled positions, account state, and audit logs, not from a single green startup message.