Predict.fun is BNB Chain’s fastest-growing prediction market — $1.5B in volume since December 2025, 120K+ users, and collateral that earns DeFi yield while you hold positions. Broad coverage across sports, politics, crypto, and pop culture. Full REST API with Python and TypeScript SDKs. Best for agents that need diversified event exposure with capital efficiency.
What Predict.fun Is
Predict.fun is a decentralized prediction market on BNB Chain where users trade binary outcome shares (YES/NO) on real-world events — NBA championships, FIFA World Cup matches, crypto price milestones, election outcomes, and pop culture predictions.
What sets it apart from Polymarket or Kalshi is yield-bearing collateral. When you deposit USDT or USDC into a market position, Predict.fun routes that capital into BNB Chain DeFi protocols (primarily Venus) to generate 5-15% APY. Your money works while your prediction cooks.
Founded by Dingaling (ex-Binance), backed by YZi Labs (formerly Binance Labs), and boosted by a social media shout-out from CZ himself in December 2025, Predict.fun launched in December 2025 and has since processed $1.5 billion in cumulative volume with over 120,000 users.
In March 2026, Predict.fun acquired Probable (incubated by PancakeSwap and YZi Labs), consolidating BNB Chain’s two main prediction platforms under one roof.
Market Coverage
Predict.fun positions itself as a casual-first, broad-coverage platform — the opposite of PredictIt’s deep-but-narrow approach:
- Sports — NBA, NFL, FIFA World Cup 2026, major league outcomes
- Crypto — BTC/ETH price targets, token launch milestones, protocol events
- Politics — US elections, global policy outcomes
- Entertainment and pop culture — Awards shows, social media milestones, cultural trends
- Macro — Select economic indicators (Fed decisions, etc.)
Markets emphasize high engagement and smaller position sizes, making it accessible to casual participants while still supporting serious traders through the order-book model.
Architecture
| Component | Implementation |
|---|---|
| Blockchain | BNB Chain |
| Order matching | Order-book model (not AMM) |
| Settlement oracle | UMA Optimistic Oracle + manual verification |
| Collateral yield | Venus Protocol integration (5-15% APY) |
| Wallet support | EOA (MetaMask, etc.) + Smart Wallet (Predict Account) |
| Collateral tokens | USDT, USDC |
| Fee model | Taker-only, settled in USDT |
Cross-chain transfers are supported with approximately 0.1% slippage. Funds are placed into an internal trading wallet separate from the user’s primary address for operational security.
Fee Structure
| Fee | Rate |
|---|---|
| Maker fee | 0% (rebates available) |
| Taker fee | Variable by market |
| Settlement | Fees finalized at market resolution, collected in USDT |
| Revenue | 100% retained by protocol |
All trading fees are taker-only, meaning market makers and liquidity providers pay nothing. Revenue retention by the protocol means no external fee-sharing — this funds ongoing development and potential future token rewards.
API and SDK for Agent Builders
Predict.fun has the most complete developer toolkit of any entertainment-focused prediction market.
REST API
| Environment | Base URL | Auth |
|---|---|---|
| Mainnet | https://api.predict.fun/ | API key required |
| Testnet | https://api-testnet.predict.fun/ | No key needed |
| Swagger docs | https://api.predict.fun/docs | — |
| Rate limit | 240 requests/minute | Both environments |
Request an API key through the Predict.fun Discord.
Official SDKs
TypeScript:
npm install @predictdotfun/sdk
Python:
pip install predict-sdk
Both SDKs are published on Context7 for LLM-assisted development:
- TypeScript:
context7.com/predictdotfun/sdk - Python:
context7.com/predictdotfun/sdk-python - REST API:
context7.com/llmstxt/dev_predict_fun_llms_txt
Python: Fetch Active Markets
import requests
API_KEY = "your_api_key"
BASE = "https://api.predict.fun"
HEADERS = {"Authorization": f"Bearer {API_KEY}"}
def get_markets():
"""Fetch all active prediction markets."""
resp = requests.get(f"{BASE}/markets", headers=HEADERS)
resp.raise_for_status()
return resp.json()
markets = get_markets()
for m in markets["data"][:5]:
print(f" {m['title']} — Volume: ${m.get('volume', 0):,.0f}")
Python: Monitor a Market with Polling
import requests
import time
API_KEY = "your_api_key"
BASE = "https://api.predict.fun"
HEADERS = {"Authorization": f"Bearer {API_KEY}"}
def monitor_market(market_id: str, interval: int = 15):
"""Poll a market for price changes."""
last_price = None
while True:
try:
resp = requests.get(
f"{BASE}/markets/{market_id}",
headers=HEADERS,
timeout=10,
)
data = resp.json()
yes_price = data.get("yes_price", 0)
if yes_price != last_price:
print(f"[{market_id}] YES: {yes_price:.2f} "
f"({'↑' if last_price and yes_price > last_price else '↓'})")
last_price = yes_price
except Exception as e:
print(f"Error: {e}")
time.sleep(interval)
TypeScript: Basic SDK Usage
import { PredictSDK } from "@predictdotfun/sdk";
const sdk = new PredictSDK({
apiKey: process.env.PREDICT_API_KEY,
// For Smart Wallet (Predict Account):
// privateKey: process.env.WALLET_PRIVATE_KEY,
});
// Fetch available markets
const markets = await sdk.getMarkets();
console.log(`Found ${markets.length} active markets`);
// Get specific market details
const market = await sdk.getMarket("market-id-here");
console.log(`${market.title}: YES=${market.yesPrice} NO=${market.noPrice}`);
Wallet Integration: EOA vs. Smart Wallet
Predict.fun supports two wallet types:
EOA (MetaMask, etc.)
→ Standard Ethereum-style wallet
→ Direct contract interaction
→ User manages gas
Smart Wallet (Predict Account)
→ Auto-created via web app
→ Programmatic access via SDK
→ Requires wallet address + signing key
For agent builders using Coinbase Agentic Wallets or Safe multisig, the EOA path is more straightforward. The Smart Wallet option is useful for agents that need account abstraction features.
Agent Infrastructure Angle
Predict.fun sits at Layer 3 — Trading in the Agent Betting Stack, with unique properties for agent portfolio management:
- Yield-bearing positions: An agent holding positions across 10+ markets passively earns 5-15% APY on deposited collateral via Venus Protocol. This changes the economics of agent capital allocation — idle capital isn’t idle.
- Broad event coverage: A single agent on Predict.fun can diversify across sports, politics, crypto, and entertainment. Most other platforms force you to specialize.
- Testnet for development: The free testnet at
api-testnet.predict.fun(240 req/min, no API key) lets agent builders iterate without spending real capital. This is the best sandbox in the prediction market space. - Cross-market arb potential: Compare Predict.fun prices against Polymarket, PredictIt, and Opinion for the same events. The BNB Chain settlement means lower gas than Polygon for high-frequency repositioning.
- Post-acquisition scale: The Probable acquisition gives Predict.fun consolidated liquidity and a stronger network effect on BNB Chain, reducing the thin-orderbook risk that plagued earlier BNB prediction markets.
Strengths and Limitations
Strengths:
- Yield on collateral (5-15% APY via Venus) — capital stays productive
- Full REST API + Python/TypeScript SDKs with testnet sandbox
- Broad market coverage: sports, politics, crypto, entertainment, pop culture
- Fast-growing: $1.5B volume in ~3 months, 120K+ users
- Low gas fees on BNB Chain
- YZi Labs / ex-Binance backing and ecosystem integration
Limitations:
- Unregulated — no CFTC, MGA, or equivalent oversight
- BNB Chain only — smaller DeFi ecosystem than Ethereum/Polygon
- Liquidity inconsistent on less popular markets (Reddit users report thin books)
- No native token yet (point system suggests future airdrop)
- Smart contract risk — DeFi yield routing adds attack surface
- Casual-first design may not attract institutional liquidity
- Relatively new (December 2025 launch)
Who Should Use Predict.fun
Predict.fun is the right platform if you’re building a diversified prediction agent that needs broad event coverage and capital efficiency, or if you want to earn yield while speculating on sports, entertainment, and pop culture outcomes.
It’s the wrong platform if you need regulatory compliance, deep political-only coverage (PredictIt), or institutional-grade macro markets (Opinion).
Last verified: March 2026