From Backtest to Live: Paper Trading and Small Size

The backtest filled every order at the price it asked for. It never hesitated at a stop, never skipped a signal after three losses in a row, and was never stuck in a meeting when the close came around. You are not a spreadsheet. The step between a backtest and full size exists to find out which of those assumptions you can meet, and it comes in stages that test different things.

Paper trading tests the routine

Paper trading means running the rules in a simulated account with no money at risk. It answers practical questions. Can you compute the signal before the close and get the order in on time? Do you place the stop the same day, at the right price? Do you remember the tenth-session exit? Does anything in the rules turn out to be ambiguous once real charts are moving?

Log every paper trade the way the backtest logged its trades: date, signal, entry, stop, exit, result in R. When a trade differs from what the rules say, note why. Missed signals count too.

Paper trading has limits. A simulator often fills at the quoted price whatever the size and whatever the liquidity, which makes its fills kinder than a live market’s. And a paper loss costs nothing, so it tells you nothing about whether you will honor the next stop.

Once the routine runs cleanly, stop paper trading. More of it adds little.

Small real size tests you and your fills

The next stage uses real money at a fraction of the planned risk. Say the plan calls for $500 of risk a trade. Trade the same rules at a quarter of that.

Everything else stays identical: the same entries, stops, exits and filters. Only the dollars shrink.

Even $124 is enough for a loss to register. Watch for the moves the rules do not allow, such as nudging a stop lower, skipping the signal after a loss, or banking a winner early that the rules say to hold for ten sessions, because each one changes the trade list and leaves the live figures impossible to compare with the backtest. Log them. They are data about you.

Decide the length of the stage before it starts, as a number of trades. The count is your choice. It should be large enough that a normal losing streak fits inside it; at the 40% win rate from the previous lesson, a run of several losses in a row is ordinary, and a stage so short that one bad run ends it will tell you about the streak, never about the rules. The losing streak calculator helps set that number.

Compare live with the backtest

While the stage runs, track the same figures the backtest produced: win rate, average win and loss in R, expectancy after costs, and the cost per trade. Before you start, write down how far each one may drift before you stop.

Costs are the easiest to check first. Suppose the backtest charged $10 a trade and your live fills and commissions average $18.

That gap alone could erase a thin edge. Re-run the backtest with the real cost figure before going further.

Limits on the other numbers work the same way. A typical set might say: stop if live expectancy after the full stage is below zero, or if the live drawdown exceeds the backtest’s worst by more than a margin you chose in advance, or if you have broken the rules on more than a handful of trades. These are your settings. Write them down with the rules.

Scaling up

If the stage finishes inside the limits, step up, perhaps to half size for another set of trades, then to full. If it breaks a limit, stop and find out why before trading again. Sometimes the answer is the costs, sometimes it is you, and sometimes the rules only ever worked on the history they were fitted to, which is the case a walk-forward test is built to catch. The stock trading hub and the lesson on reading backtest results are the places to go back to when the numbers need a second look.