One thing that surprised me when I first joined HFT was how complicated the exchange landscape was. Did you know there are 13 equities exchanges and 16 equity options exchanges in the US (not to mention over 50 dark pools for equities)? Then there’s a bunch of futures and futures options exchanges including the CME, ICE, and CBOE. Point is, it’s a free market so there’s a lot of exchanges and a lot of competition, which is a net good thing for individuals, and also extremely lucrative for HFT firms that are able to navigate the complexities of this system. The concept I want to introduce here is called market microstructure.
The academic definition is a branch of finance concerned with the theoretical, empirical, and experimental research on how exchange occurs in markets. In practice, it’s all about gaining an advantage by getting into the weeds and understanding the different rules, fee structures, nuances, infrastructure, and even the latencies of each exchange. I’ll go over some basic insights in the US equity options space because I’m most familiar with that. I also include a question for the reader in each section to give a sense of what traders would be thinking about.
All exchanges can be bucketed into price-time or pro-rata exchanges in determining the priority that quotes get filled. Price-time means whoever is first to offer the better price gets priority. Pro-rata means everyone at the same price will get a proportional allocation to how much they quoted vs. the rest of the market. Think about how you’d change your quoting strategy depending on price-time vs. pro-rata. For example, in a pro-rata exchange, if you only want to quote 5 lots but there’s already 500 lots on the bid, would you still only quote 5 lots?
The second major classification for exchanges is the fee structure, which is typically split into whether the provider of liquidity (maker) or the taker of liquidity (taker) pays fees. Makers quote and takers trade against quotes. Historically, the way the fee structure would work is only takers pay a fee, of which most of that would be given as a rebate to the maker, while the exchange pockets a small fraction. This structure rewards those who provide liquidity, and if you recall from our previous blog, are adversely selected by quoting. More recently, inverted exchanges sprung up where the taker would get the rebate while the maker paid the fee. It sounds counterintuitive but the idea is new exchanges wanted to induce more trading volumes in their exchanges so they compensated takers, which in turn would attract market makers to quote on their exchanges because higher volumes generally meant higher PnL for market makers. Nowadays there’s a healthy mix of both types of exchanges. How would these fee structures affect the way that you quote?
Rule 611 of Regulation NMS, also called the Order Protection Rule, is an extremely important rule in US equity and equity options markets that helps to ensure investors receive the best price executions for their orders. Best price means the national best bid and offer (NBBO), so even if there’s 16 exchanges, you can be sure that when you execute on Robinhood, you’re going to get at least the best price that was currently on the NBBO. More on this later, but put another way, your order cannot be routed to an exchange with worse prices. This is one reason why most exchanges are quoting at the NBBO, and every time there’s an update, you see market makers scramble to update their quotes across the board. There’s some nuance here because exchanges have different fee structures. Can you think about how different fee structures affect quoting at the NBBO?
Exchanges are very complicated pieces of software that disseminate information and receive requests from thousands of participants. It’s a real feat of engineering. They need to receive requests, send acknowledgements, build/update an orderbook, match trades, send order confirmations, and disseminate data, all in an efficient manner while preserving the fairness and rules of the exchange. The matching engine is the component that determines when a trade occurs by matching a taker with a maker. For price-time exchanges, it’s a single threaded operation because it needs to manage state in an orderly way.
While the matching engine is deterministic, there’s a lot of network variance prior to when the exchange receives the order. This is especially true during volatile moments or after an obvious trigger where everyone is sending requests to the exchange and their server is likely overloaded. Some exchanges are more deterministic than others when it comes to network variance. How would you measure it and are there any strategies to minimize this?
This is a topic we can write an entire article about but to keep it simple, all retail orders in equity options have to go through a brokerage firm (Robinhood, Schwab, Interactive Brokers, etc.) that decides how to route the order. These brokerages outsource the routing instructions to market makers, who each have a separate entity that handles order flow directions. For example, XYZ market maker might have a separate entity called XYZ Execution Services that will instruct Robinhood where to route customer orders. Their objective is to route it to an exchange where they have priority, thereby allowing them to “internalize” the order. Recall that market makers want to trade against benign retail orders as much as possible. This isn’t free of course. Market makers pay a lot of money for the privilege to give these instructions. This industry is called Payment for Order Flow (PFOF). Note that there are laws governing that the market making arm and the order flow arm are separate entities, and it’s forbidden for information to be passed from the order routing entity to the market maker entity during trading hours. The other way around is okay which is how the order flow entity knows where to route orders.
There’s nothing nefarious going on. Recall that due to Reg NMS, the retail trader is getting the best (NBBO) price. Furthermore, in many cases, the market maker would even give price improvement beyond the best price. This is because retail brokerages compete against each other and advertise that customers are getting the best price on their platform. In turn, they require market makers to be the ones who are actually offering these price improvements, and they’re all measured against each other when retail brokerages decide who to choose as partners. This is why valuation edge is so important. If you have the most confident valuations, you can give the tightest spreads and the most price improvement, which over the long run means you can build partnerships with retail brokerages and capture market share.
If you were a market maker, how would you devise a strategy for your order routing entity and your market making entity to internalize as much retail flow as possible? There are 16 exchanges each with different rules. You can come up with some pretty complicated strategies.
That’s a quick and dirty overview of market microstructure and I hope it gives you an idea of how complex and nuanced this space is. There’s a lot of edge from understanding microstructure well and most firms have optimized for microstructure based signals and execution related improvements in this area. If you like this content and want us to go in more depth or talk about different asset class microstructures, please drop us a like and a comment on what you’d like to learn about! Some other interesting microstructure topics that I didn’t have time to cover include auctions, relaying information across exchanges (microwave towers), floor/pit trading and request-for-quotes (RFQs), tick sizes, hidden liquidity (dark pools), and much more. Thanks for reading!
