Free grid bot backtest
Replay a grid trading bot over the last 7, 30 or 90 days of real hourly prices on Bybit, OKX, Bitget, BingX or Hyperliquid. No sign-up and no API keys. The result sits next to simply holding the coin, so you see whether the grid earned its keep.
How the grid bot backtest works
A grid bot places buy orders at evenly spaced price levels inside a range. When the price falls to a level, the bot buys; when it rises to the next level up, it sells that piece for a small profit. The backtest replays exactly this over hourly candles from the exchange you pick: a level fills when a candle's high and low cross it, and every fill pays the fee.
In Auto mode the range is centered on the price at the start and is as wide as one standard deviation of the coin's move over the period, measured on the 30 days before the start. The backtest never uses prices it has not reached yet, so the result is not tuned with hindsight. In Manual mode you set the lower and upper price and the number of levels yourself.
Each run spends a $1,000 budget split evenly across the levels. Results are in percent of that budget, so they scale to any amount, but very small budgets can fall below an exchange's minimum order.
How to read the result
- Net result is what the grid made after fees, including the open pieces valued at the last price. Compare it with holding: in a strong rally a grid sells too early and lags; in a sideways market it usually wins.
- Max drawdown is the deepest fall of the bot's equity from its peak. A grid holds more coin as the price falls, so drawdown grows in a downtrend.
- Hours in range shows how often the price stayed between the lower and the upper level. Outside the range the bot stops trading: above it holds only quote currency, below it holds only the coin.
- Closed trades are completed buy-sell pairs. More trades means more fees, so a dense grid does not always earn more.
Choosing grid bot settings
Range width should match the coin's volatility. Too narrow, and the price leaves it in days; too wide, and each level is so far apart that the bot rarely trades. Auto mode is a reasonable starting point: one standard deviation of the expected move each side.
More levels mean smaller steps, more trades and less profit per trade. With a 0.06% fee per fill, a step below roughly 0.3% leaves little after costs. Twenty levels on a ±15% range give a step of about 1.5%.
Try the same coin over 7, 30 and 90 days. If the grid only works in one window, the settings are fitted to that window rather than to the coin.
FAQ
Is the grid bot backtest free?
Yes. It runs in the browser with no sign-up and no exchange API keys. It uses the exchanges' public price data.
Which exchanges and coins can I backtest?
Bybit, OKX, Bitget and BingX (spot and perpetuals) and Hyperliquid (perpetuals), with more than 30 popular coins including BTC, ETH, SOL, XRP, DOGE and HYPE.
How accurate is the backtest?
It uses hourly candles: a level fills when the candle crosses it. Real fills can differ inside the hour, and slippage on thin markets is not modeled. Treat the result as an estimate of how the settings behave, not a forecast.
Are fees and funding included?
Fees are: 0.06% on every buy and sell. Funding on perpetual futures is not counted; on a long grid it is usually a small cost when funding is positive.
Why does the grid lag holding in a rally?
A grid sells each piece one level higher than it bought it, so in a steady rise it ends up holding less and less coin. It earns in sideways markets, where the price crosses the same levels many times.
Can I run the bot I backtested?
Yes. Create this bot with AI opens Tradiflux, where the AI assistant sets up a grid bot with the same range and levels and shows its backtest before you confirm. You can also start it on a demo account first.