How It Works#

Deep dive into the mechanics that power Seesaw's prediction markets. This section covers the market lifecycle, oracle integration, order matching, and settlement process.

The Core Concept#

Seesaw creates binary prediction markets - markets with exactly two outcomes:

Market Timing#

Every market follows a precise cadence based on its configured duration and Unix timestamps. Durations range from 60 seconds to 7 days (default: 15 minutes):

market_id = floor(unix_timestamp / duration_seconds)
t_start   = market_id × duration_seconds
t_end     = t_start + duration_seconds

Example Timeline (UTC) — 15-minute markets#

Multiple markets with different durations can run concurrently for the same asset (e.g., a 1-minute and a 1-hour SOL/USD market).

Complete Lifecycle#

Note: create_market atomically initializes the market account and captures the start price from Pyth, transitioning through the transient CREATED state directly into TRADING.

State Details#

Price Mechanics#

Shares are priced between 0 and 1 USDC (stored as basis points internally):

The Probability Relationship#

Since YES and NO are complementary:

P(YES) + P(NO) = 100%

If YES trades at 0.60:
└── Implied P(UP) = 60%
└── NO should trade at 0.40
└── Implied P(DOWN) = 40%

Order Book Mechanics#

Seesaw uses a single unified order book for YES shares. NO orders are automatically converted:

Conversion Rules#

Oracle Integration#

Seesaw uses Pyth Network for price data:

Sampling Rules#

Start Price (P_start):

• First Pyth update where publish_time >= t_start
• Must have positive price
• Immutable once captured

End Price (P_end):

• First Pyth update where publish_time >= t_end
• Must have positive price
• Immutable once captured

Resolution Logic#

Important: Equality (no change) resolves as UP.

Collateralization#

Every share in circulation is backed 1:1 by USDC:

Solvency Invariant#

The vault always holds enough to pay all possible winners:

vault_balance >= max(total_yes_shares, total_no_shares)

This ensures the protocol can always settle regardless of outcome.

Fee Structure#

Fees are configurable by the protocol admin. The defaults are:

Makers pay zero fees. The taker fee is split between the protocol treasury and the market creator.

Example Calculation#

Settlement Process#

After resolution, winners claim their payouts:

Related Documentation#

• Order Book Details - Matching engine deep dive
• Oracle Specification - Pyth integration details
• Settlement Rules - Claim process and edge cases