Three kinds of number, honestly labeled. Projected assumes today's rates hold for a year (they won't — they change hourly). Backtest replays our exact rules over every day these venues have existed (launched Jan 2026 — no longer history exists, and we never pick a flattering window). The chart shows the combined strategy on your total balance: the band is the execution-cost range on the pair legs (taker floor → idealized-maker ceiling), and the solid line is our actual hybrid plan — rest the quiet leg patiently, take the busy leg instantly. The weekend sleeve is modeled at full costs in all three lines, and the backtest cannot replay the live safety brake. Actual is the live track record, meaningful only after a few weeks. APR is the plain yearly rate; APY adds compounding. Past results never guarantee future ones.

The bot borrows capital from the exchange to amplify the funding income it collects. Higher leverage = more rent per dollar of your balance. Each leg of a pair runs in a separate margin account on its own exchange — the two DEXs do not cross-net margin, so if the underlying price drifts, the losing leg's collateral drains even as the winning leg gains. The rebalance alert fires at 10% leg-price drift; forced liquidation happens at ~18% (1.8× headroom). The table below shows what each level earns and the safety headroom at each step. We recommend the highest level where the warning fires well before forced-exit risk — currently 5×. The Sharpe ratio is constant at all leverage levels; the tradeoff is purely about margin safety.

Every dollar actually banked so far — hourly funding payments collected minus trading costs. It should grind, not jump; big steps usually mean a position was closed and its result got banked.