Skip to content
Allin

What Is Slippage in Crypto? Price Impact, Tolerance and What It Costs

Published 6/8/2026 · 7 min read · Finance calculators

Camille Laurent

Camille LaurentFinance writer at Allin

Tax · Personal finance

Checked against 2 sources

View profile
In short

Slippage is the difference between the price you were quoted and the price you got. On an automated market maker it has two separate parts that most pages blur together. The first is price impact, and it is pure arithmetic: a constant-product pool holding x of one token and y of the other gives you y × Δx ÷ (x + Δx) for a trade of size Δx, so your execution price is y ÷ (x + Δx) against a spot price of y ÷ x, and the impact is exactly Δx ÷ (x + Δx). Sell into a pool holding 1,000 ETH and 2,000,000 USDC and 1 ETH costs you 0.10 percent, 10 ETH costs 0.99 percent, 50 ETH costs 4.76 percent and 100 ETH costs 9.09 percent. In money, that same sequence is $2, $198, $4,762 and $18,182 — a hundred times the trade size for over nine thousand times the cost, because the loss grows with the square of the size. The second part is slippage tolerance, which is a setting in your wallet: the maximum deviation from the quote you will accept before the transaction reverts. It protects you from the price moving between quote and confirmation. It does not reduce price impact by a single basis point, because impact is a property of the pool's depth, not of your settings.

Price impact is arithmetic: on a constant-product pool it equals your trade size divided by the reserve plus your trade. Here is the formula, a table of trade size against impact, and why the setting called slippage tolerance changes none of it.

Price impact is arithmetic, and no setting removes it

A constant-product pool keeps x × y fixed. Put Δx of the first token in and the pool must hand back enough of the second to restore the product, which works out to Δy = y × Δx ÷ (x + Δx). Divide by Δx and your average price is y ÷ (x + Δx), while the marginal price before you traded was y ÷ x. The shortfall is therefore 1 − x ÷ (x + Δx) = Δx ÷ (x + Δx), which depends on one thing only: how big your trade is relative to the reserve you are trading into. Nothing about your wallet, your gas price or your patience appears in that expression.

The cost in money behaves very differently from the cost in percent. Impact is roughly proportional to trade size, so the money you lose is proportional to size squared: value × f² ÷ (1 + f), where f is your trade as a fraction of the reserve. That is why the table climbs the way it does. Going from 0.1 percent of the pool to 10 percent multiplies the trade by 100 and the cost by more than 9,000 — from $2 to $18,182. It is also why splitting one large trade into several smaller ones does not help on a single pool: the pool moves after each slice, so the slices execute at progressively worse prices and the total is very close to the one-shot cost. Splitting only helps when the slices go to different venues, or when time passes and arbitrage refills the pool between them.

Slippage tolerance is a different thing entirely

The tolerance box in a swap interface sets a minimum output: if the transaction would return less than the quote minus your tolerance, it reverts and you keep your tokens minus the gas. It exists because a blockchain transaction is not instantaneous — between the moment you sign and the moment your trade lands in a block, other people trade, and the pool you were quoted against has moved. Tolerance is protection against that delay. It is not a discount, a negotiation, or a limit on how much the pool charges you.

Both extremes cost you. Set it too tight and a trade with 5 percent price impact simply reverts against a 0.5 percent tolerance — you pay gas for nothing and try again. Set it too loose, say 20 percent on a thin pool, and you have publicly announced how much you will accept losing: a searcher can buy ahead of you, let your trade push the price up, and sell into it, pocketing the difference. That sandwich is not a bug, it is the documented economics of a public mempool, and your tolerance setting is the size of the slice. A workable habit is to read the interface's price impact estimate first, then set tolerance a little above it — enough to survive normal block-to-block movement, not enough to fund a sandwich.

What actually reduces the cost

Only two things move price impact: trading smaller, or trading against more depth. The first is under your control immediately — if a trade is 5 percent of the pool, it costs 4.76 percent, and the same money spread over three days into a pool that arbitrage keeps refilling costs a fraction of that. The second is a matter of routing. An aggregator that splits your order across several pools is genuinely reducing impact, because each pool sees a smaller f. A stablecoin-specialised curve holds far more depth near parity than a constant-product pool of the same size, which is why a large stablecoin swap on the right venue can cost single-digit basis points where a general pool would charge whole percent.

Keep the swap fee separate in your head from everything above. A 0.30 percent pool fee on a $100,000 trade is $300 and it is charged whatever the depth; the $4,762 of price impact on that same trade is fifteen times larger and comes purely from the pool's shape. Interfaces often show one and not the other, or roll them into a single estimate, which is how a trader can walk away certain they paid 0.3 percent. Run the two numbers separately before a large trade, decide whether the pool is deep enough, and if it is not, either shrink the trade or go somewhere else — those are the only levers that exist.

Sold
Price impact by trade size on a pool holding 1,000 ETH and 2,000,000 USDC, spot price $2,000
Trade size vs pool reserveSoldValue at spotReceivedPrice impactCost vs the quote
0.1 %1 ETH$2,000$1,9980.10 %$2
1 %10 ETH$20,000$19,8020.99 %$198
5 %50 ETH$100,000$95,2384.76 %$4,762
10 %100 ETH$200,000$181,8189.09 %$18,182
20 %200 ETH$400,000$333,33316.67 %$66,667

Worked with our own calculator

Slippage calculator

Given

Expected price
$200.00
Executed price
$203.00
Trade size
$10,000.00

Result

Slippage
1.5%
Cost of slippage
$150.00

These figures are produced by the calculator below, not typed in by hand — they are recomputed whenever the tool changes.

Run it on your own figures

Frequently asked questions

What slippage tolerance should I use?
Just above the price impact the interface is already showing you, with a small margin for movement between blocks. On a deep pair that is usually a few tenths of a percent; on a thin one the impact itself may be several percent and the tolerance has to clear it or the trade reverts. What you should not do is raise the tolerance until the trade goes through without checking why it was failing — a trade that needs 15 percent tolerance is telling you the pool is too small for it.
Is slippage the same as the trading fee?
No, and they can differ by an order of magnitude. The fee is a fixed percentage the pool takes — 0.30 percent on a classic pair, so $300 on a $100,000 trade. Price impact comes from moving the pool's price against yourself and depends entirely on depth: on the pool in the table, that same $100,000 trade suffers $4,762 of impact. On a small trade the fee dominates; on a large one the impact does, and by a wide margin.
Does splitting a big trade into smaller ones help?
Not within the same block on the same pool: each slice moves the price for the next one, so the slices execute progressively worse and the total lands very close to the single-trade cost, with more gas paid. It does help when the slices reach different pools — which is what an aggregator does for you — or when they are spread over hours or days, because arbitrage pushes the pool back toward the wider market price in between.

Articles you may find interesting

All guides
ComparisonCrypto Market Cap vs Trading Volume: What Each Number Can and Cannot Tell YouMarket cap is price times circulating supply — an arithmetic product, not money invested. Volume is what actually changed hands. Here is what each measures, how their ratio exposes a thin market, and where fully diluted valuation fits.ExplainerImpermanent Loss Explained: What Providing Liquidity Really CostsImpermanent loss is not a fee and it is not temporary: it is the gap between your liquidity position and simply having held the two tokens. Here is the formula, a table of price change against loss, and the fee income you would need to come out ahead.ExplainerHow Staking Rewards Actually Work: Nominal Rate, Compounding, and What Eats ItAn advertised 8 percent becomes 8.33 percent once daily rewards compound — and then 7.46 percent after a 10 percent validator commission, and less again after unbonding time. Here is each step, with the arithmetic laid out.ExplainerVolatility Is Not Risk, and the Square Root of Time Is a ChoiceAnnualised volatility = period standard deviation × √(periods per year), and that √t scaling assumes independent increments. It is a model, not arithmetic: at a daily autocorrelation of 0.1 a 60 percent annualised figure should read 66.3. The payload is volatility drag — the arithmetic mean exceeds the geometric by about σ²/2, so at 8 percent average return and 40 percent volatility the compound outcome is zero.GuideYield Farming: What an Advertised APY Actually PaysThe number on the farm's front page is a gross figure before every subtraction. Here is a 120 percent APY walked down, one layer at a time, to the 33.6 percent that actually landed — plus why APR and APY are not the same number.ExplainerDollar-Cost Averaging: What It Actually Buys YouSpending a fixed amount each period buys more units when the price is low, so your average cost is the harmonic mean of the prices while the average price is the arithmetic mean — always lower, by 4.10 percent on the path worked through here. Against a lump sum, a 200,000-path simulation puts DCA's standard deviation 41 percent lower and its expected terminal wealth $337 lower on $12,000.

Related tools

This article explains how a calculation works. It is not investment advice. Leveraged trading and mining can lose more than you put in, and past results say nothing about future ones.

Sources

Spotted a mistake in this article?