How to Create a Prediction Market
Most "how to create a prediction market" guides describe Polymarket or Kalshi, where the answer is "you can't — only their team can". This guide is for actually doing it on a creator-friendly platform.
Step 1 — Apply for creator access
Sign in with the wallet you'll use to create markets, then open /apply. The form asks for:
- Display name and X handle (Discord and Farcaster optional).
- Topic areas you'd cover (e.g. "Premier League football, La Liga, MLS"). Pick up to five.
- Two example markets you'd launch in week one — full question, time-bounded.
- How you'll bring traders to your markets — audience, newsletter, community size.
- A short paragraph (50–600 chars) on why you want to create on BlockForecast.
Reviews typically take under 24 hours. Approval criteria: clear questions, public-source resolvability, no insider info, no markets on individuals' personal lives.
Step 2 — Pick a topic with a clear answer
The strongest market questions have all four properties:
- Specific. "Will Arsenal beat Chelsea in their next league match?" not "Will Arsenal do well?"
- Time-bounded. Set the moment of resolution explicitly. "By Dec 31 23:59 UTC" beats "this year".
- Source-citable. A respectable source must publish the answer (Premier League site, CoinGecko, Pyth, AP News, etc.).
- Binary or small-N. Either yes/no, or a fixed enumeration of outcomes (Arsenal / Draw / Chelsea).
Bad examples that get rejected:
- Will my friend get married? — no public source.
- Will the new iPhone be better than the old one? — subjective.
- Will it rain in my city? — no source unless you specify station and date.
- Will [insider] do [private thing]? — insider info, gameable.
Step 3 — Write the question tightly
Templates that work:
Will [SUBJECT] [VERB] by [DATE/TIME]?Will [PRICE/STAT] be above [THRESHOLD] at [TIME]?Who will win [EVENT]: [A] / [B] / [C]?
Keep it under 60 chars if you can. The AI rewrites long questions, but starting clean speeds approval and lifts trader engagement (cards display the question prominently).
Step 4 — Pick a resolution date
Match the actual resolution moment:
- Sports: kickoff time. Trading auto-closes at kickoff; outcome resolves a few minutes after final whistle.
- Crypto: a specific UTC timestamp. The Pyth/CoinGecko snapshot at that moment is the answer.
- Politics / news: the event itself + a small buffer for sources to publish.
- Long-running: the deadline by which the answer must be known.
Step 5 — Confirm AI-generated resolution criteria
The AI proposes:
- Authoritative sources to consult at resolution.
- Tie-breaker rules.
- Edge cases (postponed match, exchange outage, ambiguous reporting).
Read carefully. The criteria are immutable once trading starts — that's a fairness commitment to traders. You can edit before confirming, not after.
Step 6 — Distribute and earn
Once live, share the market URL. Every trade earns you 0.5% in real time. Top creators stack thousands per month from a small portfolio. See the economics.
Common rejection reasons
- Ambiguous wording. "Did the team have a good game?" → rejected.
- No clear source. "Will this small private event happen?" → rejected.
- Insider information. Markets that trade well only if you have inside info → rejected.
- Person-specific. Most markets about a named individual's private life → rejected.
- Past-event. The answer is already known at creation time → rejected.
FAQ
How long does creator-access review take?
Typically under 48 hours. We respond by email.
Is there a fee to create a market?
Network gas (cents on Base). Some categories require a small refundable creator stake to discourage spam.
Can I create matchup-style markets?
Yes — 2-way and 3-way. Multi-outcome (up to 16) is supported via the multi-outcome pool path.
What if the AI oracle resolves wrong?
You or any user can file an appeal. Adversarial appeal agents review independently. Full oracle docs.