Parlays Are Coming to Purrdict: Crypto's Missing Bet Type

Why can’t you parlay in crypto?

Kalshi launched “combos” last year — parlay-style bets where you combine NFL game outcomes, point totals, and player props into a single position. Bigger risk, bigger payout. It was a smart move: parlays account for over 30% of DraftKings handle and growing. Bettors love them.

But Kalshi combos are sports-only, centralized, and completely siloed from crypto.

On Polymarket? No parlays at all. Single outcomes only.

A trader who believes “BTC breaks $100K AND ETH flips $5K AND SOL reclaims $200” has to place three separate bets on three separate markets. There’s no way to express that thesis as one position, anywhere in crypto.

We’re building it.

What’s a parlay?

Pick multiple outcomes. All of them must resolve in your favor for the bet to pay out. If even one leg loses, the whole thing fails.

The tradeoff: higher risk for significantly higher reward. A 3-leg parlay with each leg at 60% probability doesn’t pay out at 60% — the combined probability drops to ~22%, so the payout multiplies accordingly.

Parlays are the most popular bet type in traditional sports gambling. They don’t exist on-chain. Until now.

The real insight: not all markets need the same oracle

Here’s what we realized building this.

A market about whether BTC crossed $100K and a market about whether a war ended are fundamentally different. The BTC question has a deterministic answer — a price on a timestamp. The war question requires human judgment.

Most prediction platforms treat every market the same. UMA’s optimistic oracle handles everything on Polymarket — whether it’s a crypto price or a geopolitical event. Same slow propose-dispute-vote cycle for both.

That’s wasteful. Crypto price markets have the answer already on-chain.

Tier 1: Instant resolution from on-chain data

HIP-4’s highest-volume markets are recurring crypto price bets. “BTC Up or Down — 5 Minutes.” “ETH > $2,500 — 1 Hour.” These resolve based on whether a price crossed a threshold at a specific time.

That price data already exists on Hyperliquid. HyperCore’s oracle aggregates prices from Binance, OKX, Bybit, Kraken, Kucoin, Gate IO, MEXC, and Hyperliquid’s own spot — stake-weighted median across validators, updating every ~3 seconds. HyperEVM read precompiles (live on mainnet since April 2025) make this data accessible to smart contracts.

For these markets: no AI. No human governance. No external oracle. A resolution bot reads the oracle price from the chain, checks it against the market threshold, and publishes the result. The escrow contract reads that result and settles. Fast, deterministic, trustless.

This matters because crypto price parlays are the highest-frequency use case. “BTC up AND ETH up AND SOL up in the next 5 minutes” is a bet people want to make every few minutes, all day. Instant resolution makes that possible.

Tier 2: AI consensus for everything else

Markets about elections, geopolitics, community events, or anything without an on-chain price feed need external data. This is where SignalConsensus comes in.

Three independent AI models — GPT-5, Claude Sonnet 4.6, and Gemini 2.5 Pro — evaluate the outcome. 2-of-3 consensus determines the result. Every model’s vote recorded on-chain.

You can’t buy AI votes. You can’t bribe models that have no wallets. You’d need to compromise 2 of 3 independent providers (OpenAI, Anthropic, Google) simultaneously. Compare that to UMA, where a single whale used 5 million tokens to manipulate a $7M Polymarket resolution in March 2025.

Why two tiers beats one oracle

The system degrades gracefully. Tier 1 works even if every AI provider goes down. And the highest-volume markets — the ones that generate the most fees — never need to touch an oracle.

How parlay pricing works

Parlays have correlated risk that AMMs can’t price. If BTC pumps, ETH and SOL probably pump too. A 5-leg crypto parlay isn’t 5 independent bets — the legs are correlated.

Market makers who understand this correlation can offer fair odds. We use request-for-quote:

1. You build a parlay (pick your outcomes)
2. We fan out an RFQ to market makers
3. MMs quote combined odds accounting for correlation
4. You accept, both sides sign, collateral locks in escrow

The entire RFQ engine runs inside AWS Nitro Enclaves — hardware-isolated environments where nobody can see your trade parameters. Not the server. Not us. Not anyone.

How settlement works

The escrow contract handles the mechanics:

• Both sides deposit collateral, locked by cryptographic signatures
• Replay-resistant nonces prevent double-spending
• Legs resolve independently — Tier 1 price bots or Tier 2 AI consensus publish results to the resolution contract
• Early termination: one leg loses, the market maker wins immediately. No waiting for every leg.
• All legs win: you take the full pot

The contracts are written, tested (including fuzz testing), and have been through an internal security review. Third-party audit is planned before mainnet.

How we compare

Polymarket has liquidity. Kalshi has regulation. HyperMarkets has speed for simple bets. We have the full stack: parlays + two-tier resolution + private pricing + composability.

Beyond parlays

Parlays are the first product, not the last. The same infrastructure — on-chain escrow + two-tier resolution + private RFQ — supports:

• Exotic derivatives: continuous payoff, not just binary
• Conditional orders: “If ETH breaks $5K, buy 10 SOL” — prediction-triggered execution
• Cross-protocol composability: any contract on HyperEVM can read resolution data

The bottom line

Kalshi proved parlays work. They just haven’t come to crypto yet.

We’re building them on the only chain that has on-chain price data, a native prediction market protocol, and unified margin — with resolution that doesn’t depend on a committee, a governance vote, or trusting a platform.

Crypto price parlays resolve from chain data. Real-world parlays resolve from AI consensus. Both settle atomically in a single transaction.

Parlays are coming to Purrdict. Join the waitlist to get early access.