Paper Trading: Fills Are Assumed
Paper trading means executing a method in a simulator with no money at risk. It is genuinely useful for testing mechanics and process, and its fills are assumed rather than earned, which is what makes any return figure it produces unreliable as a forecast.
How it works
A simulated account with live prices. Orders are placed, positions are tracked, a balance moves, and no money changes hands.
What it establishes is real and worth having. That you can place the right order type, that your stop attaches correctly, that the platform behaves as expected, and that you can execute the method’s steps without hesitating over which button does what.
Those are genuine prerequisites and there is no reason to learn them with money at risk.
What it cannot establish
A simulator gives you the price you asked for. A resting limit order fills when price touches its level; in reality it joins a queue and may not fill at all. A stop fills at its trigger; in reality it becomes a market order into whatever depth exists.
Both of those failures are systematically favourable in a simulator, which biases every result it produces in the same direction.
Costs are frequently understated too. Some simulators apply no spread, some apply a fixed one that does not widen at the open or on news, and few model commission accurately.
On this site’s shared history the round trip is 2% of a median bar’s range and 45% of the smallest bar. A simulator that halves that turns an unprofitable method into a profitable-looking one, and the difference is invisible in the equity curve.
And it cannot reproduce the experience it is most often recommended for. Holding a losing position is uncomfortable because the money is real; a simulated loss is a number changing colour.
Which is why simulated position sizes drift upward. A hundred-thousand simulated account gets traded at sizes the same person would never take with their own money, and the resulting return is a statement about a size they will never use.
In practice: what to use it for
Test the process, not the profitability. Whether the rules are unambiguous, whether the setups are findable in real time, whether the routine fits the hours available. Those transfer.
The return figure does not transfer and should not be quoted, including to yourself.
A very small live account is the better next step. Real fills, real spreads, real costs, and a real feeling — at a size where the outcome is genuinely irrelevant. Everything a simulator cannot teach is present, and nothing important is at risk.
The one caveat is that costs are proportionally worse on a tiny account, so the returns will look poor. That is fine: you are measuring behaviour and execution, not performance.
The fill assumption is worst in thin volume, where a real order would move price and a simulated one does not.
For a slower method the distortion is smaller, because a few ticks of fill quality is a smaller share of a multi-day move. Simulated results for swing methods are less misleading than for scalping ones.
Some simulators fill a stop at its level across a gap, which cannot happen. That single modelling error removes the largest risk in overnight trading from the record entirely.
One use of a simulator is underrated and worth naming: rehearsing the days you cannot practise. A platform outage, a stop that does not attach, a position opened in the wrong direction, an order for ten times the intended size. Those events are rare enough that a live trader may meet the first one under real pressure, and a simulator is the only place to find out what the recovery procedure actually is.
Write the procedure down while it is cheap to get wrong. Which button flattens a position, what the broker’s phone number is, and what you do if the platform will not load while a position is open. That is a use of simulation nothing else replaces, and it takes an afternoon rather than six months.
What paper trading is not
It is not a performance test. The fills are assumed.
It is not psychological preparation. Nothing about it feels like risk.
It is not a backtest. That runs on history; this runs forward on live prices.
And it is not a waste of time, provided it is used for what it can establish.
When it fails
In a range the cost understatement is fatal to the conclusion. Small bars, many trades, and a simulator charging half the real spread produces a profitable-looking curve for a method that loses money live — and the range is exactly the regime where a method takes the most trades.
The second failure is the size drift. Results from a size you would never take are not results.
A third is treating a good simulated run as validation. Six profitable months on paper is evidence about the simulator’s assumptions.
A fourth is staying in it too long. The mechanics take a fortnight to learn; months of simulation afterwards mostly builds confidence that has no basis.
And a fifth is the reverse error — dismissing it entirely. Learning order types with real money is an unnecessary expense, and the platform-familiarity part of the exercise is genuinely worth doing.
The original data
On this site’s shared 576-bar history the round-trip cost is 0.0098 price units — 2% of the median bar
range of 0.493 and 45% of the smallest bar of 0.022 — and it exceeds 10% of a bar’s range on 15 of the 576
bars. The figures are in research/series-measurements.json, produced by site/measure_series.py.
That 45% figure is the specific number a simulator is most likely to get wrong. On quiet bars the real round trip consumes nearly half the range, and a simulator applying a fixed narrow spread will show a fraction of that. The check worth running before trusting any simulated result: find out what spread and commission the simulator applied, and recompute the curve with your broker’s real figures. If the method does not survive that substitution, it was never a method — it was a description of the simulator’s assumptions.
Related
Simulator covers the tools themselves and what they model. Forward testing is the disciplined version, recorded in advance. And trading plan is what a simulation is supposed to be testing.
Six months of profitable paper trading told me my platform worked and taught me nothing about myself. The first week of a very small live account taught me more than all of it, and the difference was entirely about what it feels like to be down real money.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.