A grid bot does one thing: it places a ladder of buy orders below the current price and sell orders above it, and every time price oscillates between two rungs it books a small profit. In a market that goes nowhere, it prints steadily. That is a genuinely useful property, because most markets go nowhere most of the time.
The reason grid bots have a mixed reputation is that their failure mode is severe and arrives exactly when the market gets interesting.
The mechanics
You define three things: an upper price, a lower price, and how many grid levels sit between them.
The bot divides the range into that many levels and places orders at each. As price falls through a level, a buy fills. As price rises back through the next level up, a sell fills and the pair of trades books the spacing between them, minus fees.
Nothing about this predicts direction. The bot is monetising oscillation, not forecasting.
Two spacing conventions exist. Arithmetic grids place levels an equal number of price units apart, which suits assets trading in a narrow band. Geometric grids place them an equal percentage apart, which is better for volatile assets and for wide ranges, since a fixed dollar spacing behaves very differently at 20,000 than at 80,000.
Choosing the range
This is the decision that determines whether the bot works, and most people make it badly by drawing the range around recent price.
A better approach starts with what the asset has actually done. Look at the past several months and identify the region price has been rotating within. Set the grid boundaries at levels that have been tested and held more than once, not at round numbers or at the current high and low.
Wider ranges survive more, but each grid level captures less if you keep the level count constant. Narrower ranges produce more fills and break sooner. There is no correct answer, only a trade-off you should make deliberately.
Choosing the spacing
Grid spacing has to clear costs. If your spacing is 0.15 percent and you pay 0.1 percent per side, every completed pair nets close to nothing.
A workable rule: spacing should be at least four times your round-trip fee. With 0.1 percent taker fees each way, that means spacing no tighter than 0.8 percent. If your exchange offers maker rebates and your orders rest on the book, the calculation improves considerably, which is why fee tier matters more for grid bots than for most strategies.
Beyond costs, match spacing to volatility. An asset with 2 percent daily range and 0.5 percent grid spacing will fill constantly. The same spacing on an asset with 10 percent daily range means you skip several levels per move and capture less than you could.
ATR gives a reasonable starting point: spacing around 10 to 20 percent of the average true range over your chosen timeframe.
How grid bots lose money
Three ways, and it is worth understanding each before running one.
1. Price leaves the range and does not come back
This is the big one. If price drops below your lower boundary, every buy order has filled and the bot is holding the full position at an average price inside the range. It stops trading and simply holds a losing position.
The loss is not hypothetical. A grid that ran a range from 100 to 80, fully filled, and then watched price go to 50, is holding at an average near 90 with a 45 percent unrealised loss. The small profits collected on the way down do not come close to covering it.
People describe this as the bot failing. It is not. The bot did exactly what it was told. The range was wrong.
2. Price leaves the range to the upside
Less painful but still a loss relative to holding. All sell orders fill and you are left in stablecoins while the asset keeps going. You made your grid profits and missed the move. For a bot on an asset you wanted exposure to, this is a real opportunity cost.
3. Costs quietly exceed the edge
Tight spacing on a high-fee venue produces a large number of trades that each net almost nothing. The equity curve drifts down while the trade count looks impressive.
Spot grids versus futures grids
A spot grid can only lose what you put in, and in the worst case you hold the asset. That is recoverable if you wanted the asset anyway.
A futures grid adds leverage, and leverage converts the range-break scenario from a drawdown into a liquidation. The bot fills every buy on the way down while your margin depletes, and the position closes at the worst possible point.
Futures grids have a legitimate use for traders who understand the risk and size accordingly. They are not a higher-return version of the same product. They are a different product with a terminal failure mode.
When a grid bot is the right tool
Grid bots suit specific conditions:
- An asset in a well-established range, ideally one that has held for months
- Low or falling volatility, where large directional moves seem unlikely
- An asset you are willing to hold if the range breaks downward
- A venue where your fees are low enough that spacing can be tight
They are the wrong tool during trending markets, around scheduled catalysts, and on assets you have no interest in owning.
Stablecoin pairs and large-cap assets during quiet periods are where most successful grid operation happens. Small-cap tokens are where most of the disaster stories come from, because their ranges break permanently.
Managing a running grid
A grid bot is not a set-and-forget product, despite being marketed as one.
Set a stop condition before you start. Decide in advance what price below your lower boundary means the thesis is wrong, and either close the bot there or accept the holding. Writing it down beforehand prevents the much worse decision of doing nothing while it gets worse.
Review the range monthly. Market structure changes, and a range that was valid in March may be irrelevant by June.
Track profit honestly. Many bot interfaces display realised grid profit without netting the unrealised loss on the inventory. A bot showing 4 percent grid profit while holding a 12 percent unrealised loss is down 8 percent, and the display is not going to tell you that.
A realistic expectation
In a genuinely range-bound market with sensible spacing and low fees, a grid bot can produce a steady return with limited attention. That is a real and useful thing.
What it is not is a yield product, and any presentation that quotes an annualised return from a few weeks of sideways price action is extrapolating from the most favourable possible conditions. The correct way to judge a grid bot is to include the periods where the range broke, because those are part of the strategy, not exceptions to it.