Impermanent Loss in AMMs

Some prediction markets use AMM-style liquidity instead of a pure order book. If you provide liquidity to those systems, impermanent loss is one of the main risks to understand.

Impermanent loss happens when a liquidity provider deposits assets into an AMM pool, and the price ratio of those assets changes compared with when they were deposited. In prediction markets, where contracts settle toward fixed outcomes, that effect can become especially important as resolution gets closer.

What is Impermanent Loss?

When you provide liquidity to a standard AMM, you deposit equal values of two assets (for example, $500 of "Yes" shares and $500 of "No" shares). The AMM algorithm constantly rebalances the pool as traders buy and sell against it.

If the market suddenly shifts and values the "Yes" shares at 80% probability, arbitrage traders will buy up the "Yes" shares from your pool and leave you with the depreciating "No" shares. The "loss" represents the difference in dollar value between what your pool position is currently worth versus what your portfolio would have been worth if you had simply held the initial assets in your wallet and never provided liquidity.

Why it Matters

Automated Market Makers cannot function without liquidity providers. To encourage users to deposit capital and take inventory risk, platforms usually share some trading-fee revenue with LPs.

The core question is simple: does fee income compensate you for the inventory and pricing risk you take by staying in the pool?

How it Works in Prediction Markets

In standard crypto AMMs, impermanent loss can shrink if prices move back. In prediction markets, resolution can make the path much less forgiving.

1. The Deposit

You deposit 1000 Yes shares and 1000 No shares into a market currently priced at 50/50.

2. The Price Shift

News breaks, and the market implied probability of "Yes" moves to 90 cents. Traders buy the Yes shares from your pool. The AMM algorithm raises the price of the scarcer Yes shares and lowers the price of the more abundant No shares to rebalance the curve.

3. The Resolution

The event occurs, and the "Yes" outcome wins. The Yes shares are now worth exactly $1.00, and the No shares are worth $0.00. Because the AMM let traders buy your winning Yes shares along the way, your LP position may now be tilted toward the losing side.

When you withdraw your liquidity, the total value you pull out is significantly less than the $1.00 per share you would have banked had you just held the initial 1000 Yes shares.

Example: The Math of the Curve

Assume you deposit $1000 total liquidity into a 50/50 market. If the market slowly shifts to 80/20, your Impermanent Loss relative to holding is roughly 2.0%. If the market shifts violently to 90/10, your Impermanent Loss is roughly 4.2%.

Fee income can offset this, but only if actual trading conditions are favorable and if your assumptions about volume and inventory path are realistic.

Risks: Adverse Selection and Resolution Risk

One of the largest risks to an LP in a prediction market is resolution itself.

Unlike a stock that might recover later, a prediction market contract eventually settles to zero or one. If you hold an LP position through the full move, the relative underperformance versus simply holding can become locked in at settlement.

LPs also face adverse selection. If better-informed traders interact with the pool first, the LP can end up holding more of the wrong side of the move.

FAQ

How do I protect against Impermanent Loss?

Advanced LPs manage risk by adjusting liquidity ranges, reducing exposure before major catalysts, and monitoring whether fee income is still compensating them for inventory risk.

Do I suffer Impermanent Loss on Kalshi?

No. Kalshi uses a central limit order book, not an AMM. That means you do not face AMM-style impermanent loss there, though you still face ordinary execution and directional risk.

Can trading fees actually cover the losses?

Sometimes they can, but that depends heavily on volume, fee structure, and how the price path evolves. It should not be assumed.