Paper Trading vs a Small Live Allocation

Before a new strategy gets a real capital allocation, most firms run it through one or both of two intermediate stages: paper trading, where the strategy generates real signals and simulated orders using live market data but no actual money changes hands, and a small live allocation, where the strategy trades real capital, but at a fraction of its intended eventual size. It's tempting to treat these as interchangeable steps on the way to full launch — cheaper and more expensive versions of the same test — but they actually verify different things, and skipping either one leaves a specific kind of risk unchecked.

Paper trading is built to catch operational and technical failures cheaply: does the strategy correctly connect to live data feeds, generate signals on schedule without crashing, produce orders that pass internal risk checks, and log everything correctly, all under the pressure of a live, moving market rather than a static historical backtest? Because no real capital is at risk, paper trading is a low-cost way to run a strategy for weeks before committing money to it. What it cannot test is anything to do with real execution: paper trades are typically filled at whatever price the simulator assumes (often the quoted mid-price or last trade), with no real market impact, no real slippage from actually competing with other orders for the same liquidity, and none of the psychological or organizational friction that comes from a trader or portfolio manager watching real money move.

A small live allocation exists specifically to test what paper trading cannot: real fills, real slippage, real market impact (even if small at this size), and whether the humans overseeing the strategy actually behave the way the plan assumed they would once genuine profit and loss is at stake — a manager who calmly accepted a simulated ten-day losing streak on paper sometimes reacts very differently to the same losing streak with real capital behind it. It also forces the operational plumbing around real trading — actual broker connectivity, real settlement, real compliance reporting — to be exercised for real, which a purely simulated system never has to confront.

For example, a new strategy passes six weeks of paper trading with no technical issues and simulated performance in line with its backtest. Moving to a small live allocation of $2 million then reveals that its actual fills, on a subset of less liquid names, are systematically worse than the paper-trading simulator assumed — a real execution cost that paper trading was structurally unable to detect, and that would only have shown up in full once the strategy was already running at a much larger, harder-to-unwind size.

What this means in practice

Treat the two stages as answering different questions in sequence — paper trading first, to clear the strategy of basic technical and operational failure modes cheaply, then a small live allocation, to learn what real execution and real psychology actually do to the strategy before scaling it further. Skipping straight from a backtest to a full-size live allocation, or from paper trading straight to full size without an intermediate small-money step, removes exactly the stage designed to catch the failure mode most specific to each transition.