F. HEKİMOĞLU
← cd ~/work/quant
// QUANT · ARBITRAGE, NET OF COSTS

Crypto arbitrage where the edge collapses after fees

A fee-, depth- and transfer-cost-aware cross-exchange and triangular scanner with a gross-to-net waterfall. Raw spreads look profitable; the net executable edge is negative on liquid pairs.

00Overview

Cross-exchange price gaps look like free money until you account for what it costs to actually take them. This scanner builds the full gross-to-net waterfall, taker fees, order-book depth slippage and network transfer cost, and shows that the executable edge on liquid pairs is negative.

The waterfall

NET EDGE AFTER COSTS, by manufactured gap (bps)
-51
no gap
-26
+25 bps gap
-11
+40 bps gap

01Method

  • Cross-exchange and triangular arbitrage detection.
  • A taker-fee model, order-book depth slippage via a VWAP depth-walk, and network transfer cost.
  • A gross-to-net waterfall decomposition that shows where the spread goes.
  • Deterministic synthetic books for reproducibility, with live data via CCXT and graceful degradation.

02The honest finding

Round-trip taker fees alone (36 bps at default) exceed almost every executable gross spread on a liquid pair. Even a manufactured +40 bps dislocation nets -11 bps after costs. The verdict across scenarios is no feasible edge.

Round-trip fees36 bps
Best net (+40bps)-11 bps

03Stack & links

pythonccxtarbitrageorder-bookhypothesis
SEC 00 · INDEXENDARK--:--:--⌘K / ? help