solana-arb

What I do

• •Guide Jupiter v6 API integration for quoting, routing, and swap execution
• •Document arbitrage opportunity detection and profit calculation with failure economics
• •Define MEV protection strategies with Jito bundles and anti-sandwich patterns
• •Provide bot architecture patterns, monitoring, and honest testing strategies

When to use me

Use this skill when building trading bots or arbitrage systems on Solana. Pair with solana-core for transaction construction and RPC patterns, rust-pro for Rust engineering, and perf-rust for hot-path optimization.

Jupiter v6 Integration

• •Quote API: GET /quote with inputMint, outputMint, amount, slippageBps
• •Swap API: POST /swap with quoteResponse and userPublicKey
• •Route selection: ExactIn vs ExactOut. Use maxAccounts to constrain tx size. onlyDirectRoutes is faster but may miss better multi-hop routes.
• •Direct AMM integration: interact directly with Raydium/Orca/Phoenix pool programs for lower latency (skip Jupiter API round-trip)
• •Concurrent quoting: tokio::task::JoinSet to quote multiple routes simultaneously

Arbitrage Opportunity Detection

Decision tree: Is an opportunity real?

Price discrepancy detected
  -> Is quote < 2 slots old? NO -> discard (stale)
  -> Does simulated profit > all-in costs? NO -> skip
  -> Is the route still valid? NO -> re-quote
  -> Execute

Latency matters: the first bot to land the transaction captures the opportunity.

Latency Budget Framework

Methodology: measure each phase with std::time::Instant, log p50/p95/p99, optimize the widest phase first (same principle as perf-core).

Profit Calculation

Failed Transaction Economics

Solana arb failure rate: ~30-40% of transactions fail (dropped, expired, front-run). Model this as a first-class business metric.

• •Expected profit = (gross_profit x success_rate) - (tx_cost x total_attempts)
• •A bot with 60% success rate and 0.001 SOL/tx cost needs >0.0017 SOL gross profit per trade to break even
• •Track failed tx cost as a first-class P&L metric

Atomic Execution & Jito Bundles

• •Atomic execution: all swap instructions in a single transaction -- all succeed or all revert. Order: compute_budget_ix -> swap_ix_1 -> swap_ix_2
• •Jito bundles: submit transactions as an atomic bundle to the Jito block engine. Use when gross_profit > tip_amount AND trade is sandwichable.
• •Tips: check Jito tip floor for minimum viable tip. Use Jito's published tip account addresses. Competitive tips for high-value trades.
• •Searcher API: gRPC connection to Jito block engine for bundle submission

Risk Management

MEV Defense Checklist

• • Use Jito bundles for all trades above minimum profit threshold
• • Set per-route slippage limits (not a global default)
• • Never broadcast swap intent via public mempool
• • Verify tip amount covers bundle inclusion probability
• • Monitor for unusual slippage patterns indicating adversarial activity
• • Implement circuit breaker: stop trading after N consecutive losses
• • Use separate fee-payer keypair from main trading keypair

Bot Architecture

Async event loop: stream(account_updates) -> filter(price_divergence > threshold) -> quote(jupiter OR direct_amm) -> simulate -> execute(jito OR direct_send)

• •Concurrent quoting: tokio::task::JoinSet for multiple routes simultaneously
• •State management: track in-flight transactions (avoid double-execution), cooldown per pool pair
• •Graceful shutdown: drain in-flight transactions, log final P&L, do not open new positions

Testing Strategies

The Devnet Lie: devnet pools have no real liquidity and no competing bots. A strategy profitable on devnet tells you nothing about mainnet viability. Start with localnet for correctness, skip to mainnet paper trading for strategy.

Monitoring & Observability

Anti-Patterns

Companion Skills