What is demo trading?
Demo trading means placing orders in a broker-hosted account funded with imaginary money, against the same live price feed the broker's real clients trade. Nothing is replayed: the market moves at its normal speed and you wait for setups exactly as you would with capital at risk.
Because the plumbing is the broker's own, a demo account models several things a replay tool usually approximates. The spread is whatever the broker is quoting that second, including the widening around news and the session open. Order types behave the way that platform implements them. Margin and leverage arithmetic match the live account.
What a demo account cannot model is the part that matters most to many traders, which is what happens psychologically when the money is real. It also tends to fill more generously than a live account, because there is no actual counterparty and slippage is often simulated rather than experienced.
The structural cost is throughput. A demo account produces trades at the rate the market produces setups, which for a selective strategy on the higher timeframes can be a handful a month.
How is a trading simulator different?
A trading simulator replays stored historical bars, so you control the clock: pause, step forward one candle, or run a whole month in twenty minutes. That single difference changes what the tool is good for.
Time compression is the entire value proposition. A strategy that produces four setups a month gives you forty-eight a year on a demo account and several hundred in a weekend of replay. For learning to recognise and execute a setup, that ratio decides which tool teaches you faster.
The trade-off is that everything about execution becomes an assumption. Spread is usually a fixed number rather than the one that actually prevailed. Fills happen at the price you clicked. And the resolution of the underlying data decides whether the tool even knows what happened inside the candle where your stop sat.
There is also the contamination problem a demo account does not have: replayed history is history you may have seen, and you loaded that date range for a reason.
Which should you use, and when?
Use a simulator to learn a setup and a demo account to rehearse the platform, then move to a small live position to meet the part neither one models. The sequence matters more than the choice.
Replay first, because that is where volume of repetitions is cheap. The goal at this stage is recognition and rule clarity, not a track record. Two hundred replayed instances of one setup teaches the shape of its failures in a way that eighteen months of waiting cannot.
Demo second, and briefly, with a specific and narrow purpose: confirm that your order types, sizing arithmetic and platform hotkeys do what you think under real-time conditions. This is a rehearsal of the mechanics, not a test of the strategy, and treating it as a track record is where most of the wasted months in this sequence go.
Live third and small. The gap between demo and live is almost entirely behavioural, and the only way to measure it is to put an amount at risk that is real enough to matter and small enough not to.
Does a demo account or simulator record prove a strategy works?
Neither one produces a sample large enough or clean enough to settle whether a strategy has an edge, and for different reasons in each case.
The demo account's problem is sample size. At ten trades a month, a three-hundred-trade record takes two and a half years, and by then the strategy has usually been modified several times, which resets the count. The record is honest about execution and silent about the edge.
The simulator's problem is selection and drift, covered at length in our guide to what simulators prove. The dates were chosen, the quiet periods were skipped, and the rules moved during the session.
Both records are still worth keeping, because they answer a different question well: whether you execute your own rules consistently. That is a question about the trader, and a demo or replay log is exactly the right evidence for it.
Where does QuantParadox help with the part neither one covers?
QuantParadox exists to settle the strategy question that a demo record and a replay log both leave open, by grading fixed rules across history neither tool can cover at that volume.
Concretely: the rules you clarified in replay get described in plain English or pasted as code, then graded out-of-sample by default over a decade of minute-resolution data across thirty instruments. Minute resolution matters here specifically because it decides what the tester knows about the bar containing both your stop and your target, which is the single assumption that most often flatters a hand-run result.
Spread and slippage are modelled explicitly rather than assumed away, which is the honest counterpart to the demo account's advantage. Where the underlying coverage is thin, that is disclosed rather than quietly smoothed over.
What it does not replace: the platform rehearsal a demo account gives you, and the behavioural gap that only real money reveals. Those are genuinely outside what any historical grading can measure.