Demo Accounts and Paper Trading

The short answer

A demo account is a mechanics trainer. It teaches you where the controls are, what each order type does when you use it, and whether you can follow a written plan for a sustained period. Those are real skills and they transfer intact.

What it cannot reproduce is the part of trading that happens when money is genuinely at risk: the fill you actually get, the slippage you actually take in a fast market, and the pressure that makes people abandon rules they had followed for weeks. A good demo record is therefore evidence about your process and almost no evidence about your results.

What simulation genuinely teaches

Four things, and they are not trivial.

The interface. Every platform has its own arrangement of order tickets, its own defaults, and its own way of confirming or silently rejecting something. Learning this with real capital exposed is an unnecessarily expensive way to discover that a size field was pre-populated.

Order behaviour. A limit that rests unfilled, a stop that triggers, a stop-limit that triggers and then fails to fill — you can produce all of these deliberately in a simulator and watch what happens. Reading how orders work tells you the theory; making each one happen on purpose makes it stick.

Plan mechanics. Whether your rules are actually executable is a question a simulator answers well. Many written plans contain a step that cannot be performed in the time available, or that requires information you do not have at the moment of decision. That failure shows up identically in simulation and live.

Adherence over time. Can you place forty trades in a row without deviating from your own rules when nothing is at stake? If the answer is no, the answer with real money will also be no, and you have learned something genuinely useful at zero cost.

What it cannot reproduce

Fill quality

A live order has to find a counterparty in a book with finite depth. It queues behind other orders at the same price, it removes liquidity when it executes, and in size it moves the price it is trying to get.

A simulated order does none of this. Most simulators match you against a printed price, or against the current bid or ask, without asking whether that size was available or whether you would have been at the front of the queue. The result is a systematic optimistic bias: the fills you see are the fills the best-placed participant would have received.

Slippage under stress

The bias is worst exactly where it matters most. In a fast move — the one your stop exists for — the live book empties and orders fill some distance from where they triggered. A simulator that fills your stop at the stop price has removed the entire failure mode.

This means slippage estimates taken from a demo account are not usable. If you want a real number, you have to collect it from your own live fills, recording intended price against actual price, as described in costs of a round trip.

Costs

Many simulators either omit fees or apply a simplified figure. A platform that models a single flat commission and a fixed spread has removed the maker/taker distinction, the minimum per-order charge, the partial-fill problem and the overnight financing accrual in one move. Since a round trip pays spread and commission twice, and since FINRA's day-trading disclosure spells out how quickly commissions accumulate for frequent traders, a cost-free simulation can make an approach look viable that is not.

The weight of real money

This is the irreducible one. Simulated losses are information; real losses are also information, but they arrive attached to something that makes people behave differently. Sizing up after a run of wins, refusing to take a loss because taking it makes it real, doubling after a bad day — none of these show up reliably in a simulator, because the trigger is not present.

The CFTC's forex advisory reports that most retail customers in that market lose money once all charges are included. Whatever combination of costs, leverage and behaviour produces that, a demo account will not expose you to it.

How to run a demo period properly

Six rules that make simulated practice transferable.

1. Trade the size you will actually trade. Not the size the platform funded you with. If your real account will be $2,000, set the demo to $2,000 and size every trade from it.
2. Write the plan first. Entry condition, invalidation, exit, size rule, maximum trades per day. If it is not written before the session, the session tests nothing.
3. Never reset the account. A reset erases the drawdown, and the drawdown was the lesson. If the balance is uncomfortable, that discomfort is the closest a simulator gets to the real thing, and deleting it removes the only pressure the exercise had.
4. Journal every trade, including the ones you did not take. Record the reason, the intended price, the actual price and whether you followed the plan. The journaling method applies unchanged.
5. Set an end condition in advance. A number of trades or a fixed period, decided before you start, so you cannot extend it until the result flatters you.
6. Grade the period on adherence, not on the balance. The output you are looking for is a percentage of trades that followed the plan. That number is honest. The balance is not.

Signs you are learning the wrong lesson

• You are trading sizes you would never risk live, and enjoying it.
• You have reset the account more than once.
• You have changed the plan mid-period because it was not working, without recording the change.
• You are taking trades you would skip live, because a demo loss is free.
• You quote the balance when describing how the practice is going.
• You have stopped writing down the trades you rejected, so the record only contains decisions that were easy to make.
• You have been in simulation for months and the remaining open questions are all about your own behaviour under pressure — which is the one question the simulator is structurally unable to answer.

When to stop

Stop when the mechanics are automatic and the adherence number is high and stable across a reasonable sample size. At that point the simulator has given you everything it has.

The transition should be a step down in size, not a step across. Whatever size your plan produced in simulation, start live meaningfully smaller, and expect the adherence number to fall — because the variable that was missing has just been added. Getting it back to where it was in simulation is the actual first milestone, and it has nothing to do with the balance.