Solana DEX Arb
RESEARCHAtomic multi-hop DEX arbitrage on Raydium + Meteora (1356 pools). Cycle detection, Jupiter quote validation, Jito bundle execution.
Key Numbers
At a Glance
1,356
Pools Indexed
Multi-Hop
Cycle Detection
Atomic
Execution
4-Level
Edge Tracking
Overview
About This Project
An atomic multi-hop DEX arbitrage system operating across Raydium and Meteora liquidity pools on Solana. The system indexes 1,356 pools spanning AMM, CPMM, CLMM, and DLMM pool types, computing profitable arbitrage cycles in real-time.
The cycle detection engine finds multi-hop paths (A -> B -> C -> A) where the cumulative exchange rate yields a profit after accounting for swap fees, slippage, and transaction costs. A 4-level edge tracking system classifies opportunities from raw to executable, filtering out theoretical edges that collapse under realistic execution constraints.
Jupiter's routing engine serves as a truth source for quote validation -- any arbitrage opportunity that doesn't survive Jupiter's independent quote check is discarded. Execution uses Jito bundles for atomic, MEV-protected multi-hop swaps that either complete entirely or revert with no partial risk.
Features
What It Does
Multi-Pool Math Engine
Native implementations of AMM constant-product, concentrated liquidity (CLMM), and dynamic liquidity (DLMM) pool math for accurate quote computation without RPC calls.
Cycle Detection Algorithm
Graph-based cycle finder discovers profitable multi-hop paths across 1,356 pools, pruning unprofitable cycles early using conservative fee and slippage estimates.
Jupiter Truth Source Validation
Every detected arbitrage opportunity is validated against Jupiter's independent routing engine, eliminating false positives from stale pool state or math discrepancies.
Jito Bundle Submission
Atomic execution of multi-hop swap sequences via Jito bundles ensures all-or-nothing execution with MEV protection against sandwich attacks.
Architecture
How It Works
Challenges
What Made This Hard
Pool state on Solana changes every 400ms, meaning arbitrage edges are extremely ephemeral. The system must recompute cycle profitability faster than the state changes, while accurately modeling slippage across concentrated liquidity pools where the price impact function is nonlinear. Handling four distinct pool types (AMM, CPMM, CLMM, DLMM) each with different mathematical models in a unified framework required building custom swap simulators for each.
Stack