Explainer · numbers updated 17 November 2026
What Is Slippage in Crypto Trading?
Slippage is the difference between the price a trader expects when placing an order and the average price the order actually fills at. An order larger than what rests at the best price fills at progressively worse prices, and on any venue the price can move between sending an order and its execution.
On 17 November 2026, a $1 million BTC market order cost 1.17 bps in slippage on Hyperliquid, the lowest of the perp DEXs Castle measures, against 10.32 bps on dYdX and 0.78 bps on OKX futures.
- BTC $1M slippage, Hyperliquid
- 1.17 bps
- BTC $1M slippage, OKX
- 0.78 bps
- ETH $1M slippage, Aster
- 0.58 bps
- Hyperliquid BTC depth within 0.1%
- $41.2M
Castle order-book readings, the median of the past 2 days to 17 November 2026. Before fees.
How slippage happens on an order book
An order book lists resting limit orders at each price. A market order fills against the best price first and, if it is larger than the orders resting there, keeps filling at the next prices until it is done. The average of those fills against the mid price, halfway between the best bid and the best ask, is the slippage. The more value rests close to the mid price, the market depth, the less a given order slips.
Hyperliquid's docs define a market order only as one that executes immediately at the current market price (Hyperliquid docs). Its API has no market order type: the official Python SDK sends one as an immediate-or-cancel limit order 5% away from the price, which caps how far it can fill (SDK).
Worked example: a $1 million BTC order on Hyperliquid
Using Castle's reading for the median of the past 2 days to 17 November 2026, with BTC's mid price on Hyperliquid at $64,497:
- A $1 million market order filled on average 1.17 bps away from the mid, so a buy filled at about $64,504.
- That cost about $117 in slippage.
- The taker fee on Hyperliquid's base tier, 0.045% of the trade's value (Hyperliquid docs), came to $450, about 4 times the slippage at this size.
- A $100,000 order slipped 0.58 bps, so the larger order paid more on each dollar it traded.
Slippage is the average of a market buy and a market sell, in basis points from the mid price and before fees. One basis point is 0.01%.
Slippage, price impact and slippage tolerance on AMMs
Uniswap's docs separate two effects (Uniswap docs):
- Price impact: in an AMM pool the price shifts during the swap itself, so the trade fills between the starting and ending price. More liquidity in the pool means less impact.
- Slippage: the price moving while the transaction waits to be included in a block, because other trades land first.
- Slippage tolerance: the extra price change a user accepts beyond the price impact. A swap carries a minimum amount out (or a maximum amount in) and a deadline, and if the price moves past them the transaction fails and no swap happens.
In a constant-product pool (Uniswap v2) holding 1,000 ETH and 3,000,000 USDC, the price is 3,000 USDC. Buying with 30,000 USDC, before the pool fee, leaves 3,030,000 USDC and about 990.1 ETH, since the two balances multiplied must stay at 3 billion. The buyer gets about 9.9 ETH at an average of about 3,030 USDC, a 1% price impact, and leaves the pool's price at about 3,060 USDC. The same trade in a pool ten times the size moves the price about a tenth as much.
Slippage on the largest perp DEXs today
| Venue | Asset | Depth within 0.1% | $100K order | $1M order |
|---|---|---|---|---|
| OKX (CEX) | BTC | $34.7M | 0.11 bps | 0.78 bps |
| Hyperliquid | BTC | $41.2M | 0.58 bps | 1.17 bps |
| Aster | BTC | $13.1M | 0.32 bps | 1.18 bps |
| Lighter | BTC | $12.4M | 0.86 bps | 1.80 bps |
| dYdX | BTC | $1.2M | 6.29 bps | 10.32 bps |
| Aster | ETH | $13.2M | 0.07 bps | 0.58 bps |
| OKX (CEX) | ETH | $16.8M | 0.10 bps | 1.34 bps |
| Hyperliquid | ETH | $13.9M | 0.77 bps | 1.78 bps |
| Lighter | ETH | $8.6M | 1.11 bps | 2.19 bps |
| dYdX | ETH | $1.4M | 6.17 bps | 9.31 bps |
Castle reads each venue's public order book three times a day; figures are for the median of the past 2 days to 17 November 2026. OKX futures are the centralised benchmark. Fees and funding for each venue are in Castle's perp DEX rankings.
What limits slippage
- Limit orders. A limit order fills only at its price or better, so it cannot slip, but it may not fill.
- Splitting the order. Hyperliquid's TWAP orders split a trade into suborders over time, each capped at 3% slippage (Hyperliquid docs).
- Venue caps. Hyperliquid's take-profit and stop-loss market orders carry a 10% slippage tolerance (docs), and a single market order is capped at $30 million on its most liquid markets and $500,000 on the least (contract specifications).
- Deeper venues. The same order slips less where more value rests near the mid price, which is why Castle publishes depth alongside slippage.
Questions
- What is the difference between slippage and price impact?
- Uniswap's docs separate them. Price impact is the move a trade causes in the pool's price as it executes, larger for bigger trades and thinner pools. Slippage is the price moving while the transaction waits to be included in a block. On order books the two are usually counted together as the gap between the mid price and the average fill.
- What does slippage tolerance mean?
- It is the extra price change a user accepts beyond the expected price impact. On Uniswap a swap carries a minimum amount out (or a maximum amount in) and a deadline; if the price moves past them, the transaction fails and no swap happens.
- Which perp DEX has the lowest slippage?
- On Castle's reading for the median of the past 2 days to 17 November 2026, Hyperliquid: a $1 million BTC market order slipped 1.17 bps, against 1.18 bps on Aster, 1.80 bps on Lighter and 10.32 bps on dYdX. On ETH the lowest was Aster at 0.58 bps. The ranking changes day to day.
- Does slippage include trading fees?
- No. Slippage is measured against the mid price before fees, and the taker fee comes on top. On Hyperliquid's base tier the taker fee is 0.045% of the trade's value, or 4.5 basis points, charged on top of any slippage.