Strategies

Strategies define how your AI agents analyze prediction markets, weigh evidence, and decide when to trade. Each strategy brings a distinct analytical approach — from multi-model adversarial debate to research-driven probabilistic reasoning. Every strategy shares the same risk management pipeline, rules engine, and order execution layer.

Deploying a Strategy

1. Open Strategies in the dashboard and select the strategy you want to deploy.
2. Configure deployment settings: duration (minutes, or 0 for unlimited) and cycle frequency (seconds between market scans).
3. Optionally set per-agent overrides: max trade size, max positions, daily loss limit, min confidence threshold, max trades per day, and allowed market categories.
4. Choose your environment: Training (paper trades, no real money) or Live (real orders on the exchange).
5. Click Deploy. The agent launches immediately.

Once deployed, the agent begins scanning markets, analyzing opportunities, and trading autonomously. Monitor progress in real-time from the strategy detail page — including live terminal logs streamed directly from the agent process.

Pre-flight Checks

Before every deployment, the system runs automatic pre-flight checks to prevent common errors:

• No duplicate agents: The target strategy must not already be running for your account.
• Valid credentials: Live deployments require active exchange credentials — Kalshi API keys or a Polymarket private key, depending on the strategy.
• Live trading switch: The live trading switch must be turned on for live deployments. Training mode bypasses this check.

If any check fails, deployment is blocked and a clear error message tells you exactly what to fix.

Council V2

Council V2 is the flagship strategy — a clean-sheet rewrite of the original Council that delivers faster execution, tighter edge filtering, and improved reliability. Three frontier models (Grok 4.1 Fast, Claude Opus 4.6, and Claude Sonnet 4.6) power five specialized agents that analyze every market through a sequential debate. Each agent builds on the previous output, creating a layered argument that surfaces genuine analytical edges rather than noise.

Council V2 runs on both Kalshi and Polymarket. The AI pipeline is identical across exchanges — only the execution layer differs. Polymarket orders use signed transactions on the Polygon blockchain, while Kalshi orders use the standard REST API.

For the full architecture deep dive — debate structure, edge calculation, position sizing, and risk layers — see Council V2.

The Superforecaster

The Superforecaster takes a fundamentally different approach. Instead of debating across multiple models, it uses Perplexity Sonar Deep Research to gather comprehensive web research, then a single powerful reasoning model applies structured decomposition from the superforecasting literature — breaking questions into sub-questions, establishing base rates with real sample sizes, and synthesizing inside and outside views into a calibrated probability. Every data point is sourced, never hallucinated.

Like Council V2, the Superforecaster runs on both Kalshi and Polymarket with an identical analytical pipeline. You choose the reasoning model at deployment time, giving you control over the speed-accuracy tradeoff.

For the full methodology — research pipeline, decomposition framework, and calibration approach — see The Superforecaster.

The Council (Legacy)

The original Council strategy uses a 5-agent parallel ensemble where specialized AI models — Forecaster, News Analyst, Bull Researcher, Bear Researcher, and Risk Manager — each running on a different provider, independently analyze every market opportunity. Their probability estimates are combined through a confidence-adjusted weighted average to produce a consensus, requiring at least 3 of 5 agents to agree.

For historical reference and a detailed walkthrough of the original pipeline, see The Council (Legacy).