Guide Finance 6 min read

A symmetric Uniswap v3 range holds impermanent loss at a fixed multiple while the price stays inside

Impermanent loss is what a liquidity-pool position gives up against simply holding its two assets once their prices drift apart. On a Uniswap v3 range from half to double the starting price, the same deposit loses 2 + √2 times what a full-range pool does for as long as the price stays inside.

Daniel Koh
Finance & Digital Economy Analyst
Published 15 Nov 2026, 9:22 AM (SGT)
Share:
An old brass balance scale with two pans, set on a white background. An old brass balance scale with two pans, set on a white background. Photo by Bru-nO on Pixabay
Advertisement

Impermanent loss is the shortfall of a liquidity-pool position against simply holding the same two assets after their prices move apart. Working from the Uniswap v3 whitepaper's reserve equations, we found that the same deposit on a range from half to double the starting price loses 19.5% against holding when the price reaches either edge. A full-range pool loses 5.7% on that move. The ratio is 2 + √2, about 3.41, and inside a symmetric range it does not change with the price. The 5.7% comes from the formula Uniswap's v2 documentation quotes from a 2019 article it links, 2√r/(1 + r) − 1, where r is how far one asset has moved against the other.

The loss formula is charted in Uniswap's documentation, quoted from an article it links

The formula sits on the v2 documentation page Understanding Returns, inside a quoted block that opens under the heading "Example from the article", as impermanent_loss = 2 * sqrt(price_ratio) / (1+price_ratio) - 1. An archived copy of that article is dated January 2019; we credit the documentation that quotes it.

Neither whitepaper carries it. The v2 Core whitepaper of March 2020 uses the word "impermanent" twice, both in section 2.1, with no formula; of a pool of two USD stablecoins it says providers "would generally be subject to less impermanent loss". The v3 Core whitepaper of March 2021 never uses the word.

The documentation's list of losses runs from 1.25× to 5×. We recomputed all seven entries and each rounds to the printed figure; the 5× entry is 25.464% before rounding to 25.5%.

Loss traces the same contour above and below the starting price

The documentation says a doubling and a halving cost the same, and the algebra agrees: 1/r gives the same number as r. Two assets that rise by the same percentage leave no loss at all. The quoted article explains the name by that dependence: if the prices return to their starting ratio, the gap closes. The loss is only realised if a provider withdraws while the prices are apart.

The curve climbs more gently than its first steps suggest. Per unit of price ratio, the loss stops speeding up near 2.15×. Per doubling it keeps gathering pace longer: the first doubling adds 5.72 percentage points, and the step from 4× to 8× adds 17.15, the largest among the power-of-two doublings we tested. Each doubling after that adds less. At 100× the loss is 80.2%, and it approaches 100% without reaching it.

A symmetric range redraws the map at a fixed scale for the same deposit

The v3 whitepaper gives the tokens a range position holds at each price in equations 6.29 and 6.30, built on the reserve curve of equation 2.2. Valuing them against the starting tokens, we found the gap against holding, in units of asset B, is −L(√P − √P0)²/√P0 for liquidity L and starting price P0, whatever the range. With the deposit fixed, a range from the starting price ÷k to ×k carries 1/(1 − 1/√k) times the liquidity of a full-range pool, so it loses that many times as much, in money and as a share, at every price inside.

Advertisement
RangeFactor insideEdge loss, full rangeEdge loss, this range
÷1.1 to ×1.121.490.11%2.44%
÷1.25 to ×1.259.4720.62%5.87%
½× to 2×3.4145.72%19.53%
¼× to 4×2.00020.00%40.00%
÷10 to ×101.46242.50%62.16%

The table is our arithmetic, against holding, with no fees. Over 400 prices inside each of three ranges the factor held to six decimal places, and a range stretched to 10⁻¹² and 10¹² of the start matched the full-range formula within 4.3×10⁻⁷ for moves from 0.1× to 10×. Ranges are written ÷k to ×k on purpose: 10% either side, 0.9 to 1.1, is lopsided in ratio terms and gives about 20.4× instead of 21.5×.

Beyond the border of the range the position holds a single asset and the scale falls away

Once asset A rises past the top of the range, the position has sold all of it and holds only asset B; below the bottom, it holds only A. It stops trading, so the loss keeps growing while the factor shrinks. Take the ½× to 2× range from its 19.53% at the edge to a 4× move, and the position trails holding by 51.72%, against 20.00% for a full-range pool: a factor of 2.59.

Lopsided ranges have no single scale across their width

A range set in prices, such as 80 to 120 around a start of 100, is not symmetric: 80 sits a factor of 1.25 below, 120 only 1.2 above. Across it we found the factor drifting from 10.27× at the 80 edge to 10.47× at 120, about 10.3–10.5×, with losses of 6.36% at the lower edge and 4.34% at the upper.

Loesch, Hindman, Richardson and Welch, in a 2021 paper on impermanent loss in Uniswap v3 (CC BY 4.0), describe how "this magnification also magnifies the IL" and put the maximum leverage on this same 80–120 range at "about 10×", taking the maximum loss "to 5%", between our two edges. They credit an earlier post by Lambert for the position's value, and their per-side leverage figures, about 9× and 11×, are, to their rounding, the symmetric-range factor for each side's width of 1.25 and 1.2.

This calculator surveys a full-range pool and a range side by side, with the documentation's table as a check

Until 9 November 2026 our impermanent loss calculator named no source, rounded the 5× figure to about 25%, said the loss accelerates and carried unsourced sentences about trading fees. It accepted falls below −100%: asset A at −150% against an unmoved B showed a 100% loss beside a held value of US$ 250.

The revision of that day credits its sources, prints the documentation's table beside its own (all seven rows agree) and refuses moves at or below −100%. Its defaults are asset A up 100%, asset B unchanged and a US$ 1,000 deposit, giving 5.72%: US$ 1,414.21 in the pool against US$ 1,500.00 held. Switch to the v3 range with k at 2 and the same inputs give 19.53%, US$ 1,207.11 and a factor of 3.414. Raise asset A to +300% and it reports the price outside the range, the position all asset B, with 51.72% and a factor of 2.59.

Fees, gas and tick spacing lie off the map the formula draws

All these figures are measured against a simple hold strategy and do not account for trading fees, gas costs, trade timing, or any rebalancing. Real v3 ranges snap to tick spacing, which we did not model, and on-chain balances are whole numbers in a token's smallest unit, as a separate guide on on-chain units works through.

Two stablecoins stay near the centre of the curve only while both pegs hold; a depeg is a price move, and the formula then applies in full. The v2 whitepaper states that it "does not constitute investment advice"; we offer this arithmetic on the same terms.

Advertisement
Daniel Koh
Finance & Digital Economy Analyst

Daniel Koh covers fintech, digital payments, banking technology, and ASEAN digital-economy regulation for RECATOOLS.

View author profile → · Editorial policy

About this byline Daniel Koh is a RECATOOLS editorial persona for finance and digital-economy analysis. Articles are produced and reviewed under RECATOOLS editorial supervision.

Corrections policy

Advertisement