We've already covered market monitoring (Scout) and mathematical models (AMM Math). Now, the main event—strategy layer (Strategy). This is the brain of the arbitrage bot, which needs to answer three key questions within milliseconds:

Is there money to be made? Is the price difference enough to cover gas and slippage? How should it operate? Should it buy low in Pool A and sell high in Pool B, or do the reverse? How to ensure the profit is actually realized? Lock in execution through atomic transactions and bundling.

In this episode, we'll delve into the algorithmic logic and practical engineering challenges of cross-DEX arbitrage on Solana.

**From Polling to Event-Driven**

Solana's high-performance environment dictates that arbitrage architecture must be reactive. When the pool state updates, the bot immediately enters the workflow: state change triggers → local quote calculation → compare price difference → check if it exceeds the threshold → if yes, determine direction and size → construct atomic transaction instructions → send via Jito Bundle. Conversely, if the price difference doesn't meet the threshold, it simply abandons and waits for the next opportunity.

This architecture compresses the time from listening to transaction execution into milliseconds—faster than you can imagine. It's not polling the pool state or waiting; it's truly event-driven.

**Why Use Jito Bundle**

In the Solana ecosystem, your transactions still need to queue. Jito Bundle breaks this queuing logic. The bot can bundle multiple transactions into a single atomic operation and send it to the validator, ensuring execution order and atomicity. Once in a bundle, either all succeed or all fail—no intermediate states. This is crucial for arbitrage—because you can't accept a scenario where the first transaction succeeds and the second fails.

**Core of the Decision Logic**

Price difference detection is the first hurdle. The bot must calculate all possible arbitrage paths (possibly two pools, three pools, or even more) in real-time, then select the one with the largest spread. But there's a catch: trading itself causes slippage, gas fees are fixed costs, and network congestion can cause bundle fee rates to spike. So, all these factors must be included in the calculations to ensure the final profit is positive.

Next is the direction decision. Suppose a USDC/SOL price difference appears; the bot needs to determine which pool to buy in and which to sell in. This involves liquidity depth, trading size limits, and dynamic slippage. Too small a size, and there's no profit; too large, and slippage eats into gains. The goal is to find that critical point.

The entire closed loop continuously cycles—from detecting opportunities to executing orders—with almost no delay.