🐟 Powered by MiroFish — #1 on GitHub Trending

Follow the swarm,
not the crowd.

Three independent voices on every match. ML model, agent swarm, sharpest market. We only surface a signal when they agree the odds are wrong.

Currently in validation phase · 10 live matches under calibration.

Get Early Access See the Swarm
LIVE SWARM

Consensus Engine

5 specialized agent roles. One consensus. Cross-checked against the sharpest market odds.
Click any node to see what the swarm found.

83%
HOME WIN
CONSENSUS
📊
Stats
🧠
Tactics
📡
News
🌊
Sentiment
📈
Market
↑ Click any swarm node to see its findings
📊
Stats Swarm
SIGNAL
xG Home
1.84
xG Away
0.91
Elo Delta
+87
Clean Sheet H
64%
🧠
Tactics Swarm
ANALYSIS
Press System
High>Low
Formation
3-4-3 adv.
Set Piece Edge
+14%
Transition Risk
Low
📡
News Swarm
CONFIRMED
Starting XI
Confirmed
Key Injuries
0
Home Form L10
8W-1D-1L
Travel Fatigue
None
🌊
Sentiment Swarm
CROWD
Crowd Backing H
71%
Sharp Money
+$2.4M H
Social Momentum
↑ +34%
Media Sentiment
Positive
📈
Market Swarm
EDGE
Pinnacle Fair
58%
SwarmBet Ensemble
74%
Edge vs Market
+16pp
Odds Direction
Drifting H
Demo · Bayern München vs Borussia Dortmund · Bundesliga · Matchday 28
33k+
GitHub Stars MiroFish Engine
$4.1M
Funded in 24h Chen Tianqiao backed
1M+
Agents at scale OASIS Framework
#1
GitHub Trending Above OpenAI, Google
⚽ Bundesliga
⚽ Premier League
⚽ La Liga
⚽ Serie A
⚽ Ligue 1
🏈 NFL — Coming
🏀 NBA — Coming
⚾ MLB — Coming
🏒 NHL — Coming
⚽ Bundesliga
⚽ Premier League
⚽ La Liga
⚽ Serie A
⚽ Ligue 1
🏈 NFL — Coming
🏀 NBA — Coming
⚾ MLB — Coming
🏒 NHL — Coming
THE ENGINE
MiroFish doesn't predict prices.
It predicts human reactions.

MiroFish is the open-source swarm intelligence engine that took GitHub by storm. It builds a parallel digital world from real-world data, spawns thousands of autonomous agents with unique personalities and memories, and watches what emerges.

1
Knowledge Graph
Seed data (match stats, news, odds) is processed via GraphRAG. Entities and relationships become the simulation's foundation.
2
Agent Personas
Thousands of agents are generated — each with unique personality, stance, memory, and social connections. Not clones. Individuals.
3
OASIS Simulation
Agents interact on two parallel social platforms. They follow, argue, repost, form coalitions — 23 social actions at scale.
4
Emergent Signal
A ReportAgent analyzes opinion shifts, coalition formation, and behavioral patterns — then produces a structured prediction report.
🐟
MiroFish Engine
Open-source swarm intelligence. Digital worlds with autonomous agents that have personalities, long-term memory, and social behavior. Built on OASIS by CAMEL-AI. Crowd behavior simulation at unprecedented scale.
1M+
Agents Scalable
33k+
GitHub Stars
23
Social Actions
$4.1M
Backed in 24h
Live on Polymarket — Proven Results
WHAT SWARMBET ADDS
MiroFish simulates the crowd.
We find where it's wrong.

MiroFish is the swarm. SwarmBet is the brain that turns swarm behavior into tradeable edge — with a proprietary ML layer built specifically for sports and prediction markets.

🧠
Proprietary ML Backbone
24 engineered features per match. Elo ratings, xG, form splits, implied market probabilities. GradientBoosting with sigmoid calibration. Walk-forward validated — the model only ever sees past data.
Our IP
⚡
Edge Detection Engine
We compare MiroFish consensus against our calibrated probabilities — and both against live bookmaker odds. Triple-layer edge detection. We only surface signals when all three disagree with the market.
The Signal
📊
Sports-Native Pipeline
MiroFish is domain-agnostic. SwarmBet adds the sports-specific data layer: 50k+ historical matches, 5 European leagues, real-time odds from 10+ bookmakers. The domain expertise MiroFish doesn't ship with.
Data Moat
PIPELINE
From seed data to signal
📰
Seed Data
Match stats, team news, odds, social signals feed into the pipeline
SwarmBet
🐟
MiroFish Simulates
Thousands of agents form opinions, argue, and converge in a parallel digital world
MiroFish
🧠
ML Backbone
Our proprietary model calibrates probabilities from 24 features per match
SwarmBet
🎯
Edge Signal
Swarm consensus × ML calibration × live odds = actionable edge
Signal
POLYMARKET PROOF
MiroFish is already
making money.

The MiroFish community is already using swarm simulation for Polymarket trading. One developer simulated 2,847 digital humans before every trade — and let the bot bet against the real crowd. SwarmBet takes this further with a dedicated ML layer for sports markets.

Community-reported results — not SwarmBet claims
MiroFish × Polymarket
Community Data
Trades Executed 338
Total Profit $4,266
Agents per Simulation 2,847
Best Single Trade +1,655%
Daily Avg Profit ~$669
* Community-reported. Past performance ≠ future results. SwarmBet adds its own ML layer.
API
Plug into the swarm.

Integrate SwarmBet's prediction engine into your platform. MiroFish simulation + proprietary ML backbone — delivered as a simple REST API.

Core
Predictions API
Calibrated probability distributions for every match across 5 European leagues. Pre-match and live.
  • 1X2, Over/Under, BTTS
  • Calibrated probabilities
  • Confidence scoring
  • 5 leagues, real-time
MiroFish
Swarm Insights
AI-generated match analysis from the MiroFish simulation. What the swarm thinks — and why.
  • Agent consensus breakdown
  • Key prediction drivers
  • Upset risk indicators
  • Embeddable widgets
Premium
Edge Signals
Real-time value detection. Swarm × ML × Market — triple-layer edge for operators and quant desks.
  • Live value detection
  • 10+ bookmaker comparison
  • Kelly-optimized sizing
  • Webhook alerts
Join the swarm.

Closed beta launching Q3 2026. Limited spots for operators, media companies, and builders who want early API access.

No spam. We'll reach out when it's time.

Industry conversations welcome before beta. Bookmakers, scouts, trading desks: sebastian@infty-consulting.com