Paper trading runs your trades, or your rules, on live prices with pretend money. How Hyperliquid’s testnet, the exchanges’ demo modes and TradingView’s simulator work, how far a testnet market drifts from the real one, what paper can’t show you, and a 30-day plan before real money.
November 16, 2026
18
min to read
The short answer
Paper trading crypto means trading with pretend money before you risk real money. Practice decisions in an exchange demo account or TradingView’s Paper Trading, which follow live prices, and test code on Hyperliquid’s testnet, whose faucet gives 1,000 mock USDC to addresses that have deposited on mainnet. Set pass marks first and log every trade for about 30 days.
Paper trading means placing trades, or running a trading rule, with money that isn’t real, so you can see how they would have gone before any money moves. Exchanges call it demo trading, developers call it a testnet or a sandbox, and in the life of a strategy it is the forward test: the step after a backtest, where the rule meets prices it has never seen, one at a time, as they happen.
In crypto it comes in five forms. They look alike from the outside, and they are not the same test. What sets them apart is where the prices come from.
| Form | Where the prices come from | What it tests well | Examples |
|---|---|---|---|
| Exchange demo account | The exchange’s live market data, or data built to look like it | Your decisions, on the screens and order types you will use | Binance Demo Trading, Bybit Demo Trading, OKX Demo Trading |
| Testnet | A separate network with its own order books and its own traders | Your code: keys, order formats, sizes, stops and error handling | Hyperliquid’s testnet, Binance’s Spot Test Network |
| Chart simulator | The chart’s price feed | Entries, exits and stops placed by hand | TradingView Paper Trading |
| Paper trading API | Live prices, with fills the provider simulates | A bot’s logic before it sends real orders | Alpaca’s paper API, Kraken’s command-line paper engine |
| Paper mode in a trading tool | Live prices | An automated rule before it trades real money | Ethy’s paper mode |
A demo that copies live prices tells you how your decisions would have played out in the real market, minus the real fills. A testnet tells you whether your code works and very little about whether your strategy does, because its market is a different market. The next section measures how different.
Hyperliquid runs a testnet: a second copy of the exchange where the money is mock USDC. It has its own app and its own API. Hyperliquid’s API docs say you “can make the same requests against testnet” by swapping the mainnet address, https://api.hyperliquid.xyz, for https://api.hyperliquid-testnet.xyz.
The testnet faucet only serves addresses with a deposit on the real exchange: “To use the testnet faucet, you need to have deposited on mainnet with the same address.”
The faucet page sits on the testnet app, at app.hyperliquid-testnet.xyz/drip, and gives 1,000 mock USDC. Type the address yourself from Hyperliquid’s docs; never follow a link from an ad or a message.
Privy, the login service behind email accounts, creates a different wallet address on testnet. The docs suggest exporting your email wallet from mainnet, importing it into a wallet extension such as Rabby or MetaMask, and connecting that wallet to testnet.
Same requests, different base URL. Read market ids from the testnet’s own meta response, because the asset ids differ: on November 15, 2026, BTC was asset 3 on testnet and asset 0 on mainnet.
Place and cancel orders, attach a stop loss, change leverage, check rounding and minimum sizes, and see what your code does when an order is rejected or the connection drops.
Very. We read both networks through Hyperliquid’s public API on November 15, 2026, at 12:57 UTC, one right after the other:
| Market | Network | Price against its oracle | Funding per hour | 24-hour volume | Offered in the 20 ask levels shown | A $1,000 market buy pays over the mid |
|---|---|---|---|---|---|---|
| BTC | Mainnet | −0.02% | 0.00125% | $7.91 billion | $4.89 million | 0.001% |
| BTC | Testnet | −2.66% | −0.08721% | $739,820 | $1,506 | 0.31% |
| ETH | Mainnet | −0.07% | −0.00151% | $2.40 billion | $7.46 million | 0.003% |
| ETH | Testnet | −1.07% | −0.08719% | $647,171 | $4,731 | 0.08% |
| SOL | Mainnet | −0.05% | 0.00030% | $678.05 million | $250,401 | 0.001% |
| SOL | Testnet | −10.33% | −0.39841% | $54,389 | $638 | More than the 20 levels hold |
Read it slowly. The testnet’s oracle, the reference price behind funding and margin, followed real prices: BTC’s stood at $65,484.1 on testnet and $65,484.0 on mainnet. Testnet traders weren’t pinned to it. Testnet BTC traded 2.7% under its oracle and SOL 10.3% under, so testnet funding, which exists to pull a perp back to its oracle, ran at −0.0872% an hour on BTC: shorts were paying longs about 2.1% of the position a day. On mainnet, longs on a perp that sits near spot pay 0.00125% an hour. Testnet BTC traded $739,820 in 24 hours; mainnet BTC traded $7.91 billion, about 10,700 times as much. The faucet’s 1,000 mock USDC, spent on one BTC market buy, would have paid 0.31% over the mid price on testnet and 0.001% on mainnet.
The rules differ too. Hyperliquid’s margin tiers page says “The tiers on testnet are lower than mainnet, for ease of testing”: BTC allows 40x leverage up to $10,000 of position on testnet, and up to $150 million on mainnet. On mainnet, when a position over $100,000 is liquidated, the first market order covers only 20% of it; on testnet that threshold is $10,000, “for easier testing” (liquidations). Even the lineup differs: testnet listed 280 HIP-3 exchanges, the perp exchanges outside teams deploy on Hyperliquid, against 10 on mainnet that day, and 159 live perps on its main exchange against 179.
If you write code for Hyperliquid, our Hyperliquid API guide covers the requests, the signing and the API wallet you would test here. If the exchange itself is new to you, start with how to use Hyperliquid.
Binance, Bybit, OKX and Kraken each run a demo environment. They differ in whether the prices are live, how the play money works and what the API can do. Here is what each exchange’s own developer documentation said on November 15, 2026:
| Exchange | What it is | Prices | Good to know |
|---|---|---|---|
| Binance | Demo Trading, opened from your Binance account at demo.binance.com, with its own API keys | Market data similar to the live exchange’s | Reset the balance whenever you like; the same exchange limits as live |
| Binance Spot Test Network | A developer testnet at testnet.binance.vision; you log in with GitHub | Its own prices and order books, independent of the live exchange | Reset to a blank state about once a month; balances can’t move in or out |
| Bybit | Demo Trading, “an independent account for demo trading only” with its own user ID | Public market data identical to mainnet | Request up to 100,000 demo USDT at a time, once a minute; demo orders are kept for 7 days |
| OKX | Demo Trading, opened from Trade in your OKX account, with its own API keys | Not described in its API docs | API calls carry the header x-simulated-trading: 1; deposits and withdrawals don’t work in demo |
| Kraken | A Futures demo environment at demo-futures.kraken.com, separate from production | Not described; the demo API matches production in feeds, endpoints and response structure | No existing Kraken account needed; Kraken’s command-line tool also has a paper engine that runs on live prices |
Binance’s docs add the warning that belongs on every demo screen: “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.” Binance also points out what its testnet is for: new features sometimes reach it first, while Demo Mode keeps the live exchange’s features.
Before you start on any of them, find out two things. Does the demo charge trading costs, and perps funding, the way the live account does? And do your demo orders fill against a real order book, or against a simulation of one? A demo that skips either flatters every result.
TradingView’s Paper Trading is a simulated broker built into its charts. TradingView describes it as “a tool that lets you practice buying and selling assets using simulated money,” and says the simulator covers “almost every asset class available on TradingView,” crypto included. It is a simple way to practice entries, exits and stops by hand on the same chart you analyze.
Open Supercharts, click Trade in the top-right corner or open the Trading panel at the bottom, choose Paper Trading and press Connect.
Check the commission settings in the account settings. TradingView’s help page describes a per-contract commission for futures and options, which “is inactive by default,” and a fixed or percentage type for other instruments, crypto included. Leverage comes preset per asset class at values “commonly used by real brokers and exchanges,” and you can change it.
From the order ticket (market, limit or stop, with a stop loss and a take profit), from the depth-of-market window, or by dragging orders on the chart.
The Account Manager tab at the bottom of the chart shows how much of your balance is tied up as margin. An order that needs more margin than you have available isn’t executed.
A reset deletes the transaction history, active orders and open positions, so note your results first.
TradingView also has a strategy tester for backtests on the same charts. Paper Trading is the forward test that follows it.
All three, in that order, because each one answers a question the others can’t.
| Backtest | Paper trading | Small live account | |
|---|---|---|---|
| Runs on | Past prices, all at once | Live prices as they arrive, with pretend money | Live prices, with real money at a small size |
| Answers | Would the rule have worked on the past? | Does it behave the same on prices it has never seen? | Do real fills, real costs and your own nerves match the plan? |
| Takes | Minutes | Weeks to months | Weeks to months |
| Costs you | Time | Time | Real losses, kept small on purpose |
| Misses | Bugs that peek at future data, and markets unlike the past | Fills, slippage, queue position and how a real loss feels | Rare events that haven’t happened yet |
| Its trap | Tuning until the past looks perfect | Treating pretend profit as proof | Sizing up after one good week |
A backtest covers years of history in minutes, so it is the cheap filter for bad ideas; our guide on how to backtest a trading strategy covers the method and the traps. Paper trading has no future data to leak, which makes it the check on everything a backtest can hide: a rule that fires at a different moment live, a signal computed on a candle that hadn’t closed, an order your exchange rejects. A small live account is the only test of what real money adds. Our guide to quantitative trading shows a rule that did well on past data and then failed on the months that followed.
Some rules can only start at the second step. A rule that acts on news, or on whatever a scanner finds that day, has no past to replay. Paper trading is its first test.
A paper trade fills at a price the simulator chose. Everything that happens between a real order and its price is missing. Alpaca, whose trading API comes with a paper environment that “simulates crypto trading as well,” is unusually specific about what its own simulation does not account for: market impact, information leakage, price slippage due to latency, order queue position, price improvement, regulatory fees and dividends.
US commodity regulation makes professionals say this out loud. Under CFTC Rule 4.41, a commodity pool operator or commodity trading advisor that presents simulated or hypothetical performance must attach a warning, and the one the rule itself sets out begins: “These results are based on simulated or hypothetical performance results that have certain inherent limitations.” It goes on to say that because the trades were never executed, the results “may have under-or over-compensated for the impact, if any, of certain market factors, such as lack of liquidity,” and that simulated programs “are designed with the benefit of hindsight.” Read every paper result, yours included, with that statement in mind.
Thirty days is long enough to see a rule on hourly or 4-hour bars meet more than one kind of market, and short enough to keep the discipline. A rule on daily bars may trade only a handful of times in a month: give it longer, or judge the month on whether its signals matched the backtest rather than on its profit.
The exact rule, the market, the timeframe, the size and the stop. Then the numbers that will decide the test, set before it starts: for example, at least 10 trades, a drawdown no worse than the backtest’s worst, and trades that match the backtest’s on the same dates. A test without pass marks ends in whatever you hoped to see.
Use the size, leverage and stop you would use live, not the demo’s large balance. Habits formed at the wrong size carry over.
Set the commission. If the simulator ignores slippage, add it yourself: note what the live order book would have charged your size at the moment of each fill. On perps you hold for days, add the funding the live market charged.
The time, the signal, the paper fill, the live price at that moment, and the reason the rule fired. You will compare the log, not the balance.
Run the backtest over the same dates and match the trades one by one. A trade that one shows and the other doesn’t is a bug, a data difference or a timing problem, and it is cheapest to find now.
Score the month against the marks you wrote down. A pass earns a live start at a small size, smaller than the paper size, with the log still running. A fail earns a reason before any change. Too few trades earns more time, not a looser rule.
The first real trades show what paper hid. If real fills land well away from the paper ones, measure the gap before you add size.
Here is my trading rule: [describe it]. I will paper trade it for 30 days on [venue], on [timeframe] bars, at a size of [size]. Write me a test plan: the pass marks to set before day 1 (minimum number of trades, worst acceptable drawdown, how closely paper trades must match the backtest), the fields to log for every trade, the costs I must add by hand (commission, slippage, funding), and the checks to run each week. Then list anything about my rule that a 30-day paper test can’t judge.
Ethy is an AI trading agent you ask before you trade, and paper mode is how you run one of its automations without money. You describe a rule in plain English, Ethy shows the rules it will run, and you can backtest that exact version first. When you press Activate, the automation card has a Paper mode switch. Live is the default, so switch it on.
Paper mode is a switch on Ethy’s automations, which trade Hyperliquid perps, including the stocks, indices and commodities on trade.xyz, and spot tokens. Whatever you test with, Ethy’s paper mode included, the limits in this guide still apply: fills, slippage and nerves are what a small live start is for. Ethy is software for research and analysis. It doesn’t give investment advice.
Mostly. TradingView says its Paper Trading has “no fees or catches,” and exchange demo accounts run on play money they hand out: Binance lets you reset your demo balance whenever you like, and Bybit gives up to 100,000 demo USDT per request. Hyperliquid’s testnet faucet is free, but it only serves addresses that have deposited on mainnet. On Ethy, paper automations use credits at half the live price, and the free plan includes two paper automations that run for 30 days.
Deposit on Hyperliquid’s mainnet first, then claim from the faucet on the testnet app, which gives 1,000 mock USDC to addresses that have deposited on mainnet with the same address. If you log in with email, your testnet address is different: export your email wallet, import it into a wallet extension such as Rabby or MetaMask, and connect that wallet instead.
No. It uses mock USDC and has its own traders and order books. On November 15, 2026, testnet BTC traded 2.7% under its oracle price, its funding ran at −0.0872% an hour, and its 24-hour volume was $739,820 against $7.91 billion on mainnet. Its margin tiers are lower too. Use it to test code, not results.
No. A backtest replays a rule on past prices in minutes. Paper trading runs it forward on live prices as they arrive, so it can’t use data from the future and shows how the rule behaves in today’s market. Backtest first to filter ideas, then paper trade the ones that survive.
Until you have enough trades to compare with the backtest and have seen more than one kind of market. For a rule on hourly or 4-hour bars, 30 days is a sensible minimum; a rule on daily bars may need several months. Fix the length and the pass marks before you start, not after.
No. Paper fills skip slippage, market impact and queue position, and nobody feels a pretend loss. Binance’s own demo docs warn: “Do not assume trading strategies that work in Demo Mode will work in the live exchange.” Treat paper profit as a reason to try a small live size, not as proof.
Yes. Binance’s futures API has a demo environment for its USDⓈ-M futures, TradingView’s Paper Trading supports crypto futures, Hyperliquid’s testnet runs perps on mock USDC, and Ethy’s paper mode runs automations on Hyperliquid perps. Check whether each one charges funding, which live perps pay or collect every hour.
Trading involves risk, including the loss of your money. Ethy AI is a research and trading tool and does not give investment advice or recommendations. Backtests are simulations on historical data and do not predict future results.
Deploy your AI trading agent in one click.
