Whitepaper v2.0

AltHunter

AI-Powered Multi-Strategy Trading Platform for Polymarket

Last updated: February 2026

1. Abstract

AltHunter is an AI-powered autonomous trading platform for Polymarket prediction markets. It combines five complementary trading strategies — Lag Exploit, Cluster Analysis, Fair Value, Market Making, and Combined — with advanced AI decoders (GPT-4o, Claude), real-time exposure management, and a multi-agent evolutionary arena to deliver institutional-grade trading as a SaaS product.

The platform operates in shadow mode (paper trading) or live execution mode, with full risk management including exposure limits, close-mode circuit breakers, and regression guards that automatically detect and correct performance degradation.

2. The Problem

Prediction markets like Polymarket represent a massive and growing opportunity, but individual traders face significant challenges:

Information Complexity: Prediction markets require synthesizing news, probability theory, and market microstructure simultaneously. No human can monitor hundreds of markets in real time.

Execution Speed: Oracle lag, whale cluster movements, and fair value deviations create fleeting opportunities. Manual traders consistently miss these windows.

Risk Management: Without automated exposure tracking, traders easily overconcentrate in correlated positions and face catastrophic drawdowns.

Market Making Barrier: Providing passive liquidity (market making) requires sophisticated quoting algorithms, inventory management, and volatility-adjusted spreads — traditionally reserved for institutional desks.

Strategy Optimization: Finding the right combination of strategy parameters requires continuous backtesting and A/B experimentation that is impractical to do manually.

3. The Solution

AltHunter solves these problems through a vertically integrated AI trading platform:

Multi-Strategy Engine: Five strategies work in concert — Lag Exploit detects oracle price discrepancies, Cluster Analysis tracks whale wallet movements, Fair Value uses hybrid AI+EMA estimation, Market Maker provides passive liquidity with inventory skew, and Combined aggregates signals with confidence weighting.

AI-Powered Decoder: Every trade opportunity is analyzed by frontier AI models (GPT-4o, Claude). A/B model routing compares cheap vs. premium models, auto-escalation retries with stronger models when confidence is low, and learning transfer propagates winning rules across the system.

EMA Oracle: Multi-timeframe exponential moving average (spans 5, 15, 60) provides real-time fair value estimation, trend detection, deviation measurement, and volatility tracking — reducing unnecessary AI calls and stabilizing fair value estimates.

Exposure Management: Real-time portfolio tracking with configurable notional limits, strategy concentration guards, single-position caps, and automatic close-mode with hysteresis when risk thresholds are breached.

Multi-Agent Arena: An evolutionary optimization system that spawns agent populations with different strategy parameters, evaluates their performance, culls the weakest, and promotes the best — automatically discovering optimal configurations.

Self-Improving Daily Review: Every 24 hours the system analyzes its own performance, detects regressions, proposes parameter adjustments within guardrails, requires human approval for large changes, and sends detailed Telegram summaries.

4. Technical Architecture

Data Layer

  • Polymarket CLOB API for real-time market data and order book
  • CoinGecko API for crypto price feeds (BTC reference price)
  • EMA Oracle with multi-timeframe smoothing (5/15/60 spans)
  • SQLite for persistent state, positions, and trade history
  • Intelligence Layer

  • AI Decoder: GPT-4o and Claude via OpenAI/Anthropic APIs
  • Model Router: YAML-configurable tiered/A/B routing with auto-escalation
  • Learning Transfer: Winner rules propagated to guidelines.md automatically
  • Regression Guard: Detects win rate drops, overfit flags, and exposure concentration anomalies
  • Strategy Layer

  • Lag Exploit: Oracle price discrepancy detection with configurable thresholds
  • Cluster Analysis: Whale wallet tracking and movement correlation
  • Fair Value: Hybrid AI + EMA fair value estimation with pre-filter optimization
  • Market Maker: Bid/ask quoting with inventory skew, volatility-adjusted spreads, and close-mode integration
  • Combined: Signal aggregation with hedging confidence boost and close-mode blocking
  • Execution Layer

  • Shadow mode (paper trading) and live mode with position manager
  • Trade executor with exposure checks, MM-specific sizing, and close-mode blocking
  • Agentic wallets via Coinbase CDP with per-transaction spending policies
  • Latency tracking (p50/p95/p99) across signal evaluation and execution
  • Application Layer

  • Next.js frontend with real-time dashboard, 6 KPI cards, strategy configuration
  • FastAPI backend with 20+ REST endpoints
  • Telegram integration for alerts and daily summaries
  • SaaS auth with sign-up, login, and onboarding wizard
  • 5. Pricing & Plans

    AltHunter operates as a subscription SaaS platform with three tiers:

    Starter (Free)

  • Shadow-mode trading (paper trades only)
  • 3 active strategies (Lag Exploit, Cluster, Fair Value)
  • Basic dashboard with PnL tracking
  • Daily performance summary
  • 5 AI decoder calls per day
  • Pro ($49/month)

  • Everything in Starter, plus live trading execution
  • All 5 strategies including Market Maker
  • Unlimited AI decoder calls (GPT-4o + Claude)
  • Multi-Agent Arena for evolutionary optimization
  • Real-time exposure management and close mode
  • EMA Oracle with multi-timeframe analysis
  • A/B model routing and auto-escalation
  • Telegram alerts and daily review reports
  • Regression guard with auto-rollback
  • Enterprise ($199/month)

  • Everything in Pro, plus dedicated bot instance
  • Custom strategy parameters and guardrails
  • Agentic wallets with spending policies
  • API access for programmatic control
  • Custom model routing configuration
  • Dedicated support with SLA
  • All plans include a 14-day free trial with full Pro features. No credit card required to start.

    6. Revenue Model

    AltHunter generates revenue through a subscription-first model:

    Monthly Subscriptions: The primary revenue stream. Three tiers (Free, Pro at $49/mo, Enterprise at $199/mo) provide predictable recurring revenue. The free tier serves as a funnel, allowing users to validate the platform in shadow mode before upgrading.

    Performance Fees (Future): A planned 0.5–2% fee on profitable live trades, aligning the platform's incentives with user success. This will be introduced once the live trading mode is fully validated.

    Enterprise Customization: White-label deployments, custom strategy development, and dedicated infrastructure for professional traders and funds.

    Estimated Unit Economics

  • AI API costs per user: $5–15/month (model routing optimizes this)
  • Infrastructure per user: $2–5/month
  • Gross margin at Pro tier: ~65–75%
  • Breakeven: ~25–30 Pro subscribers
  • Cost Optimization

  • EMA Oracle pre-filter reduces AI calls by 30–50% when market is stable
  • A/B model routing uses cheaper models when confidence is high
  • Auto-escalation caps at 5 premium retries per day
  • Cost guardrails track and alert on per-strategy AI spending
  • 7. Roadmap

    Q1 2026 — Core Platform (Complete)

  • Multi-strategy Polymarket trading bot (5 strategies)
  • AI decoder with GPT-4o and Claude integration
  • Market making strategy with EMA oracle
  • Exposure management with close mode and hysteresis
  • Multi-agent evolutionary arena
  • SaaS landing page, sign-up, and onboarding
  • Real-time dashboard with portfolio tracking
  • 222 automated tests passing
  • Q2 2026 — SaaS Launch & Growth

  • Public beta launch with free trial
  • Stripe payment integration (Pro and Enterprise)
  • Live trading mode (graduated from shadow)
  • Agentic wallets with Coinbase CDP
  • Telegram bot for alerts and quick actions
  • Q3 2026 — Advanced Intelligence

  • Multi-oracle price streaming
  • Custom strategy builder (no-code)
  • Backtesting engine with historical replay
  • API access for programmatic trading
  • Q4 2026 — Scale & Mobile

  • Mobile app (iOS and Android)
  • Multi-market support (crypto spot, options)
  • Enterprise white-label solution
  • SOC 2 compliance and security audit
  • 8. Security & Risk Management

    Security and risk management are foundational design principles:

    Exposure Management: - Configurable maximum notional exposure (absolute and % of bankroll) - Per-strategy concentration limits prevent overconcentration - Single-position size caps prevent catastrophic individual bets - Automatic close mode with hysteresis when exposure exceeds thresholds - Close mode blocks new positions and suggests position reductions

    Trading Safety: - Shadow mode validates strategies before committing real capital - Regression guard detects performance drops and auto-disables strategies after 3 consecutive bad days - All parameter changes require human approval above configurable thresholds - Circuit breakers for abnormal market conditions

    AI Safety: - Cost guardrails track and cap AI API spending per strategy and per day - Auto-escalation limited to 5 premium retries per day - Immutable core rules cannot be overridden by AI adjustments - All AI-suggested changes are logged and auditable

    Infrastructure Security: - Encrypted API key management - Rate limiting on all endpoints - Agentic wallet spending policies (per-transaction and daily limits) - Regular dependency audits and updates

    9. Legal Disclaimer

    This whitepaper is for informational purposes only and does not constitute financial advice, investment advice, or any other form of professional advice.

    Trading in prediction markets involves significant risk. Past performance of AI models or autonomous trading agents does not guarantee future results. Users should conduct their own research and consult with qualified financial advisors before making any trading or investment decisions.

    Shadow mode (paper trading) does not involve real money and carries no financial risk. Live trading mode executes real transactions and users are solely responsible for their trading decisions and outcomes.

    AltHunter provides tools and automation, but does not guarantee profits. Market conditions, liquidity, and external factors can impact results. Users should only trade with capital they can afford to lose.

    This whitepaper may be updated from time to time. The most current version will always be available at althunter.com/whitepaper.