In Part II, I discussed how the initial CASTrader dark market will be designed as a Continuous Double Auction (CDA) with bids and asks like virtually any stock exchange, yet unlike some stock exchanges, it would feature a centralized, open, electronic limit order book. European markets have gone this route and their microstructure has been studied. A survey of the literature indicates that such market designs don't need to have market makers (like NASDAQ) or specialists (like NYSE). Specialists have enjoyed a privileged position on the NYSE (but seem to be rapidly fading away) and have sometimes abused that position, although they apparently saved the day in the recent downturn. (Update: Mahalanobis has more). In CASTrader, software agents will be the market makers (MMs), and rather than enjoying a monopolistic position, they will have to fight to survive, like every other agent. For CASTrader, the distinction of a market maker will be a matter of semantics and design philosophy more than anything else, and a market maker (MM) is defined as follows:
A CASTrader MM is a special type of trader designed to provide a continuous bid-ask spread on a security or basket of securities. It has no obligation to provide a given level of liquidity, and as such, it enjoys little or no special privileges nor does it suffer any restrictions.
Design Philosophy. Thus, MMs in CASTrader will be designed to provide liquidity indefinately, without going broke doing so. Like real MMs, they will trade out of an inventory of stock. However, since they have no special privileges, they won't be restricted on when they can trade like specialists are. The only advantage I'm contemplating giving them now is free processor computation time. Whereas other traders must make enough profit to cover their computation costs to "live," the market makers may get a free ride in return for the service they provide.
Shannon Trader. The Shannon Trader is a natural and simple way to implement the above-described market-making capability. The Shannon Trader automatically manages an inventory of stock and can be programmed to always have a bid and ask price for trading any security, down to the smallest desired bid-ask spread, under any market condition. The Shannon Trader will survive all but the harshest markets and thrives on volatile and frictionless markets, providing more liquidity the more volatile the market is. A major drawback, however, is that liquidity is minimal relative to the inventory it carries. To the degree the traders become wealthier than Shannon MMs (and I am certainly hoping they do), they will eventually overwhelm the ability to provide liquidity. At that point, I would expect a symbiotic relationship to develop whereby the Shannon MMs extract added wealth by forcing the dark market to be more volatile than otherwise necessary in return for more liquidity. Does friction impede motion in the market like it does physical objects? It will be interesting to see how the dark market behaves. I also plan to give the Shannon MM the ability to control it's bid-ask spread and evolve via genes and the ability to reproduce, and it can potentially adapt optimally to each security and market condition.
Arbitrageurs. Some of the arbitrageurs discussed in Part I will be recruited to double as market makers in cases where it easy to program them to have a continuous bid-ask spread. These MMs will subsist on arbitrage between the real market and the CASTrader dark market. To the extent CASTrader evolves to have internal markets that do not mirror real markets, these market makers cannot be used of course. In contrast to the Shannon Trader, they will serve as high liquidity market makers with low capital requirements, since they will be able to draw virtually unlimited liquidity from the real markets - at a price. The drawback of this style of market maker is that it will act as the conduit through which CASTrader can potentially leak capital needlessly to the outside world. To the degree the dark market significantly whipsaws above and below the real market, these MMs could leak significant friction to the real market. The challenge will be to find good inventory management schemes that minimize leakage. I suspect such schemes can again be optimized by evolution. Ideally, these MMs would only trade with the real market when CASTrader anticipates a significant market move and trade out of inventory at other times.
Other Algorithms. The above may be all that is necessary for CASTrader II, but only experimentation will determine for sure. There are other market maker algorithms out there, and I have previously experimented with variations of the Shannon Trader that would mitigate the low liquidity issue in return for increased risk during a strong market move. In CASTrader, diversity is always better, and all schemes have the potential to add an edge. The challenge is finding the most bang for the buck.
Comments