·4 min
为什么加密套利里 10 毫秒数据碾压 100 毫秒
套利窗口只持续数秒。延迟不是营销数字，而是谁先成交。
Cross-exchange spreads do not wait for REST polls. When a thin alt reprices on one venue, the opposite book often lags for a few hundred milliseconds to a couple of seconds — that is the window.
A free tier that refreshes every ~100 ms still sees the move. A Pro WebSocket stream at ~10 ms sees the same move earlier and more often, which matters when you also need to size against live depth before the window closes.
Arbitrex separates research (Free) from race conditions (Pro) on purpose: the scanner is useful either way, but automated triggers on narrow edges need the faster book.