Practice: paper trading and bar replay

Nobody argues against practice. The problem is that most trading practice is indistinguishable from watching. You open a demo account, you click buy on things that look interesting, you make imaginary money, and at the end of a month you have learned the platform's keyboard shortcuts and nothing else.

Practice only teaches you something if there is a specific thing being tested and a way to be wrong. The two tools here test different things, and using them interchangeably is why so much screen time produces so little improvement.

Bar replay: the rules, one decision at a time

Bar replay steps through historical candles one bar at a time, with everything after the current bar hidden. You see what a trader on that date saw. Then you advance one bar.

Replay is for learning what a rule actually feels like to follow. Reading that the exit is 3 ATR below the running high takes ten seconds. Watching a position give back eleven consecutive days without touching the stop, and having to press the next-bar key each of those days, teaches something the sentence cannot.

Its real advantage is compression. A trend that took four months happens in a few minutes of clicking, so you can see twenty full trade lifecycles in an afternoon instead of two years. That is the only honest way to build a sense of how often a rule produces a long flat stretch, because in real time you would quit before finding out.

The failure mode is specific and worth guarding against. You know the data is historical, and some part of you knows the asset survived. You will hold through declines in replay that you would not hold through live, partly because there is nothing at stake and partly because you may recognize the period. Pick assets and dates you do not know, or the exercise degrades into confirmation.

Replay also cannot teach position sizing in any felt way, and it will not teach you anything about execution, because your fill is whatever the next bar says it is.

The simulator: the workflow, in real time

Paper trading with real prices tests something different. It is not about whether the rule works, it is about whether you can operate the whole loop: identify the condition, compute the size from the stop, place the order, set the stop, log it, and leave it alone.

Most people who lose money live are not failing at analysis, they are failing at process. The stop that never got placed. The size computed in their head and wrong by a factor of two. The position opened and then modified twenty minutes later for no written reason. The simulator is where those errors are cheap and countable.

Used properly it has rules of its own. Use the equity you would actually trade, not a hundred thousand you do not have, because size behavior does not transfer between accounts of different scale. Take every signal your plan generates, including the boring ones, because skipping the dull trades is precisely the habit that ruins a live account. Log every one in the journal with the same notes you would write live, or you are practicing half the loop.

What neither one can do

This is the part that gets skipped in every article about practice, and it decides everything.

Fake losses do not hurt. Watching a simulated position go 200 dollars against you is a data point. Watching real money do it produces a physical reaction, and that reaction is what pushes people to cut winners early, widen stops, and re-enter immediately after a loss. No simulator reproduces it, and no amount of paper trading inoculates against it.

Execution is idealized. Simulated fills are usually clean. Real fills involve spread, slippage, funding, partial fills on thin books, and occasionally an exchange that is not responding at the moment you need it. The gap between the two is largest exactly when volatility is highest.

Sample size is still small. Thirty simulated trades tell you very little, whether they went well or badly. A run of luck in either direction dominates that number completely, and a good month on paper is not evidence of anything except a good month.

Using both, in order

A workable sequence, framed as practice rather than as advice about money.

Replay first, until the rules are automatic. You should be able to state, before pressing next-bar, what the rule requires. If you are still deciding what the rule is, you are not ready for real time.

Then the simulator, for process. Run the full loop at realistic size until you go a stretch of weeks without a process error: no missing stop, no size computed by feel, no unlogged trade.

Then treat the first live period as continued practice at the smallest size that still feels real. The purpose is not the result, it is to find out what you do when it hurts, at a scale where finding out is affordable.

Practice does not remove the emotional part. It removes everything else, so that when the emotional part arrives it is the only variable left, and you can finally see it clearly instead of confusing it with not knowing the rules.