The short answer
Forward testing runs a strategy on live market data inside a demo account, and it is the step that separates a strategy that works from one that only backtests well. A backtest replays history, while a forward test faces the future, and the future includes the slippage, spreads, and execution realities the historical data smoothed away.
The test's job is to confirm the edge survives contact with live conditions before real money is on it. I cover what a backtest hides, how to run a forward test that mirrors live trading, how long it should run, and what to measure on this page, and the wider system-design context is in the prop trading strategies and systems guide.
Forward testing versus backtesting
A backtest applies a strategy's rules to historical price data and produces an equity curve, and it is the fastest way to see whether an idea ever worked. Its strength is speed and its weakness is that historical data is a cleaned-up version of what actually happened, with the messy parts averaged out.
A forward test applies the same rules to live data as it prints, in a demo account that cannot lose real money, and its strength is realism. The spread you pay is today's spread, the slippage you suffer is today's slippage, and the fills you get are today's fills, none of which the backtest can reproduce faithfully.
I treat the backtest as the first filter and the forward test as the second, because a strategy that fails the backtest is not worth testing live, and a strategy that passes the backtest has not yet earned real capital. The two stages answer different questions, and skipping the second is how traders fund ideas that worked only in the past.
What a backtest hides
The list of things a backtest smooths over is the list of things that kill strategies live. Slippage is the biggest, because a backtest assumes you filled at the price on the screen, while a live market order fills wherever the order book sends it, which on a fast move can be pips worse.
Spread widening is the second, because backtests usually run on a fixed or average spread that ignores the spikes around news and the daily rollover. A strategy that looks profitable at a 1-pip average spread can lose at the 5-pip spread that actually appears when it trades.
The third is execution latency, the gap between the signal firing and the order filling, which a backtest assumes is zero and a live account never is. I add a realistic slippage and spread cost into every backtest to narrow the gap, but only a forward test closes it.
How to run a forward test properly
A forward test only validates the strategy if it mirrors the live account exactly, which means the same position size, the same instruments, the same session, and the same execution type the real account will use. A demo that runs a larger size or a different order type than the live plan tests a different strategy.
The discipline is to treat the demo as if it were real, because the moment the trader knows it is not, they take trades they would skip live and the data is corrupted. The cleanest forward test is automated, since an EA or script removes the discretionary drift that a bored human introduces into a paper account.
I keep the demo conditions identical to the funded plan down to the broker's server time, because differences that look trivial, like a different timezone for the daily reset, change the strategy's behaviour. The whole point is to test the exact thing you intend to run, not a close cousin of it.
How long to forward test before going live
The test needs enough trades to be statistically meaningful, which is a function of the strategy's frequency rather than the calendar. A strategy that takes five trades a day reaches a useful sample in a few weeks, while one that takes five trades a month needs the better part of a year.
The rough floor most systematic traders accept is a few dozen trades, because below that the result is dominated by luck rather than edge. More is always better, and the cost of the extra time is far smaller than the cost of funding an untested idea.
I run the forward test until I have enough trades to compare the live win rate and risk-reward to the backtest's, and I do not shorten the test because the demo is boring. The boredom is the point, because a strategy you cannot run patiently in a demo is one you will not run patiently with real money either.
What to measure in a forward test
The single most important metric is whether the live results match the backtest's expectations, because a divergence is the signal that the edge does not survive real execution. If the backtest promised a 45% win rate at 1:2 and the forward test delivers 38% at 1:1.5, the strategy is not the one the backtest described.
The components to track are the live win rate, the average win and loss in risk multiples, the realised drawdown, and the execution quality measured as average slippage per fill. Comparing each to its backtest counterpart isolates where the gap is, which tells you whether the problem is the edge or the execution.
I also track my own behaviour, because a forward test reveals whether I can actually follow the strategy, which a backtest cannot. The risk-reward the live test delivers, and the drawdown it produces, are the numbers that decide whether the strategy earns real capital or goes back to the drawing board.