How to Paper Trade Properly
To paper trade properly, write the rules before the first trade, use the same size you would use with real money, and subtract real commissions and spread from every result. Assume the worse fill, because a simulator always fills and a market does not.
A simulator is a good instrument for one question and a useless one for another. It can tell you whether you followed your rules. It cannot tell you whether you would have followed them with money on the line.
Before you start
Written rules for entry, stop and size before the first trade. Rules invented during the session are not being tested; they are being fitted to what already happened.
The real commission and spread you would actually pay. From your broker’s schedule, not the simulator’s defaults, which are frequently zero.
A record you fill in for every trade, including the skipped ones. A spreadsheet is enough. The skipped trades are what proves the rules were applied rather than remembered.
The steps
1. Write the rules down before the first trade
Entry condition, stop placement, size calculation, exit condition. A rule you cannot hand to somebody else to follow is not written down yet.
2. Use the size you would actually trade
Not the simulator’s default balance. Your real account, your real risk per trade. A result at ten times your size describes an account you do not have.
3. Subtract the costs the simulator leaves out
Commission both ways, plus the spread. On the shared series a round trip is 2% of a median bar’s range, and simulators routinely charge nothing at all.
4. Assume the worse fill on every order
Entries fill at the least favourable price in the bar and stops fill through the level. A simulator grants perfect execution that no live account receives.
5. Check who else was in the market
A fill in a thin period is optimistic in the simulator and expensive in reality. Note the volume at the time of each trade alongside the result.
6. Record every trade and every skip
Date, rule that triggered, size, entry, stop, exit, cost, and result. Skips get a row too, with the reason. This record is the actual output of the exercise.
7. Run it long enough to meet a losing run
A method that has never had a bad stretch has not been tested. Continue until you have experienced a sequence of losses and recorded how you responded.
How to tell it worked
Judge it at 50 trades and not before.
Count the trades that followed the written rule exactly. 50 out of 50 is the target, and anything below 45 means you are testing improvisation rather than a method. This is the only number on the page that is genuinely under your control.
Check that costs were subtracted on all 50. A record with any zero-cost rows is describing a market that does not exist.
Then confirm you met at least one losing run of 4 trades or more. If the whole sample was comfortable, the test has not covered the conditions that actually end methods, and it needs to continue rather than conclude.
Why the sample size is not negotiable
A short run of results is indistinguishable from chance. On the shared price series, price closes higher 10 bars later on 54% of 566 occasions and higher one bar later on 52% of 571. A coin with that much bias still produces long winning streaks, and a week of paper profits sits comfortably inside what randomness generates.
Direction runs average 2.01 bars with the longest at 11. So a method with no edge whatever will still deliver runs that feel like confirmation, and stopping the test at the moment it feels confirmed is how the exercise gets wasted.
And the honest limit is worth stating plainly. Paper cannot reproduce the experience of losing money you needed, which is where most methods actually fail. It tests the rules and the arithmetic — both worth testing, neither the whole problem.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, 17 have an instruction-shaped
title mentioning paper trading, at a median of 34,273 views across 13 channels, with a maximum of
551,460. Backtesting appears in 116 instructional titles at a much lower median of 7,109. The
counts come from site/rank_howto.py.
Seventeen videos at 34,273 against 116 backtesting videos at 7,109. The manual, slower version of the same idea pulls nearly five times the audience per video from a seventh of the supply — people want to practise rather than to program, and almost nobody is teaching the practice properly.
The answer to the question on that chart is that being up is not the test. The record is. If 50 out of 50 trades followed the written rule and the costs were in, the number that matters has already been produced — and if the sample is 12 trades over a good fortnight, going live is testing the method with money for the first time.
When it fails
The core limitation is that a simulator is emotionally free. A flat, frustrating stretch that would produce impatience and over-trading in a live account produces nothing in a demo, so the behaviour most likely to end a real account is the one behaviour paper cannot test at all. That is not a reason to skip it — it is a reason to treat a good paper result as evidence about the rules and about nothing else.
The second failure is trading a size you would never use. It makes every result unreadable.
A third is a simulator charging zero costs. Costs are a large share of the outcome for anything short-term, and leaving them out inverts the conclusion.
A fourth is stopping at the first good week. That is the moment the sample is least informative and most persuasive.
A fifth is changing the rules mid-test. The record then covers several methods and measures none.
And a sixth is not recording skips. Without them there is no evidence the rules governed anything.
Related
Paper trading covers what a simulator reproduces and where it diverges from a live account. Backtesting is the automated version of the same question and has its own failure modes. And trading plan is where the written rules from step one properly belong.
I wasted a couple of months on this by paper trading at ten times the size I would ever have used, which made every result meaningless and every good week feel like proof of something. Matching the size to what I would actually trade was unglamorous and immediately useful — the wins got small enough to be believable and the costs became visible for the first time.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.