# APEX Runner — Trading Methodology

**Version:** 2.0 (May 2025)  
**Status:** Live trading on Kraken, Coinbase, Hyperliquid  
**Website:** https://apexrunner.ai

---

## Overview

APEX Runner is an autonomous AI trading system that combines four complementary strategies: grid trading, DCA accumulation, momentum breakouts, and funding rate arbitrage. Each strategy is assigned to market regimes detected in real time. Claude (Anthropic) serves as the reasoning engine for trade decisions, regime classification, and weekly strategy evolution.

---

## Market Regime Detection

Before every cycle, APEX classifies the current market into one of five regimes:

| Regime | Conditions | Strategies Allowed |
|--------|-----------|-------------------|
| RANGING | Low volatility, price oscillating | Grid, Mean Reversion |
| CHOPPY | High noise, no trend | Grid, Mean Reversion |
| TRENDING | ADX>30, vol>1.5×, BB expanded | Grid, Momentum (gated) |
| CRISIS | Sharp drawdown, F&G<15 | DCA only |
| NEUTRAL | Default/unclear | Grid only |

**TRENDING regime gate** (5 factors all required):
- ADX > 30
- Volume > 1.5× 20-period average
- Bollinger Bands expanded
- Whale wallet confirmation
- Fear & Greed between 35–75

---

## Strategy 1: Grid Trading (Primary)

Grid trading is the core revenue engine — 100% win rate on completed cycles across 2,892+ cycles since deployment.

**How it works:**
- Places buy and sell limit orders at fixed intervals above and below current price
- ATR-spaced levels: `max(1.0%, 1.5 × ATR)` — widens automatically in volatile markets
- When a buy fills, a sell is placed immediately at buy price + spacing
- When a sell fills, a new buy is placed below at sell price - spacing (true bidirectional)
- Grid rebuilds only when price drifts more than 3× spacing from center

**Pairs:** BTC, ETH, SOL, AVAX  
**Level sizes:** Kraken $25 BTC/ETH/SOL, $15 AVAX | Coinbase $8.38 BTC/ETH/SOL, $8 AVAX | Hyperliquid $5  
**Levels:** 8 per side  
**Max deployed:** 50% of portfolio  
**Compounding:** Every 20 cycles, level size increases 5% (capped at 2× base)

---

## Strategy 2: DCA Accumulation

Dollar-cost averaging triggered by extreme fear — buying when others panic.

**Entry conditions (ALL required):**
- Fear & Greed Index ≤ 20
- RSI < 45
- Stochastic Oscillator < 50
- Price below 20-period SMA
- Cash reserve > 15%

**Risk controls:**
- 20% portfolio cap on all DCA positions combined
- 5-position maximum open simultaneously
- 1% risk per trade, ATR-based sizing

**Data source:** CoinMarketCap Pro API with 55-minute cache; alternative.me fallback; neutral (50) hardcoded fallback on dual failure.

---

## Strategy 3: Momentum Engine

High-conviction directional trades during confirmed trends.

**Entry conditions (7 factors):**
1. Regime classified as TRENDING
2. ADX > 30
3. Volume > 1.5× average
4. Bollinger Band expansion confirmed
5. Whale wallet activity confirms direction (top-20 Hyperliquid leaderboard)
6. Fear & Greed 35–75 (not extreme in either direction)
7. RSI not overbought/oversold

**Sizing:** 5% of portfolio (CEX), 8% margin (Hyperliquid)  
**Pairs:** BTC, ETH (CEX) | BTC, ETH, SOL, AVAX (Hyperliquid)

---

## Strategy 4: Funding Rate Arbitrage (Hyperliquid)

Captures funding payments from perpetual futures positions.

**How it works:**
- Scans all 229 Hyperliquid perpetual markets every cycle
- Enters when funding rate > 0.05% per 8 hours
- Position sizing scaled by Fear & Greed: 1.25× in fear, 0.75× in greed
- Neutral to market direction — profit comes from funding, not price movement

---

## Risk Management

**Portfolio heat monitor:**
- Pause new entries at 5% unrealized portfolio loss
- Block all new entries at 8% unrealized loss
- Emergency flatten at 12% unrealized loss

**Per-trade risk:**
- 1% max risk per signal trade
- 10% max position size
- 2% take profit, 6-hour max hold on mean reversion trades

**Cash reserve:** Minimum 15% always maintained

---

## AI Reasoning Layer

Three Claude models serve distinct roles:

| Model | Role | Frequency |
|-------|------|-----------|
| claude-sonnet-4-20250514 | Trade decisions, regime classification | Every trade cycle |
| claude-opus | Weekly strategy evolution, urgent escalation | Sundays + as needed |
| claude-haiku | Daily reports, news scanning | Daily + per news event |

**Weekly evolution:** Every Sunday 6am UTC, Opus reviews all trade outcomes from the past week, identifies patterns, rewrites strategy skill files, and updates `apex_memory.json`. The system learns from its own performance.

**Learning layer:** Every trade outcome is logged to SQLite with entry/exit prices, regime, strategy, P&L, and market conditions. This corpus feeds the weekly evolution.

---

## Signal API

APEX exposes 24 trading signals via REST API, priced via x402 protocol (USDC on Base mainnet).

**Available signals (sample):**
- `fear_greed` — Current F&G index with APEX interpretation ($0.05)
- `regime` — Current market regime classification ($0.10)
- `grid_status` — Active grid levels and fill rate ($0.10)
- `momentum_status` — Momentum engine state ($0.25)
- `combined_alpha` — Full signal composite ($1.00)

**Endpoint:** https://apexrunner.ai/signals/  
**Payment:** USDC on Base mainnet (eip155:8453)

---

## Track Record

- **Kraken grid:** 2,892+ completed cycles, 100% win rate on closed positions
- **Combined APEX-managed capital:** ~$2,017 across three exchanges
- **AI cost:** ~$2–4/day (reduced 68% via token optimization in May 2025)

---

## Legal

APEX Runner is software infrastructure. It does not manage assets, provide investment advice, or hold client funds. Clients connect their own exchange accounts via trade-execution API keys only. No withdrawal permissions are requested or accepted. Comparable to 3Commas, Cryptohopper, and similar algorithmic trading platforms.

---

*Last updated: May 2025 | Contact: hello@apexrunner.ai*