101 arbitrage windows crossed the scanner on September 4, with MARSCOIN's 34.72% Bitunix-to-Exchange51 spread topping the board, backed by repeat setups on BR, BULLA, USELESS, STX, FLOCK and CP across majors and thinner venues.
One hundred and one arbitrage windows crossed the desk today. That's a busy tape by any measure, and it's the kind of day where a scanner earns its keep — most of the printed spreads clustered in the routine 3-9% band that professional arb books chew through as background noise, but a handful of outliers made this session worth writing up in full. The headline number is MARSCOIN at a 34.72% spread, buying on Bitunix at $0.169296 and selling on Exchange51 at $0.177000. That is not a typo-sized number for a legitimate cross-exchange spread, and it sets the tone for the rest of the report: today was a day for thin-book microcaps and newer venues throwing off outsized prints, not a day where BTC and ETH majors were doing anything interesting.
MARSCOIN alone accounted for two of the top three entries on the board — the 34.72% Bitunix/Exchange51 print and a second, calmer 13.52% Binance Futures/Exchange51 print later in the session — which tells you this wasn't a one-off fat-finger tick but a token with a genuinely dislocated order book relative to the majors. BR also showed up three separate times in the top-ten feed, all variations on a Binance Futures or Bitunix buy against a Bybit or Binance Futures sell, which is the kind of repeat-offender pattern that arb desks build standing bots around rather than manually chasing. The rest of the board — BULLA, USELESS, STX, FLOCK, CP — rounded out a session that skewed heavily toward futures-vs-futures spreads on lower-cap alts, with only STX (Binance spot vs. Coinbase) and CP (OKX vs. KuCoin) representing more traditional spot-to-spot arb structure.
The dominant pattern today was Bitunix and Gate Futures sitting on the cheap side against Binance Futures and Bybit on the expensive side. Bitunix appeared as the buy-side venue in three of today's top prints (MARSCOIN, USELESS, and one of the BR entries), which is consistent with what we'd expect from a venue that's been rapidly expanding its futures listings — new or thinner pairs there tend to lag price discovery on the majors by a few basis points to a few percent, and today that lag showed up repeatedly. Binance Futures did double duty: it was the buy-side venue against Exchange51 on MARSCOIN, but the sell-side venue against Bitunix on both BR and USELESS. That's not a contradiction — it just confirms Binance Futures sits in the middle of the liquidity stack, deeper than Bitunix and Gate Futures but still capable of trading at a discount to a thinner, more speculative venue like Aster.
Exchange51 (Aster) was the standout premium venue of the session, appearing as the sell-side leg on both MARSCOIN prints and again on FLOCK's 9.84% spread against Binance Futures. Three top-ten appearances for one newer venue in a single day is a real pattern, not noise — Aster's book depth on smaller-cap perps is thin enough that its mark price runs persistently above the majors, and any arb desk running size into that venue should expect the effective sell price to degrade as fill size increases. On the more traditional side, STX's 9.90% Binance-to-Coinbase spread is a textbook Coinbase premium print — U.S. spot demand pushing Coinbase's book above Binance's global liquidity pool, a pattern that shows up reliably during U.S. trading hours. CP's 9.70% OKX-to-KuCoin spread is smaller-venue-to-smaller-venue, and worth flagging separately since neither side has Binance-tier depth — this is a pair to size conservatively regardless of the headline percentage.
There's a real tradeoff on display across today's board between the BULLA/BR-style setups (major-venue liquidity, mid-teens-to-low-teens spread, executable at real size) and the MARSCOIN/FLOCK-style setups (thinner venue on one leg, eye-catching headline percentage, execution risk concentrated on the exit). The general rule holds: the bigger the printed spread on a low-cap or newer-venue pair, the more of that spread exists only on paper, because it's measuring the gap between a resting quote and a price that will move against you the moment you try to trade real size into it. A 34.72% MARSCOIN print and an 11.53% BR print are not the same trade wearing different clothes — one requires you to estimate how much of that number survives contact with Aster's actual order book, the other is close to fully realizable because both legs sit on venues that can absorb institutional-scale flow without much slippage.
Position sizing should scale inversely with how thin the shallower leg is, not with the headline spread. For the BULLA/BR/USELESS-style trades running through Binance Futures and Bybit, working in a small number of larger clips is reasonable — the books can take it. For the MARSCOIN/FLOCK-style trades touching Exchange51/Aster, the right approach is smaller clips worked over a longer window, checking realized fill price against the quoted price after each clip, and being willing to walk away once realized slippage eats a meaningful chunk of the edge. As a rough heuristic: if you can't get a live depth read on the thinner leg before committing capital, cap your size at whatever you'd be comfortable losing entirely to slippage, because on names like these that's a real possibility, not a tail case.
Take the BULLA setup as the clean-liquidity example: buy Gate Futures at $0.033635, sell Binance Futures at $0.038287, gross spread 13.83%. On $10,000 of buy-side notional, that's roughly $1,383 of gross edge before costs. Futures taker fees on both venues run in the neighborhood of 0.04-0.06% per side for a standard (non-VIP) account — call it 0.05% on Gate Futures and 0.04% on Binance Futures, or roughly $9 combined on this notional. Because this is a futures-to-futures basis trade run on pre-funded margin at both venues, there's no on-chain withdrawal leg and no withdrawal fee to subtract — the capital just sits, funded, on both exchanges permanently. Net result: roughly $1,374 on $10,000, or about 13.73% net — the fee drag on a spread this size is close to irrelevant.
Now compare that to a spot-to-spot structure like CP on OKX vs. KuCoin at a 9.70% spread, where the token has to physically move between exchanges. Spot taker fees typically run 0.08-0.10% per side (roughly $16-20 combined on $10,000), and then there's the withdrawal fee and, more importantly, withdrawal time — anywhere from a few minutes to over an hour depending on network congestion and each exchange's confirmation requirements, during which the spread can compress or flip entirely. That time exposure is the real cost, not the flat withdrawal fee itself. Rule of thumb for this desk: spreads under roughly 1.5-2% aren't worth chasing once you account for fees and the realistic risk of adverse price movement during a withdrawal window; spreads in the 3-6% range are worth taking on venues with genuine depth; anything above that on a thinner or newer venue should be treated as a slippage-adjusted number, not a face-value one — assume you'll realize a fraction of the headline spread and size accordingly.
Exchange51/Aster liquidity is the single biggest watch item from today's session — it was the premium leg on three separate top-ten prints, and thin order books on a newer perp venue mean the quoted price and the realized fill price can diverge sharply once you're moving real size. Bitunix showed up repeatedly on the buy side across MARSCOIN, USELESS, and BR — worth watching for any listing changes, maintenance windows, or funding-rate anomalies on that venue given how often it's generating the cheap leg of today's spreads. Standard cross-exchange caveats apply everywhere a spot leg is involved (STX, CP): confirm each exchange's current withdrawal status for the specific asset before committing capital, since deposit/withdrawal suspensions on a token can trap one leg of a trade mid-execution. And as a blanket rule for any headline spread above roughly 15-20%, treat the printed percentage as a starting hypothesis to verify against a live order book rather than an executable price — data feeds can lag on thinner pairs, and a spread that looks too good relative to the raw prices involved usually is.
BR is the name to watch first thing — three appearances in today's top ten on the same Binance-Futures-buy / Bybit-sell axis suggests a persistent mispricing rather than a closed one-off, and it's the kind of pair worth checking again at the next funding interval. MARSCOIN and FLOCK's shared exposure to Exchange51/Aster makes that venue worth a standing watch generally — if Aster's books stay thin relative to the majors, expect more premium prints there across whatever alt perps it's carrying. On the spot side, STX's Coinbase premium tends to be a U.S.-trading-hours phenomenon, so the window to watch there is the U.S. morning session when Coinbase spot demand typically outpaces Binance's global flow. Keep Gate Futures vs. Binance Futures and OKX vs. KuCoin on the standing monitor list as well — both pairs printed clean, fee-survivable spreads today, and neither is dependent on a single volatile microcap staying dislocated to keep working.
One hundred and one prints, one genuinely wild number out of MARSCOIN, and a handful of boring, fundable spreads sitting quietly underneath it where the real money gets made. Verify your books before you fire, size to the thin leg, and don't let a 34.72% headline talk you into forgetting that Aster's order book doesn't care what the scanner says. Arbitrage Hunter — September 4, 2026.