KLEROS
ORACLE · PREDICTION MARKETS
A TECHNICAL BRIEF
VOL. I · MMXXVI · WITH RECEIPTS
01 / 08
SCROLL · CLICK · SWIPE · TAP · ↓
The resolution layer
The oracle for
prediction markets.
A prediction market is only as trustworthy as the layer that resolves it. When two sides disagree on an outcome, Kleros settles it through a neutral, transparent due process that scales alongside the platform.
Scalability
A randomly drawn jury per case, from a subcourt suited to the topic. With 750+ jurors already staked, capacity is not a constraint.
Due process
Structured evidence, reasoned votes, and appeals that double the jury each round.
Credible neutrality
Contested outcomes decided by an independent, randomly selected jury.
§ 02
Proof, before the pitch
Live since 2018. Not a whitepaper.
2018
Running a live, on-chain arbitration system since.
2020
Resolving prediction-market disputes. Omen (now Presagio) was first to integrate.
2,653
Disputes resolved on-chain across Ethereum, Gnosis & Arbitrum.
+750
Jurors participating in Kleros arbitration.
five years of contested outcomes, settled by jury.
§ 03
How it works
One question. A two-layer resolution.
Reality.eth · optimistic layer
01
Question posted
With a bond, naming Kleros as the arbitrator.
→
02
Bond escalation
Anyone answers or challenges by doubling the bond. A wrong bond is forfeited.
→
Most questions settle right here.
No court needed.
↓
genuine disagreement · pay a fee to escalate
Kleros Court · arbitration
03 · Sortition
Jurors drawn
Randomly, weighted by staked PNK, from the relevant subcourt.
→
04 · Vote
Evidence & ruling
Jurors vote independently. Coherent votes rewarded; incoherent ones slashed.
→
05 · Settle
Back to the market
The ruling flows back to Reality.eth and settles the market.
06 · The appeal escalator
3
→
7
→
15
→
31
→
…
Jurors per round. Each appeal doubles the jury, so overturning a well-evidenced answer becomes progressively more expensive.
Jurors cannot be chosen, and they cannot choose their case. More PNK raises your probability of being drawn, but it never buys the seat.
§ 04
Why not UMA
Same game. Different table.
Both reward voters for converging on the answer the panel is expected to choose, a Schelling point. The difference is who sits on the panel.
UMA
Kleros
Who answers
The entire token-holder base votes on every dispute.
A fixed-size jury, drawn per case from a topic-relevant subcourt.
Who decides
Token-weighted vote of holders.
A random, stake-weighted jury draw, per case.
Scalability
Everyone votes on everything. As volume climbs, voters can't study each case.
Each case gets a dedicated jury drawn from 750+ active jurors, so review depth holds as volume grows.
Due process
No appeal; escalation to the DVM, where ~9 wallets hold 50% of voting power.
Evidence + appeals. The jury doubles each round, raising the cost to overturn a coherent answer.
Speed
~2 days.
Most cases resolve in a day; heavily appealed cases can take a couple of weeks.
§ 05
On real money
Four disputes, settled by a drawn jury.
#532 · OMEN / PRESAGIO
↗ KLEROSBOARD
Will Joe Biden win the 2020 US presidential election?
General Court · 500 votes
Ruled Yes
#302 · OMEN / PRESAGIO
↗ KLEROSBOARD
≥1,000 reported US COVID deaths in a single day, first half of July 2020?
General Court · 500 votes
Ruled Yes
#1661 · SEER
↗ KLEROSBOARD
Was Zelenskyy photographed or videotaped wearing a suit between May 22 and June 30, 2025?
General Court · 31 votes
Ruled Yes
#1657 · SEER
↗ KLEROSBOARD
Winner of the run-off in the 2024 Romanian presidential election?
General Court · 31 votes
Resolved too soon, run-off annulled
The hard cases, including the one where the right call was that the market resolved too soon. Every ruling is a live, on-chain link.
§ 06
When oracles fail
Not every contested outcome is an oracle failure.
Some contested Polymarket outcomes trace to the market's own rules or a late clarification, not the oracle. But the oracle layer can be a structural weak point in its own right.
On UMA, voting power is concentrated in a handful of large holders, and a majority of active voters are linkable to Polymarket trading accounts, a direct positional conflict in several markets.
Under Kleros
25% of stake is a 25% chance of being drawn as a juror, not 25% of the verdict.
And every appeal raises the cost of holding a dishonest result, making concentration expensive rather than decisive.
§ 07
Who's already in
Prediction markets resolving with Kleros.
Seer
Decentralization, composability, and community-driven resolution.
Omen / Presagio
Built on Gnosis' Conditional Tokens Framework. First to integrate, 2020.
Foresight
Kleros's own forecasting experiment, running on Seer.
Precog
Collective intelligence for human and AI participants.
Futarchy.fi
Prediction markets & conditional tokens for DAO decision-making.
Deep Funding
Allocation proposals run through conditional prediction markets.
Ringo
Open markets on real-world events, integrated with X.
Resolving disputes since 2020.
SEER · GNOSIS · ETHEREUM
§ 08 · Integrate
Make Kleros your
resolution layer.
Plug Reality.eth and Kleros into your market in a few steps. Optimistic by default, with a credible court behind the hard cases.
Available on
Ethereum · Arbitrum · Gnosis · Polygon · Base · Optimism · Unichain
Solana (integration in progress)
Solana (integration in progress)
Contact [email protected] · t.me/juanrah
KLEROS · ORACLE FOR PREDICTION MARKETS · MMXXVI