When High-Profit Windows OpenFour days of live stablecoin spread trading

What the four-day ledger shows

The observation window is 27 October 2026, 15:30, to 31 October 2026, 15:30, in Los Angeles (PDT): 27 October, 22:30, to 31 October, 22:30 UTC. The comparison is the immediately preceding 96 hours. These are fixed historical windows, not a rolling live dashboard.

Eight anonymized accounts contributed completed traded cycles. One basis point is 0.01%. For this note, a high-bps cycle means a completed traded cycle earning at least 2 bps after recorded trading fees.

Buy-leg notional in the observed window totaled USD 5,462,573.25. Strategy PnL deducts actual fees on both trading legs, values USD and USDC at 1:1, and is before platform service fees. Notional includes repeated use of the same funds; it is not invested capital. Cycle bps are not an account return or APY.

The 683 high-bps cycles made up 19.4% of completed traded cycles and contributed 257.99 USDC, or 36.1% of net strategy PnL. The extra profit came from a group of actual fills.

The window opens and then fades

Each row runs from 15:30 on the first date to 15:30 on the next date, in Los Angeles time. The 29–30 October window earned 1.56 bps on a notional-weighted basis, with 373 cycles reaching at least 2 bps. Over the following 24 hours, the weighted result fell to 0.42 bps, with no cycles reaching 2 bps.

Higher-profit windows appeared more than once during these four days, and they did not persist. The final day’s contraction is part of the result.

Wider realized spreads explain the direct improvement

nomantrade buys USDT in the USDT-USD market and sells it in USDT-USDC. A quoted opportunity becomes realized profit only when the actual fills leave enough spread to cover costs.

The realized gross spread rose from 1.70 bps to 1.91 bps, while recorded trading fees stayed near 0.60 bps in both windows. Almost all of the roughly 0.21-bp improvement in the net result corresponds to a wider realized buy–sell spread.

Public candles point in the same direction. Across fully contained hourly observations, the mean USDT-USDC close relative to the USDT-USD close increased from about 1.51 bps to 1.84 bps. This is a relative-pricing measure, not a simultaneously executable quote.^[1]

Excluding the account with a special target setting, the weighted net spread still rose from 1.12 bps to 1.29 bps. Among cycles with the same 0.3-bp target, it rose from 1.07 bps to 1.19 bps. The sample mix still changed, but a higher configured threshold alone does not explain the improvement.

The Binance incident may have amplified the later window

Binance confirmed the 29 October security incident and temporarily suspended withdrawals. TRM Labs places the first anomalous transfers at 20:05 UTC, or 13:05 in Los Angeles. The following comparison uses that on-chain event time to split the preceding and following 24 hours.^[2][3]

Over the same comparison, the mean five-minute high–low range in USDT-USDC increased from about 0.54 bps to 1.14 bps. More active stablecoin trading and larger price movements coincided with the later concentration of high-bps cycles.^[1]

A plausible mechanism is that the shock changed stablecoin conversion demand while the withdrawal pause made inventory harder to move between venues. If some market makers reduced quote size or risk exposure, price gaps could have taken longer to close.

That mechanism remains a hypothesis. These data do not establish fewer market makers, a sustained fall in Coinbase depth, or actual net investor withdrawals. Higher trading volume is not evidence of net capital outflow.

There were already 310 cycles earning at least 2 bps within the four-day window before the event. The Binance incident may have amplified the later opportunity; it cannot explain all four days. Coinbase’s 30 October conversion arrangements for certain EEA stablecoin balances are another possible influence on USDT and USDC flows, whose contribution we cannot isolate.^[4]

A high win rate through volatility still includes losses

All 3,518 completed traded cycles were recalculated from fill prices, sizes and actual commissions, matching recorded strategy PnL. There were 3,508 profitable cycles and 10 losing cycles, for a 99.72% profitable-cycle win rate.

In the first six hours after the on-chain event, six of 339 completed traded cycles lost money. The win rate fell to 98.23%, while the weighted net spread remained positive at 1.46 bps. The sharper market move did reduce the win rate: opportunity and adverse execution risk increased together.

Losing cycles cost a combined 2.18 USDC, already included in the 713.88-USDC net strategy PnL. The worst cycle earned about −2.60 bps. Typical losses followed a buy fill and further declines in both markets: the intended sell could not fill, and the exit rules subsequently realized the loss.

Of the 3,519 cycles started in the window that had bought USDT by the cutoff, 3,518 had completed and one remained incomplete. That open cycle is excluded from realized PnL and the win-rate denominator. The win rate must still be read alongside loss size, capital tied up and open-position risk.

An occasional winning ticket

These four days show intermittent windows of wider profit and nomantrade’s ability to capture some of those spreads in actual trades. “Extra” means higher realized bps than the comparison window or usual entry threshold; it is not a promise of long-term returns.

Spreads can contract quickly as markets settle, competition returns and inventory is redistributed. The final 24 hours at 0.42 bps provide a direct example. Extrapolating the best window into daily or annualized returns would overstate its persistence.

Higher returns are an occasional winning ticket, not the daily norm. They depend on the opportunity appearing, the strategy being available, sufficient funds and both legs completing. Everyday expectations belong with ordinary spreads and actual costs. The extra bps count only after the trades deliver them.

Definitions and sources

The ledger combines central event records with local journals, deduplicated by account and cycle with central records taking precedence. A completed traded cycle requires completed buy and sell legs and positive buy-leg notional. A further 816 zero-fill cycles are listed separately and do not count as profitable cycles. Win rate is profitable cycles divided by completed traded cycles.

Cycle bps equal realized strategy PnL divided by buy-leg notional, multiplied by 10,000. Aggregate bps are notional-weighted; buy and sell legs are not double-counted as volume. The 2-bp threshold is a descriptive choice for this note, not a universal market standard.

Market comparisons use Coinbase’s official candles, with Kraken as a cross-check. Each four-day comparison uses 95 fully contained hourly intervals. The event comparison uses five-minute bars over each 24-hour window; USDT-USDC returned 269 and 281 intervals, respectively. Relative-price calculations use matching intervals in both markets. No-tick intervals may have no candle, so these data cannot replay a continuous order book. Accounts can share the same shock; cycle counts are not counts of independent opportunities.

1. Coinbase candle data and interval rules
2. Binance security incident update, 30 October 2026
3. TRM Labs on-chain investigation and event timeline
4. Coinbase conversion arrangements for restricted stablecoins in the EEA

Public candle responses: USDT-USDC before the event · USDT-USDC after the event.