Oracle Specification

1. Optimistic Resolution (UMA-style, on Solana)

Inspired by UMA's Optimistic Oracle (whitepaper, 2020). Adapted for Solana because the existing UMA bridge to Solana is an event log without dispute primitives on the Solana side.

stateDiagram-v2
    Proposed --> Confirmed: challenge_window elapsed, no challenge
    Proposed --> Challenged: dispute(bond)
    Challenged --> Settled: DVM tally

Bonds

The configurable challenge_window defaults to 24h. Markets that resolve on slow data (election results) can set it to 72h.

2. Aggregated Feeds (Pyth / Switchboard)

For continuous-data markets (price thresholds, weather, scores), the resolver pulls M-of-N aggregated samples and takes the median across non-stale entries.

def resolve_binary(samples, now, threshold, max_staleness, m):
    fresh = [s.value for s in samples if now - s.timestamp <= max_staleness]
    assert len(fresh) >= m
    median = sorted(fresh)[len(fresh) // 2]
    return 0 if median >= threshold else 1

Default settings: m = 3, max_staleness = 60s, sources sourced from Pyth product feeds + Switchboard v3 oracles + an optional in-house attestation node.

3. Manual

For test markets and off-chain games (esports brackets, governance polls). resolution_source = Manual { authority }. The authority signs a single transaction setting winning_outcome. There is no challenge window -- this mode is for low-stakes flow only.

4. Dispute Escalation: Reputation DVM

When the optimistic layer transitions to Challenged, control hands off to the Reputation Module.

flowchart TD
    A[Challenge raised] --> B[Open DVM round]
    B --> C[Voters stake $WAGR]
    C --> D[Cast vote on outcome]
    D --> E[Round closes after voting_window]
    E --> F{Stake-weighted majority}
    F -->|winning side| G[Score *= 1.10, stake intact]
    F -->|losing side| H[Score *= 0.5, stake slashed]

Reputation Mechanics

• score starts at 1.0 (REP_ONE = 1_000_000 in fixed-point).
• Correct vote: score *= 1.10, capped at REP_MAX = 4.0.
• Incorrect vote: score *= 0.5, floored at REP_MIN (so a fully-slashed voter can recover).
• Effective stake in the next tally = stake * score.
• Slash rate per incorrect vote = slash_bps (default 1000 = 10%).

5. Feed Inventory

The off-chain service exposes /oracle/feeds so frontends can render available sources. Defaults:

6. Why Not Just Defer to Pyth?

Pyth is excellent for liquid price feeds and unsuitable for almost anything else. "Did SOL close above $300 on the last day of Q3?" is a multi-step query (which timezone? which close? which aggregator?) that the market author should be able to bind unambiguously at creation. UMA-style optimistic resolution lets the author write the question in human terms and lets the crowd enforce the spec via bonded disputes.