A paper account can answer useful questions before real money is involved. Can you enter the intended order? Does the exit reduce the position? Can you follow a rule after several losses? Those are different questions from whether the strategy will make money.
The best use of simulation is to turn each uncertainty into a test. A rising simulated balance is much less useful if you cannot explain what produced it.
Learning the order sequence
Use a balance and position sizes similar to the account you expect to trade. A million-dollar default balance can conceal mistakes that would dominate a smaller account. Keep the same balance through losing periods rather than resetting the experiment when it becomes uncomfortable.
Practice a complete trade: enter, receive a partial fill, cancel the remainder, modify the exit, and close only the intended position. Confirm the resulting position after each step. A submitted cancellation is not the same observation as a confirmed cancellation.
Deliberately test awkward cases. What happens when an order is too small, margin is insufficient, a stop triggers during a fast move, or a connection is lost? These exercises are useful even if the strategy itself has no forecasting advantage.
How a simulator decides the fill
A simulator must decide whether an order would have filled. It cannot simply replay your actual place in a live queue because the real market never saw the order.
Interactive Brokers' paper-trading documentation, for example, describes simulated fills based on the top of the book rather than deep-book liquidity, and differences in supported order behavior. Those are implementation limits, not flaws unique to one simulator.
Suppose the ask offers 20 units at $100 and the next 80 at $100.10. Buying 100 units would cost $10,008, for an average of $100.08, if those quantities remained available. A simulation that gives all 100 units at $100 understates entry cost by $8. Repeated hundreds of times, small differences can dominate a modest trading advantage.
A resting limit order has the opposite problem. A chart touching your price does not prove that enough orders traded ahead of you to fill yours. Record whether the simulated result depends on fills that would have required unusually favorable queue priority.
A rule fixed before the result
Write the entry condition, exit condition, size, and allowed exceptions before the observation period. Record each qualifying opportunity, including the ones you missed. Otherwise a discretionary record can quietly become a collection of remembered successes.
Keep two results where practical: the simulator's reported P&L and a conservative estimate using realistic spread, slippage, fees, and funding. The CFTC's discussion of hypothetical trading results explains why simulated performance must be distinguished from an actual track record.
Also compare with a simple alternative. If the strategy held a rising asset almost continuously, how much did the entry and exit decisions add beyond owning the asset? If it earned money by accepting occasional large losses, has the sample contained enough adverse conditions to reveal that pattern?
Three kinds of evidence
An operational test establishes that the workflow behaves as intended. A prospective paper record tests whether a fixed rule behaves plausibly on new data. A small live record reveals actual fills, costs, and decisions when losses are real. Success at one stage does not automatically establish the next.
There is no magic number of paper trades that proves readiness. Ten trades with the same underlying exposure may contain less independent information than the count suggests. A quiet month also says little about behavior during the event the strategy is most vulnerable to.
From simulated fills to real ones
If moving to limited live exposure is appropriate, compare expected and actual fills trade by trade. Did slippage widen at the same times? Were losses harder to accept? Did the strategy require attention you could not provide?
The purpose of paper trading is to make those differences smaller and easier to explain. A simulation is doing its job when it exposes a weak assumption before that assumption becomes an expensive habit.