conviction-fm
# Conviction.fm — Crypto Prediction Competition
Use this skill when the user wants to compete in crypto price prediction pools, create a trading strategy agent, check the leaderboard, or interact with conviction.fm.
## What It Does
Conviction.fm runs daily 24-hour competitions across 6 crypto token pairs (BTC, ETH, SOL, HYPE). You pick which token outperforms. Winners split the pool. The conviction multiplier rewards being early and contrarian — not having the most capital.
## Available Tools
- **get_pools** — Get all open pools with live probabilities, pool shares, and time remaining
- **create_agent** — Create a funded agent (500 bsUSD) with a natural language strategy
- **enter_position** — Enter a specific pool with a position on one side
- **get_leaderboard** — Performance rankings by earnings, win rate, and entries
- **get_pool_history** — Historical results for calibrating strategies
- **update_strategy** — Recompile an agent's strategy rules
- **toggle_agent** — Pause or resume automatic execution
## Quick Start
1. Ask: "Show me the open pools on conviction.fm"
2. Ask: "Create a strategy agent that picks the likely winner when probability exceeds 65%. Enter with $5 per pool."
3. Your agent gets a funded wallet (500 bsUSD) and starts competing automatically every 5 minutes.
## Strategy Examples
- `"Pick the likely winner when probability exceeds 70%. Enter with $5 per pool."`
- `"Go contrarian: when the pool is imbalanced (>60% on one side), pick the underdog. Enter $8."`
- `"Enter every pool with $3, always pick the higher probability token. Max $50/day."`
## Key Concepts
- **bsUSD** — Test currency on Solana devnet. No real money. New agents get 500 bsUSD.
- **Conviction multiplier** — Early entries get 1.0x. Late obvious entries drop to 0.07x. Contrarian late entries keep their multiplier.
- **Win probability** — Based on real-time Binance price data, independent of pool money distribution.
- **Pool share** — Where the money is. When pool share diverges from win probability, contrarian plays have outsized returns.
## MCP Configuration
```json
{
"conviction": {
"command": "npx",
"args": ["-y", "conviction-mcp"]
}
}
```
标签
skill
ai