What this covers

Brad Katsuyama, the former RBC trader who went on to found IEX, tells Patrick how he came to understand what happens between pressing the buy button and receiving a fill. In 2006 and 2007, shares visible on quote screens could no longer be bought in full, and the price ticked up with each attempt. Brad tested the coincidence explanation by pressing buy after counting delays, and the stock rose every time. The cause was routing: an order from RBC reached BATS in Weehawken about two milliseconds before NYSE in Mahwah, enough for fast traders to spot it and buy ahead. Staggering the orders so they landed together brought fill rates to 100 percent.

From there the conversation turns to exchange economics. Brad says market data and technology sales now outearn trading, so exchanges sell co-location, cabling of varying quality, and microwave connections, effectively monetizing a gap they could close. He explains maker-taker pricing, which Island and the ECNs invented to break the NYSE-NASDAQ duopoly, and calls the resulting rebate a kickback that creates an agency conflict in order routing. He also argues that exchanges cannot buy or sell shares, so "liquidity" is a word used to shut down criticism.

The contested material is mostly about definitions. Brad separates classic market making and useful arbitrage from predatory strategies, and says about 95 percent of high-frequency activity at IEX is liquidity taking. He then describes IEX's answer, 38 miles of coiled fiber adding a 350-microsecond delay, and why a regulated exchange cannot be built as a minimum viable product. He also explains why RBC could not build it, since rival brokers would not use a competitor's venue.

Sharpest takeaway

Katsuyama argues that U.S. exchanges profit by selling speed, data, and tiered access that let fast traders extract value from slower investors, and that exchange design should minimize friction instead.

  • Exchanges earn more from market data and technology sales than from trading, creating incentives to fragment the market and sell latency advantages
  • Staggering orders by each exchange's latency restored fill rates to 100 percent, proving the latency arbitrage problem was real
  • Maker-taker rebates create an agency conflict in which brokers route customer orders to the longest queues with the worst execution quality

The argument · threads2 threads · 82 claims
0.71

Buy-side PMs overlook execution quality as HFT profit pools shrink for remaining firms.

2 pointscentrality 4/5
  • Patrick proposes that HFT firms' profit pools have shrunk as costs rose, and Brad partly agrees, saying the pool has shrunk for fewer players but that the remaining firms carve up a larger share.

    “I think it's true to an extent. I think the costs have gone up. So there's fewer players.”

  • Buy-side research PMs focus on company fundamentals and macro factors and typically do not understand market microstructure, even though execution quality materially affects returns on large orders.

    “a lot of buy-side PM's, especially on the research side, that are dealing mostly with companies, not so much with execution and trading, actually don't know a ton about this world”

0.29

Founders need a clear mission rather than money, because unforeseen difficulties will arise.

1 pointcentrality 2/5
  • Patrick frames Brad's advice as a classic missionary approach, meaning founders must know what they are driving toward because they will face unforeseen difficulties, in contrast to founders motivated mainly by money.

    “this is like classic missionary. You need to know what you're driving towards”