When a track record comes from a demo account

The image is a futures position on an exchange you recognise. The pair is real, the interface is the one you use, the profit line is green, the timestamp is yesterday. Nothing in it has been edited. It is also not a record of anyone's money, because it was taken in the exchange's practice environment — and the practice environment ships the same interface on purpose.

This is not an accusation aimed at practice environments. They exist for good reasons: learning an order ticket without paying tuition, wiring up a bot before it can do damage, checking that a webhook fires. The problem is narrow and entirely mechanical. A screenshot cannot tell you which environment it came from, and the difference between the two is exactly the part of trading that costs you money.

What is a practice environment, and what does the venue call it?

The naming is inconsistent across venues, which is itself a small trap. As of September 2026, Binance documents two separate practice products for spot: a Testnet and a Demo Mode, with different endpoints and different data behaviour (Binance developer docs). Its consumer help page describes demo trading as being conducted "using mock digital protocol tokens which have no economic or monetary value" (Binance help centre). Bybit runs a Testnet and, separately, a Demo Trading service on its mainnet domain, and its API documentation is explicit that "it is an independent account for demo trading only, and it has its own user ID" (Bybit API docs). OKX calls it demo trading and states plainly that "these are virtual assets for simulated trading only and don't represent real money" (OKX help centre).

Three venues, three vocabularies, one shared property: the balance is issued by the exchange, not deposited by the trader. Everything else follows from that.

Behaviour and interfaces change, and each of these statements belongs to the venue that publishes it rather than to the industry. Binance's own USDS-margined futures documentation now gives the testnet REST base as https://demo-fapi.binance.com, a different host from the one that circulated in older tutorials — so treat any specific endpoint below as a check to re-run against the venue's current documentation, not a permanent fact.

Why can a demo fill happen at a price that never traded?

Because the price feed and the order book in a practice environment are not always the live ones, and at least one venue says so in a single sentence. Binance's developer documentation draws the line between its two products directly: "Testnet's prices and order books are independent from the live exchange," while "Demo Mode's prices and order books are similar to the live exchange." The same page adds the warning that matters most to anyone reading someone else's results: "Realistic market data is not equal to 'real' market data. Do not assume trading strategies that work in Demo Mode will work in the live exchange."

OKX states the separation more bluntly on its help centre: "The demo trading mode and live trading are completely separate — market data can't be used as a reference across the two, and price movements in demo trading mode have no relation to live market conditions."

Binance's help centre says the same thing about its own demo platform in the language of a disclaimer: "There may be discrepancies in the chart data, actual order book pricing, and trade order execution when compared to the live trading environment."

Read those together and the consequence is concrete. In a practice environment, a limit order can fill at a level the live book never printed. A stop can be honoured at exactly the stop price on a candle where the live venue would have gapped past it. A large order can be absorbed instantly by a book that owes no one an inventory. None of this requires anyone to cheat — it is a simulator behaving like a simulator. But it means a demo fill is not evidence that the same order would have filled at that price with real money, which makes every downstream number in the record — entry, exit, percentage, drawdown — an artefact of the environment rather than a measurement of the market.

Note one asymmetry that catches people out: a venue can serve real public market data into a practice environment while still simulating your fills. Bybit's demo documentation says the private streams are demo-specific but "public data is identical to that found on mainnet." So a chart in the screenshot can be genuinely live while the trade drawn on it is not.

What happens to costs and to size?

Two more things separate the two environments, and both flatter the record in the same direction.

Costs. Fees, funding and borrow are the slow leak that turns a marginally positive strategy into a negative one, and a practice environment is a narrower thing than the venue it imitates. What the documentation states plainly is that whole functions go missing: OKX's API documentation notes that in demo trading "some functions are not supported, such as withdraw, deposit, purchase/redemption, etc." (okx.com), and Bybit's says "not every API is available for demo trading account because demo trading service is mainly for trading experience purpose" (bybit-exchange.github.io). Neither statement says how fees and funding are treated inside that environment, and that silence is the point: the cost treatment behind a demo record is not something you can assume matches live. You do not have to know which cost model a given demo uses. You have to check whether the record you were shown contains a fee column and a funding column at all — because if it does not, a gross number is being presented to you as a net one.

Size and consequence. This is the deeper difference and it is rarely discussed. A demo balance is replenishable. Bybit exposes an endpoint for requesting demo funds, documented with per-request maxima of 100,000 USDT, 100,000 USDC, 15 BTC or 200 ETH. Binance's help centre says "Virtual funds can be reset separately for both Spot and Futures Accounts." A funded account has no such button. That single fact changes the behaviour being recorded: in a practice environment a wiped account is an inconvenience, so there is no cost to holding a losing position, no cost to adding to it, and no cost to carrying a size no funded trader would carry. The record you are looking at may be the record of a person who was never once afraid.

That is not a hypothetical. When we ran an automated account against live signals ourselves, the damage came from exactly this seam — the mechanics of execution and risk under real conditions, not from the calls being wrong.

What are the tells in an image or an export?

None of these is proof on its own. Together they usually settle it in a few minutes.

• The host, but only when there is one to read. A visible demo- or testnet marker in an address bar settles the question immediately. A clean address bar settles nothing, because on several venues practice is a mode inside the ordinary site at the ordinary URL: Bybit's API documentation describes it as something you "Switch to", adding that it "is an independent account for demo trading only, and it has its own user ID" (bybit-exchange.github.io). So read the address bar first, and treat a normal-looking one as a question still open rather than an answer.
• The endpoint in an API export. Hosts do separate cleanly at the API layer, and that is where an export gives you a direct answer. As of September 2026 the documented bases include https://demo-api.binance.com for Binance spot Demo Mode (developers.binance.com) and https://api-demo.bybit.com for Bybit demo trading (bybit-exchange.github.io). OKX goes further and puts the flag in the request itself: its V5 documentation requires the header x-simulated-trading: 1 for demo trading (okx.com). If someone sends you a raw response or a script alongside their record, those lines answer the question outright.
• Interface furniture. Practice environments generally label themselves somewhere — a banner, a coloured badge, a differently-tinted header, a menu item that says demo. A screenshot cropped so that exactly that strip of pixels is missing is worth a question.
• Round-number balances and step-shaped equity. An equity curve that starts at a suspiciously clean figure, or that jumps upward with no corresponding closed trade, is consistent with a top-up rather than a profit.
• Missing cost lines. No fee rows at all, and no funding rows on perpetual positions held across funding intervals.
• Symbols that are not listed on the live venue, or contract names that do not match the venue's live instrument list.
• No partial fills anywhere. A real account trading meaningful size accumulates partial executions. A history in which every order fills once, whole, at the stated price describes a book with infinite depth.
• A history that cannot go back as far as claimed. Retention in a practice environment is set per venue and can be short: Bybit's API documentation states that "Orders generated in demo trading keep 7 days" (bybit-exchange.github.io). That limit governs the queryable order record on that venue, not screenshots, which anyone can accumulate for months — so it is a check to run against an export, and a months-long export from an environment documented to keep a week has a problem with its own provenance.

This is the demo-versus-live axis specifically. The separate question of whether the arithmetic inside a profit image hangs together at all is covered in our guide to checking a P&L screenshot, and it is worth running both.

Is a demo record worth nothing?

No — and this is where the topic usually gets over-corrected.

A seller who tells you up front that a record is from a demo account is not lying to you. That is a materially different situation from one who lets you assume otherwise, and it deserves a different response. A demo record is real evidence of process: that a strategy exists and is written down, that entries and exits follow stated rules, that the person can operate the instrument they claim to trade, that a bot does what its author says it does. For a system being tested before it goes live, that is precisely what a practice environment is for.

What a demo record cannot do is stand in for the four things that determine what happens to your money: slippage, funding and fees, liquidity at the size actually traded, and the behaviour of a human being who can lose something. The regulator that governs US futures marketing writes the same limitation into the disclaimer it prescribes for simulated results — that "Unlike the results shown in an actual performance record, these results do not represent actual trading," and that simulated or hypothetical programs are "designed with the benefit of hindsight" (17 CFR 4.41). The same prescribed wording adds that because such trades were never actually executed, the results may have under- or over-compensated for the impact of market factors such as a lack of liquidity. Binance's own help centre reaches the same conclusion about its product: "Your performance on the Demo Trading Platform is not a reliable indicator of your future performance or returns when live trading."

Take a demo record as a description of a method. Do not take it as a measurement of an outcome.

What do you ask for instead?

The request that settles this politely, without accusing anyone, asks for a different form of evidence rather than more of the same:

Happy to look at the strategy — can you send a read-only export from the live account, or switch on the exchange's shared performance page? A screenshot can't tell me which environment it came from.

Read-only exports and exchange-hosted performance pages carry their environment with them; a picture does not. If the answer is that no live record exists yet, that is a usable answer too, and it tells you exactly what you are being offered. Our breakdown of the three kinds of verification covers what each form of proof can and cannot carry, and where each one stops.

If none of it is forthcoming, you still have the option that costs nothing and answers the question directly: follow the feed yourself and keep your own book. Paper-trading a channel before you pay produces a record with the one property the seller's screenshot lacks — you know which environment it came from, because you were standing in it.