Lesson 7 — Impermanent loss: the math, the misconception, the cases where LPing wins
Impermanent loss is the most-misunderstood concept in DeFi. Today: the actual math, when it matters, and the structural cases where fee yield genuinely compensates.
Liquidity providing is sold to retail users as 'earn yield on your tokens.' The yield is real. So is the impermanent loss — the gap between holding the LP position and just holding the underlying tokens. Most users discover this after a price move, by which point the gap has already realized. This lesson is about understanding it before you commit capital.
**The math, plain.** Provide liquidity to a Uniswap V2 ETH/USDC pool when ETH = $2,000. You deposit 1 ETH and 2,000 USDC (total $4,000). The pool's constant-product invariant means your position automatically rebalances as the price moves: if ETH rises, the pool sells ETH for USDC; if ETH falls, the pool buys ETH with USDC. Now ETH = $4,000 (doubled). Your LP position is worth √(price ratio) × initial value × 2 / (1 + price ratio) — for a 2× price move, the position is worth about $5,657. If you had just held 1 ETH + 2,000 USDC, you would have $4,000 + $2,000 = $6,000. The gap, $343, is impermanent loss. It is roughly 5.7 percent for a 2× move. For a 4× move it's about 20 percent; for a 5× move, about 25 percent. Symmetric for downward moves.
**'Impermanent' is misleading.** The loss is impermanent only in the sense that if the price returns exactly to where you started, the loss reverses. Any time you withdraw or any time the price ends at a different point, the loss is realized. The word survives historically; the substance is closer to 'rebalancing loss' or, in academic terminology, LVR (loss-versus-rebalancing).
**When LPing wins anyway.** The fee yield from a pool can be larger than the impermanent loss, particularly in pools with stable assets (where prices don't move) or in pools with high trading volume relative to TVL. A Curve stablecoin pool earns moderate fee yield with near-zero impermanent loss because the assets are designed to trade at near-1:1. A V3 concentrated-liquidity position in a heavily-traded pair earns substantially more fees per unit of capital than V2, because all your capital is active around the current price. The question is always: does the fee yield exceed the impermanent loss across the price range I expect over the holding period?
**Concentrated liquidity changes the calculation.** V3 positions are like V2 positions but levered: capital is concentrated, fee yield is amplified, and impermanent loss is also amplified. A V3 ETH/USDC position with a tight ±10 percent range around $3,500 earns roughly 10× the fees of a V2 equivalent — but when ETH moves outside that range, the entire position has been converted to one asset and earned no fees. V3 LPing is closer to active trading than passive yield earning.
**When LPing structurally loses.** Volatile pairs with mismatched directional exposure (e.g., LP-ing a meme coin against ETH) face the worst impermanent loss because the price moves are large in one direction. Low-volume pools collect minimal fees, so even small price moves leave LPs net-negative. Pools that aren't well-arbitraged with external markets ('stale' pools) suffer most of the LVR cost without compensating fee volume.
**Loss-versus-rebalancing (LVR) in practice.** Academic work since 2023 (Milionis et al., 'Automated Market Making and Loss-Versus-Rebalancing') has quantified what LPs lose to informed traders who arbitrage the pool's stale price against external markets. The finding: for most volatile-asset pools, LPs are structurally adversely-selected, and a substantial fraction of nominal fee yield is offset by LVR. V3 concentrated liquidity makes this worse, not better, because the same LP capital is exposed to a narrower price window where it can be more efficiently picked off. The takeaway: pool-level APY numbers are often the gross yield before LVR; the net is meaningfully lower for non-stable-pair LP positions.
**The decision framework.** For an LP position, you need: (a) an expectation of the price range across the holding period, (b) an estimate of fee volume during that period, (c) honest accounting of LVR against external markets. If volume × fee rate exceeds expected impermanent loss + LVR, the position is net positive in expectation. For stablecoin pairs, this is almost always true. For volatile pairs in heavily-arbitraged markets, it is often not.
Example
A common LP trap: a user deposits $10,000 into a V2 pool of a small-cap token paired with ETH at the token's launch. The pool advertises 200 percent APY in fees. The token doubles in price over a month. The user expects to be up substantially. Their LP position, however, has experienced ~5.7 percent impermanent loss against a hodl baseline. The 200 percent APY translates to maybe 16 percent monthly in fees — significantly larger than the IL, until you account for LVR: most of the volume came from informed arbitrageurs picking off stale prices, not retail traders paying fees on round-trip swaps. The user's net result is +5–10 percent versus a baseline that would have been +50–80 percent if they'd just held the underlying tokens. The headline 200 percent APY was real; the realized return after IL and LVR is what the user actually got, and it was far worse than the alternative of not LPing.
Common mistakes
- Treating fee APY as the realized return. Gross APY less impermanent loss less LVR is the realized return; the gap is often material.
- LPing into volatile pairs hoping for both fee yield and directional appreciation. The position-rebalancing structurally sells the winning asset, so directional upside is capped.
- Ignoring LVR. For volatile pairs, LVR can account for half or more of the nominal fee yield in heavily-arbitraged markets.
- Using V3 concentrated liquidity passively. The model is closer to active trading; positions that go out of range stop earning until repositioned.
- Treating 'impermanent' literally. The loss is only impermanent if price returns to entry; for any other exit, it is realized.
Check your understanding
You provide $4,000 of liquidity to a Uniswap V2 ETH/USDC pool when ETH = $2,000 (so $2,000 ETH + $2,000 USDC). ETH then moves to $4,000. The pool has collected zero fees in this period. Compared to simply holding the original tokens, where does your LP position stand?
Key terms covered
Sources & further reading
- Primary
- PrimaryMilionis et al. — Automated Market Making and Loss-Versus-Rebalancing (2023)
Academic foundation for LVR analysis.
- Secondary
- Secondary
We prioritise primary sources. Where a topic moves quickly (regulation, security incidents), we re-check sources on the cadence shown by the page's "Next review" date.