Definition
Live trading is the deployment of a strategy against real markets with real capital — orders that route to an exchange, fills that settle on a brokerage statement, profit and loss that hits a balance sheet. It sits at the end of a pipeline that begins with research, passes through backtesting, and graduates through paper trading. The label looks deceptively similar to its predecessors but the engineering problem is qualitatively different.
The distinguishing feature is consequence. In a backtest a bug rewrites a CSV; in paper trading a bug invalidates a simulated track record; in live trading a bug moves money. Every operational concern that backtests can finesse — latency, partial fills, broker outages, market-data gaps, malformed orders, daily margin calls — becomes a first-class engineering requirement.
Why it matters
How it works
A production live-trading stack has at minimum five layers. A market-data layer ingests the feed, validates timestamps, and routes ticks to the strategy. A strategy layer computes the desired position. An execution layer translates desired position into orders and routes them to the broker via API, tracking each order through its lifecycle. A risk layer sits in front of execution and rejects any order that would breach a position, exposure, or drawdown limit. An observability layer logs every event, computes live P&L, and pages a human when anything looks wrong.
The transition from paper to live is the moment most quant projects fail. The strategy that delivered a beautiful simulated Sharpe ratio in paper mode begins generating real slippage, real rejections from the broker for malformed orders, real losses from a hard-coded assumption that the market is always open. Successful transitions stage capital gradually — start at a tenth of the intended size, watch the operational metrics for a few weeks, scale only after the live behaviour matches the paper behaviour to within a tolerance set in advance.