Paper Trading: How to Practise Properly

At a glance

Good for: Validating process, platform mechanics, and rule clarity
Bad for: Testing your psychology or predicting live returns
Minimum useful run: 30 to 50 trades, or one full quarter
Graduation test: Rules followed on every trade, journal complete

Key takeaways

Paper trading proves that your rules are unambiguous and that you can operate your platform. It cannot prove you will follow rules when money is at risk.
Simulated fills are optimistic: demo engines usually fill at the quoted price with no queue position, no partial fills, and no market impact.
Run the simulation at the size you will actually trade, so the numbers translate directly and the discipline transfers.
Judge the run on rule adherence and execution quality, not on profit. Profit over 30 trades is mostly noise.
Move to live money in small size sooner than feels comfortable: the last 20 percent of learning only exists when losses are real.

What paper trading genuinely tests

A simulated account is a process rehearsal. It answers questions that have nothing to do with market prediction and everything to do with whether your operation works.

Are the rules unambiguous? You will discover within a week which rules require a judgement call you never specified. That discovery is the main value of the exercise.
Can you operate the platform? Order types, bracket orders, position sizing tools, and alerts should become automatic before money is involved. Fat-finger errors are common and expensive.
Does the routine fit your life? A strategy that requires attention at 3pm on weekdays fails immediately when you are in meetings, and paper trading exposes that in days rather than months.
Is the data what you expect? Splits, dividends, contract rolls, symbol changes, and time zones all cause surprises that are much cheaper to meet in simulation.
Does the strategy produce roughly the number of signals the backtest predicted? A large discrepancy usually means a bug in the backtest or a misunderstanding of the rules.

What paper trading cannot tell you

Aspect	Why simulation misleads	What to do instead
Emotional response	Simulated losses cost nothing, so loss aversion is absent	Trade live at minimal size as early as possible
Fill quality	Demo engines usually fill instantly at the quote	Assume worse fills; model half-spread plus slippage
Queue position on limit orders	Simulators often fill limits when price merely touches the level	Require price to trade through your level before assuming a fill
Market impact	Your simulated order does not exist in the book	Size checks against average volume; stay small
Short availability and borrow cost	Demo accounts allow shorting anything, free	Verify borrow before treating a short strategy as viable
Behaviour in a crisis	A simulated 15 percent gap produces no adrenaline	Study the backtest drawdown and pre-commit to the response

How to run a paper trading period properly

1
Fix the parameters in advance
Starting equity equal to your real intended capital, the same instruments, the same strategy rules, and a fixed end date or trade count. Write them down so you cannot revise mid-run.
2
Use realistic assumptions where you control them
Enter orders at prices you would actually get: for market orders assume the far side of the spread; for limit orders require the price to trade through your level, not merely touch it.
3
Log every trade as if audited
Timestamp, rule that triggered it, intended entry and exit, actual fills, position size calculation, and whether you deviated. A journal is the output of this exercise; profit is not.
4
Track process metrics, not just returns
Rule adherence rate, number of missed signals, average delay between signal and order, and slippage versus assumption. These predict live performance far better than simulated profit.
5
Compare against the backtest
Signal counts, win rate, and average win and loss should be in the same neighbourhood. Large gaps indicate a bug, a data difference, or a misunderstood rule, and must be resolved before going live.
6
Set an explicit graduation test
For example: 40 trades completed, rule adherence above 95 percent, journal complete every day, no platform errors in the last 20 trades. Profitability is not on the list.

When and how to go live

Go live once the process is clean, not once the simulation is profitable. Waiting for a profitable paper run usually means waiting for a favourable random sequence, which teaches nothing and delays the only learning that matters.

Start at minimum viable size. One share or one micro contract per position. The goal is to introduce real loss aversion at a cost of pennies.
Keep the same rules exactly. Changing the strategy at the same time as changing to real money makes the comparison meaningless.
Scale in defined steps. For example, double position size after every 25 trades in which rule adherence stays above 95 percent, and step back down after any deviation.
Expect degradation. Live results are typically worse than simulated ones because of fills, timing, and hesitation. If live expectancy is close to the simulation, that is a good result.
Run simulation in parallel for new ideas. Once live, any new strategy goes through the same pipeline rather than being added directly to the funded account.

Alternatives and complements to paper trading

Bar replay: Replaying historical data bar by bar with the future hidden. It compresses months of practice into hours and is the fastest way to build pattern recognition, though it lacks live execution realism.
Backtesting: Testing rules across years of history. It answers whether the edge exists at all, which simulation over 40 trades cannot. Do this first, always.
Walk-forward analysis: Repeatedly fitting on one period and testing on the next. It is the closest statistical analogue to forward testing and uses far more data.
Monte Carlo simulation: Reshuffling your trade sequence thousands of times to see the range of drawdowns that the same edge could plausibly produce. It prepares you emotionally in a way that a profitable paper run does not.

Frequently asked questions

How long should I paper trade before using real money?

Long enough to complete 30 to 50 trades and to encounter the routine problems of your platform and data, which is typically four to twelve weeks depending on style. Beyond that the returns diminish sharply, and continuing usually reflects avoidance rather than caution.

Why do I make money on paper and lose it live?

Three usual causes. Simulated fills are better than real ones, especially with limit orders and in fast markets. Real losses trigger hesitation, early exits, and rule breaking. And paper runs are short, so a profitable one may simply have been a favourable sequence. Compare rule adherence and slippage between the two to identify which is dominant.

Is a demo account the same as paper trading?

Broadly yes: a demo account is a broker-provided simulated account, usually with live prices. Quality varies considerably. Some simulate spreads, partial fills, and latency; many do not. Check what your provider actually models before trusting results.

Should I paper trade with a large simulated balance?

No. Use exactly the capital you intend to fund. A 100,000 USD demo account for a 5,000 USD trader produces position sizes, risk percentages, and instrument choices that will not transfer, and it makes losses feel trivially small.

Can I skip paper trading if my backtest is strong?

You can skip the profitability question but not the operational one. A short forward-testing period catches data problems, rule ambiguities, and platform mistakes that no backtest exposes. Running it at live minimum size instead of in simulation is a reasonable compromise.

Test this idea before you trade it

Describe the rules in plain language and AlgoTrader AI turns them into a structured strategy blueprint with a configurable historical backtest, cost assumptions, and exportable code.

Build a backtest

Keep reading

Referenced by

Forward Testing: The Only Truly Clean Evidence

Educational use only. This guide explains how a strategy works. It is not investment advice, not a recommendation, and no result described here is a forecast. Test any approach on historical and out-of-sample data, size positions conservatively, and never risk money you cannot afford to lose.