Slippage
The gap between the price you saw when you sent a trade and the price it actually filled at. It shows up as a slightly worse rate than the one on screen, and it grows the thinner the market is and the faster the price moves.
How it works
An order book is a queue of offers at different prices. A market order eats through them in turn, starting with the best. If your size is bigger than what sits at the best price, the remainder fills at worse ones, and your real price becomes the average across all those pieces.
The same thing happens on a DEX, except against a liquidity pool: every unit you buy shifts the pool's ratio, so the price walks away from you while the trade is still going through.
Two things set the size of it: how much liquidity sits near the current price, and how far the price travels in the seconds between your click and the fill.
A worked example
You buy 5 ETH with $2,400 showing on screen. The book holds 1.5 ETH at $2,400, then 2 ETH at $2,403, then 1.5 ETH at $2,409.
| 1.5 × $2,400 | $3,600.00 |
| 2 × $2,403 | $4,806.00 |
| 1.5 × $2,409 | $3,613.50 |
| Total for 5 ETH | $12,019.50 |
The average came out at $2,403.90 instead of the $2,400 you expected. Slippage cost $19.50, or 0.16%. On a liquid pair that is normal. On a thin altcoin the same order can walk 3–5% up the book, and instead of $12,000 you pay around $12,500.
How to keep it small
- Cap your toleranceExchanges and DEXs both let you set a maximum: the trade reverts instead of filling at a price you never agreed to. For major pairs 0.1–0.5% is normal. A pair that needs 5% or more is telling you about the pair, not about the setting you should raise.
- Use a limit orderIt names your price and waits. It may go unfilled, but it cannot fill worse.
- Split a large orderSeveral smaller fills spread out cost less than one order eating the whole book.
- Check depth before sizeThe order book, or the pool's TVL, tells you how much you can take before the price moves.
- Wait out the newsIn the first minutes spreads widen and depth disappears. A high tolerance on a thin token is exactly what sandwich bots look for.