84 arbitrage windows opened across CEX and perp venues today, led by an 11.38% HEMI spread between Gate Futures and Hyperliquid. BTR, FLOCK, and 龙虾 (Longxia) rounded out a day dominated by low-cap, thin-orderbook listings — exactly the conditions arb hunters live for, and exactly the conditions that punish anyone who sizes up carelessly.
Eighty-four arbitrage windows crossed the desk today, and the pattern is unmistakable: this was a low-cap and micro-cap day, not a majors day. You will not find BTC or ETH anywhere near the top of this list. Instead, the spreads clustered around newer, thinner-order-book tokens — HEMI, BTR, FLOCK, and the freshly-listed 龙虾 (Longxia) — the exact profile that produces double-digit percentage gaps because a handful of large orders on one venue can blow the local price out of line with the rest of the market.
The headline number is HEMI's 11.38% spread between Gate Futures and Hyperliquid — buy at $0.017730 on Gate, sell at $0.018360 on Hyperliquid. That is a real, executable-looking gap on paper, and it is also a textbook example of why arb traders need to separate 'spread size' from 'spread capturability.' A perp-to-perp spread across two different funding and margin systems is not the same trade as a spot-to-spot transfer, and today's top five make that distinction the whole story.
Ten confirmed opportunities cleared the 7.7% threshold today, all concentrated in four tickers: HEMI (twice), BTR (three times), FLOCK (twice), 龙虾 (twice), and FF (once). When a single asset shows up multiple times with shifting exchange pairs — HEMI on Gate/Hyperliquid, then again on Bitunix/Bybit — that is a signal the token's liquidity is genuinely fragmented across venues right now, not a one-off wick. That is the kind of setup worth building a repeatable playbook around rather than chasing as a single trade.
Bitunix is the connective tissue of today's list — it appears on four of the ten logged opportunities (FLOCK buy leg, BTR sell leg, HEMI both buy and sell legs across two separate trades). That's a strong signal Bitunix's order books on these specific low-cap tickers are systematically thinner or slower-updating than peers, which is exactly the profile arb desks should keep on a standing watchlist rather than treating as one-off noise.
Bybit shows up on both sides of the ledger — as the cheap buy leg on BTR (#4) and as the expensive sell leg on HEMI (#5) and FF. That dual role tells you Bybit's pricing on these tokens moves fast enough to be a source of edge in either direction depending on which way local flow is pushing it, rather than being consistently 'ahead' or 'behind' the market. Binance Futures, by contrast, only ever appears as the higher-priced sell leg (FLOCK, BTR) — consistent with Binance Futures generally carrying a slight premium on newer, hype-sensitive listings due to leveraged long demand.
The 龙屋 (Longxia) pair is worth flagging separately: both its listed spreads route through Gate Futures, once as the cheap leg (buy vs. Binance Futures sell) and once as the expensive leg (sell vs. Exchange51 buy). Gate Futures appears to be currently mispriced relative to at least two other venues simultaneously on this token — a pattern that usually resolves once Gate's own market makers catch up, so this window is more likely to close within the day than to persist.
Today's list splits cleanly into two trade archetypes. The first is the fast, small-size grab: HEMI on Bitunix/Bybit and BTR on Bybit/Bitunix both sit on standard CEX-to-CEX rails with reasonable liquidity depth, meaning you can likely execute $5,000-$15,000 notional with acceptable slippage in a matter of seconds if capital is pre-positioned. These are the trades to run through automated execution — by the time a human confirms the spread on a chart, a meaningful chunk of it is usually gone.
The second archetype is the slow, larger-size play: the HEMI Gate/Hyperliquid spread and the BTR KuCoin/Exchange24 spread both involve either cross-architecture settlement (CEX perp vs. DEX perp) or a smaller venue with unclear depth. These can theoretically support larger notional before slippage kicks in — precisely because fewer bots are competing for them — but the execution risk (bridge delay, thin counter-liquidity, one-sided fills) is materially higher. Position sizing here should scale down, not up, relative to the headline spread percentage: a good rule of thumb is to cap size at whatever the second or third order-book level shows, never the top-of-book quote alone.
General guidance for today's conditions: on the four-way Bitunix/Bybit/Binance Futures cluster, size up to standard limits since fills are fast and rails are proven. On anything touching Exchange24, Exchange51, or Hyperliquid, treat the headline spread as a ceiling, not a target — assume 30-50% slippage decay on size beyond a few thousand dollars until you've personally verified the book depth.
Walking through the HEMI Bitunix→Bybit trade (#5, 8.99% gross) as the cleanest example: on a $10,000 notional buy at $0.012152 on Bitunix, gross spread capture at Bybit's $0.012681 sell price is roughly $899. Subtract taker fees on both legs — a typical 0.10% spot taker fee on each side removes about $10 on the buy and $10 on the sell, for $20 total. If capital must move between exchanges rather than sitting pre-funded on both, add a stablecoin or token withdrawal fee, typically $1-$8 depending on network (TRC-20 is cheap, ERC-20 can run $5-$15 at busy times). Net result: roughly $870-$880 on $10,000 deployed, or an effective net spread around 8.7-8.8% — the fee drag here is small because both legs are liquid CEX rails.
Compare that to the KuCoin→Exchange24 BTR trade (#2, 10.89% gross): same fee structure on the trading side, but the withdrawal-and-settle step to reach a smaller venue like Exchange24 both costs more in absolute terms and — more importantly — costs time, during which the 10.89% figure is a snapshot, not a guarantee. If the spread decays by even 3-4 percentage points during a 10-15 minute transfer window (common on thinner venues), your realized capture on $10,000 can land closer to $650-$750 net rather than the naive $1,050 the headline number implies.
Minimum spread worth chasing: for pre-funded, same-architecture CEX pairs (Bitunix/Bybit/Binance-style trades), anything above roughly 0.4-0.6% gross clears fees with room for modest slippage — that's your bread-and-butter automated range. For trades requiring a live withdrawal or a bridge between different settlement systems (spot-to-perp-DEX, or routing through a smaller exchange), do not engage under 3-4% gross, because fee drag plus decay risk during the transfer window can erase anything smaller. By that standard, all ten of today's logged opportunities cleared the bar on paper — but only the Bitunix/Bybit and Bybit/Bitunix legs clear it with high confidence in practice.
Keep HEMI and BTR on the top of the watchlist — both showed multiple, independent spread events today across different exchange pairs, which is the strongest signal of persistent (not one-off) liquidity fragmentation. Specifically monitor Gate Futures vs. Hyperliquid for HEMI and the full five-way BTR complex (KuCoin, Exchange24, Bybit, Bitunix, Binance Futures) for a repeat pattern in the early session.
FLOCK and 龙虾 are second-tier watches — both are newer listings where market-maker coverage is still filling in, so expect continued but likely narrowing spreads as more liquidity providers arrive. Best times to watch based on today's cluster: the opening hours of the US session and the first hour after Asia opens tend to be when smaller-venue books (Exchange24, Exchange51, Bitunix) lag the majors most, creating the widest windows. Set alerts on the Bitunix/Bybit pair specifically — it produced the cleanest, most executable trade today and is the most likely rail to repeat.
Eighty-four windows, ten worth writing home about, and one clean lesson: the biggest number on the sheet isn't always the best trade on the desk. Pre-fund your rails, respect the order book past level one, and let the fragmented low-caps do the work for you. Watch HEMI and BTR into tomorrow's open.
Arbitrage Hunter — September 2, 2026