Roughly speaking, there are three types of market structures, and all can coexist: Order Books, Dealer (quote-driven), and auction models. Only auctions require a single physical matching algorithm, but those only are central at discrete moments in time. That said, some of the “decentralized exchange” proposals I have seen (many of which have attracted very large investments in their ICOs) are nonsensical, as they severely misunderstand the nature of price discovery and do a lot of “arm waving” about price formation.

There are two reasons why this is true: First, price formation requires the controlled interaction of orders and trading interests. The phrase used by market structure experts is “order competition,” which is only possible when displayed orders can interact on one or more order books. In modern markets, this is facilitated by Smart Order Routers (SORs), which scan all displayed markets before deciding where to trade. As a result, investors should be very wary of “solutions” that do not have an SOR or routing capability, as that lack indicates that pricing on that venue can become well out of line with other venues.

Second, price formation requires that participants trust the process. As a result, the bane of all pricing mechanisms is manipulation, where “bad actors” distort the market and profit from the distortion. Manipulation, therefore, is a major issue for many market models, and has been largely ignored by many of the decentralized exchange whitepapers I have read so far.

Collectively, the CFTC and the SEC have assessed large fines for activities such as “spoofing,” which they often define as posting orders for the purpose of influencing price movement, without a true intent to trade. These fines are NOT a government overreach, as the knee-jerk conclusion might be in the libertarian crypto community. In fact, such regulation is necessary to protect market integrity. (Even if the crypto community DOES believe in a winner take all, caveat emptor approach, it would never be allowed by regulators, so isn’t worth discussing.) In addition to spoofing, there are other manipulative activities that need to be addressed, including misrepresenting prices, front running, and providing unequal access to different client types. All of these should be addressed by each exchange or market model, whETHer centralized or decentralized, yet many have not.

Let’s consider one decentralized exchange proposal that was cited in both Bloomberg articles noted above: Airswap. The Swap protocol facilitates a double blind, multi/step interaction between buyers and sellers along with anonymous price providers. This sets up a game, like poker, where achieving best price could entail a great deal of “bluffing.” (While poker is a favorite pastime of mine, I acknowledge its shortcomings as a model for an investment ecosystem.) AirSwap also incorporates “last look” features, and its mETHod of displaying initial trading interests is reminiscent of “flash orders,” which attracted significant regulatory ire in the past. The company’s whitepaper described no incentives either to police gaming or safeguard pricing for honest investors, nor do they say how their market will interact with other markets participating in the price discovery process. (There is no mention of manipulation in the whitepaper at all.) It may well be, with the $36 million they raised from investors in their token offering, that they will develop such features, but those issues should have been addressed in the initial design.

There are other markets which have also attracted significant investment, such as the Republic Protocol. On the one hand, we at CoinRoutes are happy about their success, as it indicates the importance of aggregating dark liquidity, as our own proposed network using the RouteCoin protocol will do. It also will provide a dark pool for our SOR to utilize, which will be beneficial to our clients. That said, without an SOR to link Republic to displayed markets, it will lack the diversity and transparency to be a dominant trading venue. This is because price discovery requires the interaction of trading interests from as many participants simultaNEOusly as possible.

In more mature markets, such as Equities, there is still consternation over fragmentation, despite the ubiquitous availability of sophisticated routing strategies which tie the market togETHer. In the U.S., for example, smart order routers (SORs) simultaNEOusly scan the 13 exchange order books and interact with large numbers of alternative trading systems that have “dark” orders as well. There is, however, a constant fear that, if too much liquidity migrated off the displayed order books at the exchanges, price discovery would fail and trading costs would rise. Thus, it is simple to understand why the notion of crypto moving to mostly non-displayed, decentralized exchanges would be concerning, unless those decentralized exchanges supported displayed order books and facilitated SOR interaction.

Interestingly, one excellent idea to solve this conundrum is the use of decentralized protocols, such as 0x, which enables users to safeguard their own keys while trading with decentralized exchanges. This would only be a complete solution, however, if all exchanges with meaningful displayed liquidity operated in a distributed manner. Essentially, 0x allows for limit order books, most of which are currently operated by “centralized exchanges” such as GDAX, Gemini, Bittrex & Kraken, to operate without also being custodians of the crypto-assets they trade.

Ultimately, I believe that this will be the way that the market evolves, since, trading on multiple displayed “central” order books with distributed settlement makes the most economic sense overall. Separating the order books from settlement could be done in a number of ways (including a full distributed model such as 0X), but one consideration should be cost. In equities, it is not uncommon for participants to execute thousands of trades daily across exchanges in a specific symbol, but only settle one trade at the end of the trading day. The trade they settle, of course, is the net amount they bought or sold from all the trades done during that day. Equity trades, all in, cost between $10 to $20 per net trade, so the savings from such netting is substantial.

While it is true that the cost of a crypto transfer is up to 1/100th of that amount, it means that participants that trade more than 100 times in a given asset during the day would be worse off if they were forced to settle every trade. As a result, my expectation is that decentralized models will develop, which incorporate some form of session netting. Such platforms will win in the end, as that will be more efficient for the market makers and other participants that do most of the trading. This, of course, requires that the mechanisms for price discovery and for settlement need to become untangled, which will require significant evolution of the markets.