From backtest to paper trading to live
Xavi ·
A strong backtest does not earn a promotion to live trading.
It earns a chance to face a different test.
Historical testing can show how rules behaved on past data under modeled fills. It cannot show whether today’s signals arrive as expected, whether the broker accepts the order, or how a live spread behaves in the queue.
Paper mode exists to expose those gaps without putting capital at risk.
Freeze the research version
Before moving on, save the strategy exactly as tested.
Record the rules, date range, data settings, assumed slippage, commissions, and position size. Write down what you expect to happen during a normal session: signal time, contract selection, order type, exits, and any daily limits.
If you change the rules during observation, call it a new version. Otherwise the evidence becomes a mix of different strategies.
Paper mode tests the decisions and the plumbing
Paper trading runs the strategy against current markets through a broker’s simulated order workflow. Two different things get tested at once, and it helps to read them separately.
The first is whether the logic travels from history into the present. Did the signal occur at the intended time? Was the correct expiration available? Did the strategy choose sensible strikes? Did it stay out when a filter said it should?
The second is the plumbing. Paper mode can reveal rejected order types, contract-permission issues, connection problems, cancellations, and unexpected order-state behavior. Those are operational findings, not minor details. A strategy that cannot reliably manage its orders is not ready for real capital regardless of its backtest.
Paper fills deserve skepticism. Broker simulators may fill orders more easily than a live market, handle queue position differently, or use a different data feed. A profitable paper period does not validate the historical expectancy.
Keep a log. Compare expected and observed behavior rather than staring at hypothetical P/L.
Live trading changes the question
Live mode introduces real liquidity, market impact, partial fills, connectivity risk, and emotional pressure.
The decision to trade live is personal and separate from the software test. A backtest cannot decide whether the possible loss fits your finances or risk tolerance. Neither can an AI assistant.
If a trader chooses to proceed, smaller size reduces exposure while the operational assumptions meet reality. It does not make the strategy safe. Direct broker monitoring and a written response for rejected orders, disconnects, and unexpected positions are still necessary.
Automation should never be treated as permission to stop watching the account.
Promote behavior, not recent profit
Set criteria before you move on.
For paper mode, you might require a defined number of sessions with correct signal timing, contract selection, and exits, together with successful order submission, cancellation, reconnection, and position reconciliation.
Do not make a short run of profits the gate. A few outcomes can be luck in either direction. The purpose of this stage is to exercise the process.
Rare strategies need more calendar time because they produce fewer observations. Frequent strategies generate events quickly but may still go weeks without encountering a connection failure or a disorderly market. Pick the observation window in advance so a pleasant result does not tempt you to shorten it.
Compare the same strategy across stages
Backtest and paper results will not match perfectly. The data frequency, current quotes, and broker simulator can differ.
Investigate the differences:
- Did both versions produce the same signal?
- Did they choose the same type of contract?
- How far was the observed paper fill from the quote midpoint?
- Did an order remain open longer than the historical model assumed?
- Did any manual intervention change the outcome?
This comparison is much easier when the strategy logic stays the same.
StratVerra is built around that continuity. You can create a strategy in plain English or with the no-code builder, backtest it on hosted minute-level data, and run the same strategy through supported brokers in paper or live mode.
Start the free trial with one idea you already understand. Write the expected behavior first, run the backtest under more than one cost assumption, and keep live trading out of the decision until paper mode has answered its questions.