Paper trading automated strategies
What is paper trading for?
Paper trading runs the system against live market data with simulated money. It sits between a backtest, which replays history, and live trading, which risks capital. It answers a specific set of questions well and a different set badly, and knowing which is which is most of the value.
The correct order is backtest, then paper, then live, and the most common mistake is compressing the middle step because it is boring.
What does paper trading prove?
That the machinery works
Orders are formed correctly, protection is placed, positions reconcile against the broker's record, the system restarts cleanly and resumes. None of this can be established by a backtest and all of it must be true before capital is at risk.
That it behaves as the study suggested
You can compare live behaviour against what the backtest predicted. Divergence early is the most valuable signal available, and it is only visible if the research was reproducible in the first place.
What it refuses
Watching what a system refuses is the observation most people skip. Watching what a system declines, and why, tells you far more about the quality of its judgment than watching what it takes.
Whether you can leave it alone
Running unattended for weeks reveals whether you actually trust it, and whether it needs attention it should not need. If you find yourself checking constantly, something is wrong with the system or the arrangement.
What does paper trading hide?
Your fills are better than they would be
Simulated fills do not compete for liquidity and do not move the market. Expect real execution to be worse, particularly on anything thin or multi-leg.
Nothing is at stake, including for you
The behavioural test is absent. Watching a simulated drawdown is not watching your own money fall, and how you respond to the second is not predicted by the first.
Rejections and edge cases may not appear
Margin calls, restricted account states and borrow failures often do not surface in simulation, and they are ordinary live.
A quiet period proves very little
If nothing interesting happened, the run has confirmed that the system runs. It has not told you how it behaves when conditions change.
How long should you paper trade?
Longer than feels necessary, which is months
Two weeks tells you the software works. Seeing more than one kind of market takes a quarter or more, and that is what you are waiting for rather than a target number of trades.
Until you have seen it refuse and recover
Two specific events are worth waiting for: a stretch where it declines repeatedly, and a day where data quality dropped. How it handles both is the substance of the evaluation.
Until boredom, not until a win
Stopping the moment results look good selects for a favourable run. The discipline is to keep going after it has convinced you.
How do you move from paper to live?
Start at a size where total loss is annoying
Whatever paper showed, first live capital should be an amount whose complete loss would not matter. You are buying the option to keep going after a bad stretch.
Keep a paper account running alongside
A parallel simulated account gives you a reference for whether unusual behaviour is the market or the system. It costs nothing and answers a question you will eventually have.
Expect the first live weeks to feel different
Real fills, real rejections and real emotion arrive together. Sizing small is what makes that survivable rather than instructive in an expensive way.
Which mistakes waste the exercise?
Treating it as a demo rather than the product
If what you are running in paper is a reduced version, you are evaluating the part you were shown. The evaluation should be the complete system with the account restricted, not a limited build.
Intervening when it does something you dislike
Overriding a paper decision destroys the only clean record you will get. The point is to observe what it does unattended, including the parts you disagree with.
Not reading the record
A paper run nobody examines has produced nothing but reassurance. Set the same reading cadence you intend to keep once live, and find out now whether you will actually keep it.
Common questions
- How long should I paper trade an automated strategy?
- Longer than feels necessary. Two weeks confirms the software works; seeing more than one kind of market usually takes a quarter or more. Wait specifically for two things: a stretch where the system repeatedly declines to trade, and a day where data quality dropped. How it handles both is the substance of the evaluation.
- What does paper trading not show you?
- Realistic fills, since simulated orders do not compete for liquidity or move the market. The behavioural test, because watching a simulated drawdown is not watching your own money fall. And edge cases such as margin calls, restricted account states and borrow failures, which often never surface in simulation and are ordinary live.
- Is paper trading better than backtesting?
- They answer different questions and the order matters. A backtest tests the approach across history; paper trading tests the machinery against live data, whether orders form correctly, protection is placed, positions reconcile and the system restarts cleanly. Backtest first, then paper, then live, and do not compress the middle step because it is boring.
- Should a paper trading evaluation be the full product?
- Yes, with the account restricted rather than the software reduced. If a vendor offers a limited build for evaluation, you are assessing the part they chose to show you and buying the part you did not see. The evaluation should include the reproducible research and the written constraints as well.
TradeAgentic is an autonomous trading desk for macOS and Windows, licensed to operators and firms who intend to run it themselves.
Request licensing terms