Arb Desk Report: ONE Rips Three Legs to 24% as OKX Bleeds Liquidity Across the Board

🎯 Arb Desk Report

Forty-seven arbitrage flags crossed the desk today, and the shape of the tape is unmistakable: OKX was the cheapest place to be a buyer almost all session. Five of the ten opportunities we're tracking route through OKX as the buy leg, and three of those are the same ticker — ONE — repricing at different moments as the spread widened, partially closed, and widened again. That's not a single arbitrage opportunity, that's a standing structural dislocation between OKX's order book and everywhere else trading ONE.

The headline print is a 24.03% spread on ONE, buying on OKX at $0.001885 and selling on Bitget at $0.002012. That is an enormous number for a large-cap-adjacent altcoin and it should immediately raise your eyebrows about liquidity depth rather than your pulse about free money — the two usually move together, and not in the direction you want. Behind it, a second ONE print at 20.83% (OKX $0.001922 / Binance Futures $0.002065) and a third at 16.52% (OKX $0.001214 / KuCoin $0.001301) confirm this wasn't a one-off wick; OKX's ONE order book was mispriced relative to the rest of the market for an extended stretch.

Outside the ONE cluster, CNPY delivered the single best non-ONE opportunity at 17.00% between Bitunix and Gate Futures, and JELLYJELLY posted a clean 10.57% between KuCoin and Bybit. Six more setups clustered between 6.49% and 8.77% on GENIUS, G, CROSS, and 龙虾 — smaller edges, but on pairs with generally tighter books, which matters more than the raw percentage once you start sizing a position. Today's session rewards traders who move fast on thin books and traders who are patient enough to let the mid-single-digit spreads run without chasing the scary-big ones into a liquidity trap.

🏆 Top 5 Arbitrage Opportunities

1. ONE — 24.03% spread. Buy OKX at $0.001885, sell Bitget at $0.002012. This is the widest print of the session on the widest-traded ticker in the batch, which is exactly why it's the one to treat with the most suspicion. A spread this size on a token trading in the sub-$0.002 range typically means one venue's book is several price ticks away from the other on size, not that a genuinely fillable path exists to move meaningful capital at that exact print. Liquidity risk: high — at these price levels, OKX's top-of-book depth for ONE is usually shallow enough that even a mid-five-figure buy order walks the price up and erodes half the spread before you're done filling. Withdrawal risk: ONE typically clears OKX-to-Bitget in minutes if the network isn't congested, but any delay beyond 5-10 minutes at this spread size is enough for Bitget's ask to have moved on you. Our take: executable in small size (low four figures) for a trader who already has capital pre-positioned on both venues; not executable as a discovery-to-fill trade because the window this wide rarely survives the time it takes to move funds cross-exchange from scratch.

2. ONE — 20.83% spread. Buy OKX at $0.001922, sell Binance Futures at $0.002065. Note this is a spot-to-futures arb, not spot-to-spot — you're not actually delivering the same instrument on both legs, you're buying spot ONE on OKX and selling (or shorting) the futures contract on Binance. That changes the risk profile entirely: you're not settling a matched pair, you're putting on a basis trade that only converges to your locked-in spread if you eventually close both legs, and in the meantime you're exposed to Binance funding rate flips against a short futures position. Liquidity is somewhat better here than the pure spot version because Binance Futures order books are typically deeper than Bitget spot at this size. Our take: executable as a basis trade for someone comfortable running short futures exposure overnight, but it is not a simple buy-here-sell-there arbitrage — treat the funding rate as a cost line, not a footnote.

3. CNPY — 17.00% spread. Buy Bitunix at $0.526764, sell Gate Futures at $0.568200. This is the standout non-ONE trade of the day and, notably, a token where a full dollar of price gives you real tick granularity to work with — much easier to place limit orders precisely than on the sub-cent ONE prints. Bitunix is a thinner, higher-fee venue than the majors, so slippage risk sits on the buy leg rather than the sell leg; Gate Futures is comparatively liquid for a token this size. Withdrawal consideration: if you're moving CNPY itself rather than running a basis trade, confirm Bitunix's withdrawal isn't gated behind a minimum holding period, which is common on newer-listing tokens on smaller exchanges. Our take: the most legitimately executable trade in the top five — real price granularity, one thin leg and one liquid leg, and a spread wide enough to survive normal fee drag.

4. ONE — 16.52% spread. Buy OKX at $0.001214, sell KuCoin at $0.001301. Third ONE print of the session and notably at a lower absolute price than the two above it — this looks like the tail end of the same OKX mispricing working its way down as OKX's book got hit repeatedly by arbitrageurs closing the earlier, wider spreads. KuCoin is a reasonably deep venue for ONE, which helps the sell side; the constraint is still OKX's buy-side depth at sub-$0.0013. Our take: executable in the same small-size band as opportunity #1 — treat this less as a fresh signal and more as confirmation that OKX's ONE market was broadly dislocated for a sustained window today, which is useful information for tomorrow's setup even after this specific print closes.

5. ONE — 13.04% spread. Buy OKX at $0.001224, sell Bitget at $0.001301. The fourth ONE appearance in the top five, and the narrowest of the cluster — this is the spread you'd expect to see as the dislocation compresses back toward normal. Fee-adjusted, this is close to the minimum-viable threshold discussed below, so execution quality (limit vs. market orders, minimizing the number of hops) matters more here than on the fatter prints. Our take: marginal — worth taking only if you already have inventory positioned on both OKX and Bitget; not worth initiating fresh capital movement for a 13% headline spread once fees and withdrawal time are priced in.

📊 Exchange Spread Patterns

The dominant pattern today is OKX-as-cheap-buy-side, full stop. OKX appears on the buy leg in five of the ten flagged opportunities — all four ONE prints in the batch. That's a single-venue, single-asset story more than a broad structural pattern across the exchange landscape, and it's worth asking whether OKX's ONE market maker stepped back or whether a large sell order sat on OKX's book without being arbed away fast enough by the usual bots. Bitget shows up on the sell side twice (both against OKX-bought ONE) and as a buy-side venue once (G, buying at $0.004655 against Binance Futures at $0.004980), suggesting Bitget's books were generally priced fair-to-rich relative to peers rather than showing a directional bias. Gate Futures appears on both sides of the ledger — buying GENIUS cheap against Binance Futures, and selling CNPY rich against Bitunix — which tells you Gate's futures pricing was simply tracking spot fundamentals accurately while the counterparties on each trade were the ones out of line, not Gate itself. Binance Futures shows up as the sell (rich) side three times, against OKX ONE, Gate GENIUS, and Exchange24's 龙虾, a pattern consistent with Binance Futures carrying a persistent slight funding premium today that pulled its mark price above spot-referenced venues. We didn't see a Hyperliquid entry in today's batch, so no read on the perp-DEX-vs-CEX basis this cycle — worth flagging for tomorrow since that pair often produces the cleanest, most mechanically explainable spreads.

⚡ Speed vs Size Analysis

The two ends of this book require completely different playbooks. The sub-cent ONE spreads (13-24%) are fast, thin, and unforgiving on size — the percentage looks enormous, but the absolute price per unit is a fraction of a tenth of a cent, meaning the depth available at the quoted price is typically exhausted by a surprisingly small notional order. Trying to push five figures through a 24% ONE spread will realistically capture a fraction of that headline number once you've walked both books; this category rewards speed and small clip sizes over trying to maximize dollar capture. The mid-range prints — CNPY at 17%, JELLYJELLY at 10.57%, GENIUS at 8.77% — sit at price levels with real decimal granularity ($0.05-$0.57), which means slippage per dollar traded is meaningfully lower and you can size up without your own order being the thing that closes the spread. Our position sizing rule of thumb for this batch: cap ONE-cluster trades at whatever size your historical fill data shows moves OKX's book by less than 2-3%, and size CNPY/JELLYJELLY/GENIUS trades based on standard 1-2% of available top-three-levels depth rather than a fixed dollar cap. Chasing the biggest printed percentage with the biggest size is the single fastest way to turn a paper 24% into a realized 3% after slippage on both legs.

💰 Profit Calculations

• CNPY example — Gross spread: buy Bitunix $0.526764, sell Gate Futures $0.568200 = 7.87% mechanical gross ((568200-526764)/526764), well below the 17.00% headline figure because listed spread metrics often reference deeper/rolling book levels rather than best-bid-best-ask; always recompute off the actual fill price you'd get, not the flagged percentage.
• Trading fees both sides — assume 0.10% taker on Bitunix buy + 0.05% taker on Gate Futures sell (typical futures taker tiers) = 0.15% combined, leaving roughly 7.72% after trading fees alone.
• Withdrawal / transfer cost — if you're physically moving CNPY between venues rather than running a basis trade, budget the network's flat withdrawal fee as a percentage of trade size: on a $1,000 clip a $2-3 flat withdrawal fee is ~0.25-0.3%, shrinking further as size increases. On a basis-trade version (long spot, short futures, no token transfer) this line drops to zero but you inherit funding-rate exposure instead.
• Net profit, CNPY spot-to-spot on $1,000 notional: ~7.72% minus ~0.3% withdrawal ≈ 7.4% net, or roughly $74 — a real, bankable arb if execution matches the assumed fill prices.
• ONE example (24.03% headline, OKX $0.001885 / Bitget $0.002012) — mechanical gross is 6.74%, not 24.03%; after 0.10% + 0.10% taker fees (0.20% total) and a flat withdrawal fee that on a sub-cent token can eat 1-3% of trade value depending on network gas at time of transfer, net profit compresses to roughly 3.5-5.5% — still worthwhile, but a fraction of the flagged number.
• Minimum spread worth chasing — as a floor, we don't act below 1.5-2% mechanical (best-bid/best-ask) spread after accounting for ~0.2% round-trip taker fees and typical withdrawal drag; below that, normal price noise and execution latency has a real chance of erasing the edge entirely before both legs settle.

⚠️ Risk Alerts

• Headline spread percentages in this feed run well above what a direct best-bid/best-ask calculation produces — always verify against live order book depth before committing capital, don't trade off the flagged number alone.
• OKX ONE liquidity: four separate ONE opportunities today all use OKX as the cheap buy side, which smells like a persistently thin or stale OKX book rather than four independent opportunities — expect your actual fill price on OKX to be worse than the quoted $0.001885/$0.001922/$0.001214/$0.001224 the moment you place size.
• Sub-cent token withdrawal delays: ONE, CNPY, and JELLYJELLY are all priced low enough that network congestion or exchange-side withdrawal queues can turn a 10-minute transfer into an hour, during which the sell-side price is fully exposed to market movement.
• Spot-vs-futures legs (the OKX/Binance Futures ONE trade and the Gate Futures/Binance Futures GENIUS trade) are basis trades, not matched arbitrage — funding rate flips can erase the locked-in spread if you hold the position past a funding interval.
• Bitunix and smaller venues in the CNPY leg often carry higher withdrawal minimums or holding-period restrictions on recently listed tokens — confirm before assuming instant transferability.
• No volume figures were reported alongside these spreads this cycle (feed shows $0.0M across pump/dump/buy/sell pressure) — treat every spread above as unconfirmed for size until you've checked live order book depth yourself; do not assume six-figure liquidity behind any of these prints.

🔮 Tomorrow's Setup

Keep OKX's ONE book on a dedicated watch — a venue that mispriced the same asset four times in one session against four different counterparties (Bitget twice, Binance Futures, KuCoin) is a candidate to do it again until whatever's causing it (a stale market maker, a delisting rumor, a liquidity provider pulling back) resolves. Watch the OKX-Bitget and OKX-Binance Futures pairs specifically in the first hours of the Asia session, when OKX order books have historically thinned out before Western market makers refresh quotes. On the mid-tier side, CNPY and JELLYJELLY both showed real, trade-worthy spreads today on venues (Bitunix, Gate, KuCoin, Bybit) that aren't chronic mispricers — if either token sees continued listing-driven volatility, expect follow-through spreads in the same pairs rather than new venues. GENIUS and CROSS both showed futures-vs-spot basis dislocations against Gate Futures and Bitget respectively; if Binance Futures funding stays elevated overnight, that basis gap on GENIUS is likely to persist or widen into tomorrow's session. Best watch windows: the first 30-60 minutes after major exchange funding settlements (00:00 and 08:00 UTC) and immediately following any large single-exchange order book sweep, which is usually what kicks off these dislocations in the first place.

Sign Off

Forty-seven flags, one venue doing most of the damage, and a reminder that the biggest number on the screen is rarely the best trade on the desk — the real edge today was in the boring, deep-decimal CNPY print, not the flashy 24% headline. Size small on the thin books, size normal on the thick ones, and always check the order book before you check the headline percentage. Arbitrage Hunter — September 18, 2026.