How the agents perform.
Three independent AI agents analyze every active Polymarket market. Each prediction is graded when the market resolves. Track records update continuously — accuracy determines weight in the consensus.
30+
Markets analyzed
per run cycle
~40%
Consensus rate
of markets reach agreement
6
Categories covered
politics to crypto
3+
Agents per market
independent scores
THE AGENTS
Three independent perspectives. One consensus.
Anthropic
Claude
Excels on multi-step inferential questions where the answer requires chaining several conditional probabilities. Strong on regulatory and legislative outcomes.
Google DeepMind
Gemini
Integrates live search results before scoring each market. Strongest on breaking news events and rapidly evolving situations where training cutoff matters most.
xAI
Grok
Trained on a corpus with higher social media representation. Demonstrates above-average calibration on markets driven by public sentiment and crowd behavior.
CALIBRATION DATA
Per-agent accuracy, by category.
Accuracy scores update after every market resolution. Agents with higher historical accuracy on a category carry more weight in the consensus for that category.
MARKET CATEGORIES
Where the agents focus.
Politics
Elections, legislation, regulatory outcomes
Crypto
Token prices, protocol events, exchange listings
Sports
Match outcomes, championships, athlete events
World Events
Geopolitics, conflicts, international agreements
Economics
Macro data, Fed decisions, market benchmarks
Science & Tech
AI milestones, launches, regulatory approvals
How the consensus is calculated
Each agent independently produces a probability estimate for every active market. Those estimates are combined using a proprietary accuracy-weighting model — agents that have demonstrated better historical calibration on a given category contribute more to the final consensus probability.
When the consensus probability diverges from the live Polymarket price by more than a minimum threshold, the system identifies it as a potential edge opportunity. The size of the gap determines position size via Kelly Criterion — larger gaps produce larger positions, within hard portfolio caps.