The part where market makers really make money often isn’t posting orders to capture the spread, but using low fees to take stale quotes before others have time to cancel them. This is what I observed in Arcus’s public trade-by-trade data for the largest market maker address, 0x00cb, based on PANews’ analysis of the address.
First, the facts. Over 6.3 days, the address made about $1,643 in gross profit on NVDA-USD. Taker trades accounted for 88% of that, while maker orders contributed just $268; directional positions also lost $71. Its taker trades received an average price about 0.99 bp better than fair value, and in the 300 milliseconds before each trade, Binance’s mid-price had already moved 0.81 bp in its direction. In other words, Binance moved first and Arcus’s quotes followed, allowing it to trade before the stale quotes disappeared. Fees are make-or-break: at the base fee tier, the same set of trades would have resulted in a net loss of $1,655; it only turned profitable after reaching VIP status.
Inventory management is even more noteworthy. The account had about $45,774 in equity while holding positions across 60 markets, with roughly $678,000 in notional exposure—about 14.8x leverage—and only 7% of its margin free. Its NVDA inventory changed direction 392 times in a week, and changes in its NVDA positions showed no clear correlation with its SPY or QQQ positions. This suggests it may be managing inventory through high turnover and diversification across markets rather than hedging, but the source material explicitly says it is impossible to determine whether it hedges, so this remains uncertain.
My observation: this strategy’s moat isn’t in the formula, but in “providing liquidity without becoming someone else’s profit.” When other market makers’ orders were filled, Binance’s price had already moved an average of 0.80 bp against them—effectively leaving them to take the other side at stale prices. 0x00cb faced just 0.34 bp of adverse movement, thanks to faster cancellations: in 20-minute snapshots, it had an average of 111 orders posted, but only 18% were still there 30 seconds later.
When would this conclusion no longer hold? If its actual fee rate rises, its taker-trading edge narrows, or slower inventory turnover leads to greater directional losses, the strategy stops adding up. Things to monitor include changes in fee tier and whether inventory skew and cancellation speed move in tandem. Note that the profit-and-loss attribution is an estimate from the original article, not an exchange settlement record.
#加密市场 #CryptoWatch
Sources:
PANews (originally published 10-26 at 10:00 Beijing time)
https://www.panewslab.com/zh/articles/01a1725d-7b5b-7039-b8b4-4c4e6afb3550
First, the facts. Over 6.3 days, the address made about $1,643 in gross profit on NVDA-USD. Taker trades accounted for 88% of that, while maker orders contributed just $268; directional positions also lost $71. Its taker trades received an average price about 0.99 bp better than fair value, and in the 300 milliseconds before each trade, Binance’s mid-price had already moved 0.81 bp in its direction. In other words, Binance moved first and Arcus’s quotes followed, allowing it to trade before the stale quotes disappeared. Fees are make-or-break: at the base fee tier, the same set of trades would have resulted in a net loss of $1,655; it only turned profitable after reaching VIP status.
Inventory management is even more noteworthy. The account had about $45,774 in equity while holding positions across 60 markets, with roughly $678,000 in notional exposure—about 14.8x leverage—and only 7% of its margin free. Its NVDA inventory changed direction 392 times in a week, and changes in its NVDA positions showed no clear correlation with its SPY or QQQ positions. This suggests it may be managing inventory through high turnover and diversification across markets rather than hedging, but the source material explicitly says it is impossible to determine whether it hedges, so this remains uncertain.
My observation: this strategy’s moat isn’t in the formula, but in “providing liquidity without becoming someone else’s profit.” When other market makers’ orders were filled, Binance’s price had already moved an average of 0.80 bp against them—effectively leaving them to take the other side at stale prices. 0x00cb faced just 0.34 bp of adverse movement, thanks to faster cancellations: in 20-minute snapshots, it had an average of 111 orders posted, but only 18% were still there 30 seconds later.
When would this conclusion no longer hold? If its actual fee rate rises, its taker-trading edge narrows, or slower inventory turnover leads to greater directional losses, the strategy stops adding up. Things to monitor include changes in fee tier and whether inventory skew and cancellation speed move in tandem. Note that the profit-and-loss attribution is an estimate from the original article, not an exchange settlement record.
#加密市场 #CryptoWatch
Sources:
PANews (originally published 10-26 at 10:00 Beijing time)
https://www.panewslab.com/zh/articles/01a1725d-7b5b-7039-b8b4-4c4e6afb3550
