Skip to content
Candle candle docs

Slippage

What slippage is, how Candle sets it, and why a loose limit is the real exposure. Not investment advice.

Two different things

Section titled “Two different things”
  • Price impact is how far your own trade moves the price. Bigger trades against thin liquidity move it more. It is in the quote you see before signing.
  • Slippage is how far the price you get can fall short of the quote between quote and execution. Your slippage tolerance is the limit you set. The trade carries a minimum output (minOutRaw), computed as the quoted output times (10000 − maxSlippageBps) / 10000. If the fill would be below it, the transaction reverts instead of filling.

What Candle does

Section titled “What Candle does”
  • The agent API takes maxSlippageBps from 0 to 10,000 and defaults to 100 (1%) (Agent trading).
  • A Candle-launched Hood token that has graduated to its Candle v4 pool is also held to min(maxSlippageBps, 300) bps of that pool’s output, so a route on another pool cannot fill far from it.
  • After a fill, slippageBps on the trade reports how far the fill came in below the quote taken at build time.

Why a loose limit matters

Section titled “Why a loose limit matters”

Your tolerance is the most a bad fill can cost you. A tolerance of 25% means the trade will still land if the fill is 25% worse than quoted, whether because the price moved, a pool was thin, or someone traded around you (see Priority fees and MEV). A tight limit makes the trade fail instead. The trade-off: too tight on a fast-moving token and it simply reverts.

Candle’s own check of 30 days of Solana agent trades (2026-10-25): 57% of trades allowed 5% to 25% slippage, mostly small ones, while larger trades mostly used 100 bps. Across the 1,720 recorded Solana fills in the source’s full sample, the reported median shortfall against the quote was 0 bps and the 10th percentile was 74 bps worse than the quote. These statistics have known measurement defects: tiny sells under 0.02 SOL included rent or fee artifacts in the recorded output, even showing apparent fills below the on-chain minimum; separately, 507 of 1,576 rows had recorded output exactly equal to their quote, so some outputs were not truly measured. The source (mev-evidence.md, §3, Candle’s data team, 2026-10-25) reports those caveats alongside the statistics; these figures do not establish the distribution of actual execution slippage.

Practical points

Section titled “Practical points”
  • Keep the default unless you have a reason; raise it for one trade, not as a standing setting.
  • Small trades in thin tokens are the usual case for wide tolerances. Check price impact first.
  • A reverted trade for slippage cost you the network fee, not the trade.