CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

trading-manifold

Place bets on Manifold Markets using their REST API

56

Quality

71%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./src/skills/bundled/trading-manifold/SKILL.md
SKILL.md
Quality
Evals
Security

Manifold Markets Trading Skill

Real, working methods to bet on Manifold Markets using Mana (play money that can be donated to charity).

Setup

Get your API key from: https://manifold.markets/profile (API key section)

import os
import requests

API_URL = "https://api.manifold.markets/v0"
API_KEY = os.getenv("MANIFOLD_API_KEY")

def headers():
    return {
        "Authorization": f"Key {API_KEY}",
        "Content-Type": "application/json"
    }

Search Markets

def search_markets(query: str, limit: int = 10):
    """Search for markets"""
    r = requests.get(f"{API_URL}/search-markets", params={
        "term": query,
        "limit": limit,
        "filter": "open",
        "sort": "liquidity"
    })
    r.raise_for_status()
    markets = r.json()

    for m in markets[:5]:
        prob = m.get("probability", 0.5)
        print(f"\nMarket: {m['question']}")
        print(f"ID: {m['id']}")
        print(f"Probability: {prob*100:.1f}%")
        print(f"URL: {m.get('url', '')}")

    return markets

markets = search_markets("AI")

Get Market by ID or Slug

def get_market(id_or_slug: str):
    """Get market details"""
    # Try by ID first
    r = requests.get(f"{API_URL}/market/{id_or_slug}")
    if r.status_code == 404:
        # Try by slug
        r = requests.get(f"{API_URL}/slug/{id_or_slug}")

    r.raise_for_status()
    return r.json()

market = get_market("will-gpt5-be-released-before-2025")
print(f"Question: {market['question']}")
print(f"Probability: {market.get('probability', 0.5)*100:.1f}%")

Place a Bet

def place_bet(
    market_id: str,
    amount: int,           # Mana amount to bet
    outcome: str = "YES",  # "YES" or "NO"
    limit_prob: float = None  # Optional limit order probability
):
    """
    Place a bet on Manifold

    Args:
        market_id: The market ID (not slug!)
        amount: Amount of Mana to bet
        outcome: "YES" or "NO"
        limit_prob: Optional - if set, creates a limit order at this probability
    """
    payload = {
        "contractId": market_id,
        "amount": amount,
        "outcome": outcome
    }

    if limit_prob is not None:
        payload["limitProb"] = limit_prob

    r = requests.post(f"{API_URL}/bet", headers=headers(), json=payload)
    r.raise_for_status()
    result = r.json()

    print(f"Bet placed!")
    print(f"Shares: {result.get('shares', 0):.2f}")
    print(f"Probability after: {result.get('probAfter', 0)*100:.1f}%")

    return result

# Market bet - buys at current price
result = place_bet(
    market_id="abc123",
    amount=100,  # 100 Mana
    outcome="YES"
)

# Limit order - only fills at 40% or below
result = place_bet(
    market_id="abc123",
    amount=100,
    outcome="YES",
    limit_prob=0.40
)

Cancel Bet (Limit Orders Only)

def cancel_bet(bet_id: str):
    """Cancel a limit order"""
    r = requests.post(f"{API_URL}/bet/cancel/{bet_id}", headers=headers())
    r.raise_for_status()
    return True

cancel_bet("bet123")

Sell Shares

def sell_shares(
    market_id: str,
    outcome: str = "YES",
    shares: float = None  # None = sell all
):
    """
    Sell shares in a market

    Args:
        market_id: The market ID
        outcome: "YES" or "NO" - which shares to sell
        shares: Number of shares to sell (None = all)
    """
    payload = {
        "contractId": market_id,
        "outcome": outcome
    }

    if shares is not None:
        payload["shares"] = shares

    r = requests.post(f"{API_URL}/market/{market_id}/sell", headers=headers(), json=payload)
    r.raise_for_status()
    return r.json()

# Sell all YES shares
sell_shares("abc123", "YES")

# Sell specific amount
sell_shares("abc123", "YES", shares=50.0)

Get Your Bets

def get_my_bets(market_id: str = None):
    """Get your bets"""
    params = {}
    if market_id:
        params["contractId"] = market_id

    r = requests.get(f"{API_URL}/bets", headers=headers(), params=params)
    r.raise_for_status()
    bets = r.json()

    for b in bets[:10]:
        print(f"Bet: {b['outcome']} {b['amount']}M @ {b.get('probBefore', 0)*100:.0f}%")

    return bets

bets = get_my_bets()

Get Your Positions

def get_positions():
    """Get current positions across all markets"""
    # Get user info first
    r = requests.get(f"{API_URL}/me", headers=headers())
    r.raise_for_status()
    user = r.json()

    # Get bets to calculate positions
    r = requests.get(f"{API_URL}/bets", headers=headers(), params={"limit": 1000})
    bets = r.json()

    # Aggregate by market
    positions = {}
    for bet in bets:
        mid = bet["contractId"]
        if mid not in positions:
            positions[mid] = {"yes": 0, "no": 0, "invested": 0}

        if bet["outcome"] == "YES":
            positions[mid]["yes"] += bet.get("shares", 0)
        else:
            positions[mid]["no"] += bet.get("shares", 0)

        if not bet.get("isSold", False):
            positions[mid]["invested"] += bet["amount"]

    return positions, user.get("balance", 0)

positions, balance = get_positions()
print(f"Balance: {balance} Mana")
for mid, pos in positions.items():
    if pos["yes"] > 0 or pos["no"] > 0:
        print(f"Market {mid}: YES={pos['yes']:.1f}, NO={pos['no']:.1f}")

Get Balance

def get_balance():
    """Get your Mana balance"""
    r = requests.get(f"{API_URL}/me", headers=headers())
    r.raise_for_status()
    user = r.json()
    return user.get("balance", 0)

balance = get_balance()
print(f"Balance: {balance} Mana")

Complete Trading Bot Example

#!/usr/bin/env python3
"""
Manifold arbitrage bot - finds mispriced markets
"""

import os
import time
import requests

API_URL = "https://api.manifold.markets/v0"
API_KEY = os.getenv("MANIFOLD_API_KEY")

def h():
    return {"Authorization": f"Key {API_KEY}", "Content-Type": "application/json"}

def search(query):
    r = requests.get(f"{API_URL}/search-markets",
                    params={"term": query, "limit": 20, "filter": "open"})
    return r.json()

def bet(market_id, amount, outcome, limit_prob=None):
    payload = {"contractId": market_id, "amount": amount, "outcome": outcome}
    if limit_prob:
        payload["limitProb"] = limit_prob
    r = requests.post(f"{API_URL}/bet", headers=h(), json=payload)
    return r.json()

def get_balance():
    r = requests.get(f"{API_URL}/me", headers=h())
    return r.json().get("balance", 0)

# Strategy: Buy extreme probabilities (likely to revert)
MIN_LIQUIDITY = 1000  # Only trade liquid markets

while True:
    try:
        balance = get_balance()
        print(f"\nBalance: {balance} Mana")

        # Search trending markets
        markets = search("2024")

        for m in markets:
            prob = m.get("probability", 0.5)
            liquidity = m.get("totalLiquidity", 0)

            if liquidity < MIN_LIQUIDITY:
                continue

            # Buy YES on very low probability (< 10%)
            if prob < 0.10:
                print(f"LOW: {m['question'][:50]} at {prob*100:.1f}%")
                if balance > 50:
                    bet(m["id"], 50, "YES", limit_prob=0.15)

            # Buy NO on very high probability (> 90%)
            elif prob > 0.90:
                print(f"HIGH: {m['question'][:50]} at {prob*100:.1f}%")
                if balance > 50:
                    bet(m["id"], 50, "NO", limit_prob=0.85)

        time.sleep(300)  # Check every 5 minutes

    except Exception as e:
        print(f"Error: {e}")
        time.sleep(60)

Multiple Choice Markets

def bet_multiple_choice(market_id: str, answer_id: str, amount: int):
    """Bet on a multiple choice market"""
    payload = {
        "contractId": market_id,
        "amount": amount,
        "answerId": answer_id
    }

    r = requests.post(f"{API_URL}/bet", headers=headers(), json=payload)
    r.raise_for_status()
    return r.json()

# Get market with answers
market = get_market("who-will-win-2024-election")
for answer in market.get("answers", []):
    print(f"{answer['text']}: {answer['probability']*100:.1f}% (ID: {answer['id']})")

# Bet on specific answer
bet_multiple_choice("market123", "answer456", 100)

Create a Market

def create_market(
    question: str,
    description: str = "",
    close_time: int = None,  # Unix timestamp
    initial_prob: float = 0.5,
    ante: int = 100  # Initial liquidity
):
    """Create a new binary market"""
    payload = {
        "outcomeType": "BINARY",
        "question": question,
        "description": description,
        "initialProb": int(initial_prob * 100),
        "ante": ante
    }

    if close_time:
        payload["closeTime"] = close_time * 1000  # Milliseconds

    r = requests.post(f"{API_URL}/market", headers=headers(), json=payload)
    r.raise_for_status()
    return r.json()

# Create market
new_market = create_market(
    question="Will it rain tomorrow in NYC?",
    initial_prob=0.3,
    ante=100
)
print(f"Created: {new_market['url']}")

Important Notes

  1. Mana is play money - Can be donated to charity
  2. No limit on bets - Unlike real money markets
  3. Market maker AMM - Prices move based on bets
  4. Limit orders - Use limitProb for better prices
  5. API rate limits - Be gentle, ~60 requests/minute
  6. Shares ≠ Mana - Shares vary based on probability
Repository

alsk1992/CloddsBot

Path

src/skills/bundled/trading-manifold/SKILL.md

Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.