John Rust, John Miller, and Richard Palmer. Behavior of trading automata in a computerized double auction market. In Friedman and Rust [1993] , pages 155--198.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....principles by which robust bidding strategies can be designed that perform well against both human and computerized opposition. Our focus on competition in heterogeneous bidder populations is similar to that of the agent vs. agent competition held at the Santa Fe Double Auction Tournament (SFDAT) [Rust et al. 1992] . The SFDAT was an intrinsically discrete time auction, with nonpersistent orders, synchronized bidding of all agents at every time step, and a coarse time step size deliberately chosen to allow all agents enough time to calculate and place their bids. Thus the conclusions may not apply to ....

J. Rust, J. Miller, and R. Palmer. Behavior of trading automata in a computerized double auction market. In D. Friedman and J. Rust, editors, The Double Auction Market: Institutions, Theories, and Evidence, Redwood City, CA, 1992. Addison-Wesley.

....of auctions, including the Continuous Double Auction institution (CDA) which is pervasive in real world markets and which is known for its high market eciency. A signi cant competitive analysis of agent trading strategies was performed in the Santa Fe Double Auction Tournament (Rust et al. [8]) The primary conclusion was that an extremely simple sniping strategy outperformed all other submitted strategies. Subsequently, other algorithms have been proposed (Cli 1997; Gjerstad and Dickhaut, 1998) but have only been studied in homogeneous populations where all agents use the same ....

.... with human subjects in laboratory simulations of CDAs have found reliable and rapid convergence of prices to values extremely close to the theoretical competitive equilibrium values [10, 11] There have also been several studies of CDAs with various types of computerized bidding agents [1, 2, 3, 5, 6, 8, 9]. A major comparative study of several di erent agent trading strategies was performed in the Santa Fe Double Auction Tournament (SFDAT) Rust et al. 8] Over 30 di erent programs were submitted, encompassing a wide range of approaches. The surprising outcome of the tournament was that an ....

[Article contains additional citation context not shown here]

J. Rust, J. Miller, and R. Palmer. Behavior of trading automata in a computerized double auction market. In D. Friedman and J. Rust, editors, The Double Auction Market: Institutions, Theories, and Evidence, Redwood City, CA, 1992. Addison-Wesley.

....sellers submit bids, and it is continuous in the sense that the market clears instantaneously on receipt of compatible bids. The CDA has also been widely employed in experimental economic studies, and even in a prior trading agent competition conducted at a Santa Fe Institute workshop in 1990 [15]. Over the last ten years, CDA markets have served as a basis for many studies of artificial trading agents (Cliff s recent example [4] provides an extensive bibliography) A CDA exercise typically commences with a random assignment of good valuations and endowments to agents, followed by a ....

....potential to incorporate real world market data of some sort. On the other hand, the basic scenario may be too simple and familiar to induce novel and interesting agent designs. Indeed, a lesson from prior experience is that it is difficult to make sophisticated reasoning pay off in this context [8, 15]. For the same reason, the scope of new insights that might be expected from the exercise could be limited. Some incremental extensions and variants to the basic CDA may overcome some of these disadvantages in part. The possibilities include: Complex activity rules. The winning trader in the ....

[Article contains additional citation context not shown here]

John Rust, John Miller, and Richard Palmer. Behavior of trading automata in a computerized double auction market. In Friedman and Rust [7], pages 155-198.

.... agent bidding strategy is based on stochastic modelling of the auction process using a Markov chain [106] 107] A sequential bidding method that uses dynamic programming is proposed in [146] The steal strategy involves waiting in the background and bidding when the bid and ask get very close [117]. In [58] heuristic fuzzy rules and fuzzy reasoning mechanisms are used to determine the best bid to make given the state of the marketplace. Multiple auctions. In this type of auction, the agent needs to monitor all the relevant auctions, decide which one to bid in and determine what to bid in ....

J. Rust, J.H. Miller, and R. Palmer. Behavior of trading automata in a computerized double auction market. In D. Friedman and J. Rust, editors, The Double Auction Market: Institutions, Theories, and Evidence, pages 155 198. Addison-Wesley, 1991.

....not depend on the behavior of the other entrants. However, there have also been several competitions in which the agents do interact directly. Examples include Axelrod s iterated prisoner s dilemma (IPD) tournament from the late 1970 s [3] the Santa Fe double auction tournament in the late 1980 s [16]; and the recent RoShamBo (rock paper scissors) programming competition [4] All three of these competitions led to interesting results despite the fact that entered programs faced very limited sensation and action spaces in domains that have been well studied and understood in isolation ....

J. Rust, J. Miller, and R. Palmer. Behavior of trading automata in a computerized double auction market. Addison-Wesley, Redwood City, CA, 1992.

....for some kinds of mechanisms, including continuous double auctions [10] the mechanisms are too complex to admit analytical solutions. Because of these complexities, economists are increasingly turning to computational methods in an attempt to take an engineering approach to microeconomic design [23, 25]. We follow such an approach in this paper. 2 Co evolution One approach to computational microeconomic design is to use techniques from machine learning to explore the space of possible ways in which agents might act in particular markets. For example, reinforcement learning has been used to ....

J. Rust, J. Miller, and R. Palmer. Behavior of trading automata in a computerized double auction market. In D. Friedman and J. Rust, editors, The Double Auction Market: Institutions, Theories and Evidence, Santa Fe Institute Studies in the Sciences of Complexity, pages 155-198. Perseus Publishing, Cambridge, MA, USA, 1993.

....for some kinds of mechanisms, including continuous double auctions [t0] the mechanisms are too complex to admit analytical solutions. Because of these complexities, economists are increasingly turning to computational methods in an attempt to take an engineering approach to microeconomic design [23, 25]. We follow such an approach in this paper. 2 Co evolution One approach to computational microeconomic design is to use techniques from machine learning to explore the space of possible ways in which agents might act in particular markets. For example, reinforcement learning has been used to ....

J. Rust, J. Miller, and R. Palmer. Behavior of trading automata in a computerized double auction maxket. In D. Friedman and J. Rust, editors, The Double Auction Market: Institutions, Theories and Evidence, Santa Fe Institute Studies in the Sciences of Complexity, pages 155-198. Perseus Publishing, Cambridge, MA, USA, 1993.

.... Experiments with human subjects in simulated CDAs find reliable price convergence close to theoretical equilibria [9, 10] Several studies examined CDAs with various computerized bidding agents [1, 2, 4, 5, 8] including a major comparative study in the Santa Fe Double Auction Tournament (SFDAT) [7]. The principal conclusion was that Todd Kaplan s simple sniping strategy, which waits for the bid ask Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for prot or ....

....distribution between the unit s limit price and either a maximum allowable price for sellers, or a minimum allowable price for buyers. One wouldn t expect this algorithm to perform well, but it does provide a useful baseline level of performance. 2. 2 Kaplan s Sniping Traders Kaplan agents [7] tend to let other agents do all the negotiating. Kaplan buyers will bid at the best ask only when one of the following three conditions are met: 1) Juicy offer: the best ask is less than the minimum trade price in the previous period. 2) Small spread: the best ask is less than the maximum trade ....

[Article contains additional citation context not shown here]

J. Rust, J. Miller, and R. Palmer. Behavior of trading automata in a computerized double auction market. In D. Friedman and J. Rust, editors, The Double Auction Market: Institutions, Theories, and Evidence, Redwood City, CA, 1992. Addison-Wesley.

....these parameters. The training with the GA requires the agent to know the final convergence price of the whole auction. It is not clear how such GA training can be applied to online settings. Other types of intelligent agents have also been designed for double auctions. In the Santa Fe Tournament [13], 30 different intelligent programs competed in a synchronized double auction, and a simple non adaptive agent won. That simple strategy is as follows: wait in the background and let the others do the negotiation, but when bid and ask get sufficiently close, jump in and steal the deal. This ....

John Rust, John Miller, and Richard Palmer. Behavior of trading automata in a computerized double auction market. In Friedman and Rust

....Rather than rejecting market efficiency altogether, some ideas for a less static definition of efficiency are proposed in section 8. The interaction of market actors following different strategies can also be simulated in the sense of experimental economics. A step in this direction is done in (Rust et al. 1993), where the interaction of different computerized traders in a simulated double auction market is studied. For future research, we suggest modeling the different traders of a market simulation study according to the heterogeneous market hypothesis: each market component with its own time horizon, ....

Rust J., Miller J. H., and Palmer R., 1993, Behavior of trading automata in a computerized double auction market, in The Double Auction Market -- Institutions, Theories, and Evidence; Santa Fe Institute Proceedings, 14, 155--198.

....demand and supply, largely ignored earlier, are shown to play a critical role in market dynamics and convergence to equilibrium. 1 Testing the effect of conjectured trading strategies by coding these conjectures into computer programs designed to mimic human behavior can be very complex [Rust, Palmer, and Miller (1993)] Since these conjectures specify only a few aspects of behavior and trader behavior varies widely, the possible set of programs is bewilderingly large. Even if market outcomes with these traders are similar to those with human traders, it is often difficult to understand 2 Price controls such ....

Rust J., Miller, J.H., and Palmer, R (1993). Behavior of Trading Automata in a Computerized Double Auction Market. In The Double Auction Market: Institutions, Theories, and Evidence, edited by D. Friedman and J. Rust, Santa Fe Institute Series in the Sciences of the Complexity, Proceedings Volume XV. New York, NY, Addison-Wesley.

....a decision theoretic approach to game theory, which is advocated by some as a practical solution concept for developing an agent strategy. Our p strategy can be viewed as 1 level RMM, as it models the behavior of other agents (but not the other agents interior reasoning processes) Rust et al. [18] have carried out a double auction tournament to investigate appropriate agent strategies. Their results show that a very simple waiting in the background trading strategy emerges as the winner of the tournament. We have not used this winning strategy in our experiments, as the auctions are ....

Rust, J., J. Miller, and R. Palmer, Behavior of Trading Automata in a Computerized Double Auction Market, in The Double Auction Market, D. Friedman and J. Rust, Editors. 1993, 155-198.

....rates of exploration, the market settled to a predictable equilibirum, but higher rates led to complex dynamics, including temporary bubbles and crashes. We know of only two other papers comparable to the trading agent work we advocate here: there are by Easley and Ledyard [25] and Rust et al. [49]. We discuss these papers in detail in [15, 20] and also explain why other work, such as Epstein and Axtell s recent book [26] is not relevant. The significant issue here is that, to the best of our knowledge, there are very few papers in the economics literature that bear any worthwhile ....

J. Rust, J. Miller, & R. Palmer. Behavior of trading automata in a computerized double auction market. In [29, pp.155-198], 1992.

....features have been said to be very important for analysis of emergent phenomena in markets. Therefore, alternative approaches apart from REH which describe agents adaptive behavior, are said to be necessary. Several alternative approaches are proposed. Among them, there is a multiagent approach [4, 36, 64, 86, 98, 104]. Previous studies based on this approach, make market models with artificial adaptive agents and conduct computer simulations. Then they analyze the evolution of models and use the results of the analysis to understand the actual markets. There are, however, two problems in the previous ....

....prediction process. So, their results can t explain rate dynamics in the real markets well. 2.4 Multiagent models: Integration of Micro and Macro In order to establish linkage between micro and macro, several alternative approaches are proposed. Among these, there is an multiagent models approach [4, 36,64, 86, 98, 104]. Previous studies in this approach make market models with artificial adaptive agents and conduct computer simulations (fig. 2.5) Then the studies analyze the dynamics of the market model and use the results of the analysis to understand the actual markets. This approach is inspired the ....

John Rust, John H. Miller, and Richard G. Palmer. Behavior of trading automata in a computerized double auction market. In Daniel Friedman and John Rust, editors, The Double Auction market: Institutions, Theories, and Evidence, pages 155--198. Addison-Wesley, 1993.

....learn to play Nash Markov equilibria. There are many other cases of evolutionary algorithms (EAs) being applied to decision situations, such as the prisoners dilemma by Ho [16] Miller [17] and Axelrod [18] sequential decisions by Oliver [19] and double auctions by Rust, Miller, and Palmer [20]. This broad stream of literature, including much else not mentioned, suggests that the goal of practical, automated negotiating agents is ambitious but attainable. 4 Implementation The success of evolutionary algorithms in diverse domains, but especially in decision and search problems [21] ....

Rust, J., J.H. Miller, and R. Palmer, Behavior of Trading Automata in a Computerized Double Auction Market, in The Double Auction Market, D. Friedman and J. Rust, Editor^Editors. 1992, Addison-Wesley: Reading, MA. p. 155-198.

.... [17] Epstein and Axtell [18] Gode and Sunder [22] Guriev and Shakhova [24] Holland [26] Holland and Miller [27] Kirman [29, 30, 31] Kollman et al. 32] Lane [33] Mailath et al. 37] Marimon et al. 38] Marks [39] McFadzean and Tesfatsion [41] Miller [42] Routledge [49] Rust et al. [50]; Sargent [52] Shubik [54] Tesfatsion [57, 59, 60, 61] Vriend [62] and Young [64, 65] How might ACE be taught to graduate students in a typical department of economics today In keeping with the subject matter of ACE, as well as the newness of the methodological approach 1 As explained more ....

Rust, J., Miller, J., and Palmer, R., "Behavior of Trading Automata in a Computerized Double Auction Market," pp. 155--198 in D. Friedman and J. Rust (eds.), The Double Auction Market: Institutions, Theories, and Evidence, Addison-Wesley, New York, 1992.

....the DA does not square with considerable experimental evidence that demonstrates that even untrained undergraduates are able to quickly learn to trade in DA markets and observed outcomes are very close to the predicted competitive equilibrium outcomes. Research by Gode and Sunder (GS) 1993) and Rust, Miller and Palmer (RMP) 1993) suggests a potential resolution to this paradox: outcomes in the DA market are typically collectively rational even if some traders are not individually rational . 20 This robustness property of the DA institution vastly simplifies the computational problems of traders. Instead of solving ....

....strategies employed by individual opponents, so that it is unlikely that rational traders could identify and exploit irrational traders unless they were repeatedly trading with these agents in the same market. Furthermore the evolutionary computer tournaments conducted in Miller, Palmer and Rust (1993) show that the irrational traders are rapidly driven into extinction anyway, so in the absence of a steady stream of noise traders the long run outcomes of DA markets are dominated by the more intelligent or rational traders. Unfortunately, given the apparently intractable computational problem ....

[Article contains additional citation context not shown here]

Rust, J. Miller, J. and R. Palmer (1993) "Behavior of Trading Automata in a Computerized Double Auction Market" in D. Friedman and J. Rust (eds.) The Double Auction Market: Institutions, Theories, and Evidence Volume XIV in the Santa Fe Institute Studies in the Sciences of Complexity, Addison-Wesley, Reading, Massachusetts.

No context found.

John Rust, John Miller, and Richard Palmer. Behavior of trading automata in a computerized double auction market. In Friedman and Rust [1993] , pages 155--198.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC