Guide · 6 minutes
What is a grid bot, and when does it work
A grid bot buys and sells on its own, at prices set in advance, while the price moves up and down. It makes money when the market goes sideways or moves the way you chose, and loses when it breaks hard the other way. Here's how it works, what you decide, and a real example with 100 Bitcoin bots from the past year.
The idea, in one picture
That's the whole trick: it doesn't try to guess where the price is going. It spreads buy orders below the current price and sell orders above it, and keeps refilling them. The more the price zigzags inside the range, the more cycles it completes and the more it earns.
What you decide
Range
The lowest and highest price it trades between. If the price leaves the range, the bot sits still until it comes back.
Number of grids
How many levels fit inside the range. More grids: more frequent cycles, each one earning less, and more fees.
Direction
On spot the bot starts partly in coins. On futures you pick long (earns more if it rises), short (if it falls) or neutral (neither).
Leverage
Futures only. It multiplies what the cycles earn, and the losses too. If the price moves far enough against you, the exchange liquidates the bot and you lose what you put in.
Where it wins and where it loses
A grid bot's result has two parts: the cycle profit, which always adds up, and the value of the position the bot holds at any moment, which can take away much more.
Works well
When the price goes back and forth inside the range for days: every swing is a cycle banked. And when the trend goes the bot’s way: a long grid in a rally, or a short one in a drop, earns from the cycles and from the move as well.
Works badly
When the price runs against the bot and doesn’t come back. If it falls below the range, you’re left holding coins bought higher (unless it’s a short). If it rises above, it depends: on spot or a long grid you don’t lose, because you already sold at a profit, though you earn less than if you had just held the coin; a neutral or short futures grid, on the other hand, has been selling into the rally and loses.
That's why a grid bot isn't easy money. Cycles give a steady trickle; a strong trend against you can take it all at once. And in a strong rally, simply buying and holding usually beats the bot.
A stop loss just outside the range, on the side that goes against you (and a take profit on the other), cuts the fall before it can liquidate you. In our simulations (3,561 thirty-day bots) it removed liquidations without hurting the average result, although it sometimes closes the bot on a spike that then reverses. The oracle gives you the exact price.
A real example: 100 BTC bots
With the simulator, we launched the same bot at 100 different moments between November 2025 and October 2026: BTC futures, neutral, range ±10% around the entry price, 40 grids, 3x leverage, 30 days. With fees, funding and liquidation, minute by minute.
ended in profit
typical (median) result over 30 days
the best: 7 Feb 2026, 496 cycles, the price never left the range
the worst: 18 Jan 2026, BTC fell 29% and spent almost two thirds of the time outside the range
None were liquidated, and in 61 of the 100 the bot did better than buying BTC and holding for those same 30 days. But notice the asymmetry: the best case gained half of what the worst one lost. What decides the result isn't so much the number of grids as whether the price stays inside the range.
In our validation across 23 pairs and six years of data, a neutral bot with its range sized to each coin's volatility ended in profit between 65% and 79% of the time, depending on the pair. The most volatile coins, such as PEPE or NEAR, have much harsher worst cases. Details are on the validation page.
Common beginner mistakes
- A range that's too narrow. Lots of cycles at first, but the price soon leaves it and the bot stops with a losing position.
- Too many grids. If the gap between levels is tiny, fees eat a big share of every cycle.
- High leverage. At 10x or more, an ordinary week's move can liquidate the bot.
- Trusting the annual return the exchange shows. It usually comes from the last few days, or from bots that did well. It says nothing about a bot you launch today.
- Going long right when everyone is already long (or short when everyone is selling). That's when the price is most likely to turn against you.
How GridReplay helps
GridReplay is independent of any exchange. Use it to decide before you launch, and to understand afterwards what happened.
Oracle →
Every day, which direction the market favours (long, short or neutral), with what range and how many grids. BTC and ETH, free.
Simulator →
Test your setup at 100 moments over the past year: how often it wins, the typical result and the worst cases.
Analyzer →
Upload your Pionex bot history and see how much came from cycles and how much from the position. It runs in your browser.
Where do you create a grid bot?
The big exchanges offer them ready-made: pick a pair, range, grids and amount, and the bot runs on its own. Pionex was built around bots, and it's the one we used to validate the simulator against real bots.
Create a grid bot on PionexAffiliate link: if you sign up through it, GridReplay gets a commission from Pionex at no cost to you. Start small and without leverage until you understand how it behaves.
FAQ
Does a grid bot always win?
No. It wins in a sideways market and loses when the price moves far from the range and doesn't come back. In the example above, 42 of 100 bots ended at a loss.
Spot or futures?
On spot there's no liquidation and the risk is that of holding the coin. On futures you can choose direction and leverage, but a strong move can liquidate the bot. To start, use spot or futures without leverage.
How many grids should I use?
Enough for frequent cycles, but with a gap between levels clearly larger than the round-trip fees. The oracle suggests a number for each pair based on its volatility, and the simulator lets you compare several.
When should I stop a bot?
Once the price leaves the range and doesn't look like coming back, the bot stops cycling and all you have left is the position. Decide in advance at what price you'd close it (the stop) so you don't improvise.