Skip to main content
What LPs can expect.

How LPs Make Money

1. Trading Fees

Every trade on Skepsis pays a fee between 0.3% and 1%, depending on the market’s risk category. A portion flows to the Vault.
Daily trading volume: $500,000
Fee rate: 0.3%-1% (avg ~0.5%)
Daily fees: ~$2,500
LP share of fees: ~$1,000

On a $100,000 Vault position, that's 1%/day
(Hypothetical. Actual volume varies.)

2. Market Maker Spread

LMSR has a built-in spread that benefits LPs, but it’s not what you might think. The marginal prices across all outcomes always sum to exactly 1.0. That’s a core property. The spread comes from convexity: when a trader buys a chunk of shares, the price moves up as they buy. The average cost per share ends up higher than the starting price.
Trader sees: $96K-$97K at 25% probability
Trader buys $500 worth of shares
Price moves from 25% → 28% during execution

Average cost per share: higher than 25%
That difference = the spread = LP's edge
The bigger the trade relative to alpha, the more slippage. This is the pool’s natural defense, and it’s what keeps LPs solvent.

3. Losing Trader Capital

When traders bet wrong, their capital stays in the pool. After winners are paid, the remainder goes back to the Vault.
Market pool: $15,000
Winners claim: $8,000
Remaining: $7,000 → Returns to Vault

If the Vault deployed $10,000, it lost $3,000 on this market.
But across many markets, winners and losers balance out.

How LPs Lose Money

The Winning Range Problem

If a very popular range wins, the market owes a lot of USDC. The Vault covers the gap.
Everyone bets on $96K-$97K
BTC lands at $96,500
Many winners, many payouts

Pool can't cover it all from trader bets alone
Vault-deployed capital covers the rest.
LP loss on this market.

The Bound

The maximum LP loss per market is mathematically capped.
Max LP loss = α × ln(bucketCount)

Typical values:
- α = 3,333 USDC, 16 buckets → Max loss ≈ $9,240
- α = 1,000 USDC, 8 buckets  → Max loss ≈ $2,079

This is known BEFORE the market starts.

Alpha Decay: The LP’s Friend

Markets on Skepsis can use alpha decay, where the liquidity parameter shrinks over time.

Why This Helps LPs

Market opens: α = 3,333
Required reserves: $9,240

Market matures: α = 1,000
Required reserves: $2,772

Surplus released: $6,468
Vault harvests this. LPs get capital back early.
Early in a market’s life, spreads are wide and the LP takes more risk, but earns more from the spread. As the market matures, spreads tighten (better for traders) and the LP’s risk decreases. The excess capital gets recycled.

Solvency: Three Layers

Layer 1: Market Level

Every market enforces the P-Z invariant:
poolBalance ≥ maxLiability + 2 × α × ln(N)

Translation: The market always has enough to pay winners.
If a trade would violate this, it reverts. No exceptions.

Layer 2: Vault Level

The Vault keeps a 20% liquid buffer: capital not deployed to any market.
Vault total: $100,000
Deployed: $80,000 (across markets)
Liquid buffer: $20,000

This covers:
- Withdrawal queue processing
- Emergency scenarios
- New market seeding

Layer 3: Per-Market Cap

No single market gets more than 20% of the Vault.
Vault total: $100,000
Max per market: $20,000

Even if a market goes maximally wrong,
the Vault loses at most ~$9K, not $100K.

Expected Returns

LP returns depend on:
FactorGood for LPsBad for LPs
High trading volumeMore fees
Balanced bettingSpread across ranges
Alpha decaySurplus harvest
Many marketsDiversification
One-sided betting
Low volume
Many popular ranges winning
Bottom line: LPs are betting that across many markets, trading fees and spread income will outweigh payouts to winners. History of LMSR-based systems suggests this is generally true, but not guaranteed.

Summary

AspectDetail
Revenue sourcesTrading fees, spread, losing trader capital
RiskPopular winning ranges, bounded per market
Max loss per marketα × ln(N), known in advance
WithdrawalQueue-based, FIFO
Best forPatient capital, believers in prediction market volume

Next Steps

The VaultHow to deposit and get startedThe Vault
LMSR ExplainedThe math that makes it workLMSR Explained