What is the the Santa Fe Double Auction Tournament?

What is the Santa Fe Double Auction Tournament?


This was a tournament, conceived, designed, and held in 1990 at the Santa Fe Institute, that created a large scale competition between program traders playing the roles of buyers and sellers in a computerized version of the double auction market known as the Santa Fe Double Auction Market. The Santa Fe Double Auction Market is a web-based, computerized version of the double auction played in a finite number of discrete time steps to enable human traders and computerized traders to play on a more equal footing. The organizers of the Santa Fe Double auction tournament wanted to enable humans to trade against program traders prior to the tournament in order to get experience in how the market actually works, and to be able to observe in real time how their own strategies as well as strategies of other contestants perform in training runs prior to the actual tournament. It is important to have some brief delay or discretization of time steps, otherwise traders who can execute and communicate more quickly have an inherent advantage. For example, if trading were done in continuous time the computerized traders would have an advantage by being able to submit bids and asks much more quickly than any human trader, given the relatively slow response times of humans playing the game from computer terminals connected to the Santa Fe double auction market. In the Santa Fe double auction, a fictional commondity called a token is traded between buyers and sellers. Trading occurs in a finite number of alternating bid-ask and buy-sell steps. Bid-Ask Step: During a bid-ask steps all traders have a finite period of time when they can submit bids (an price offer to buy a single unit by a buyer), and asks (a price offer to sell a single unit by a seller). No trader can see the other bids and asks until the end of the bid-ask step, at which point all traders are shown all bids and asks (and the identities of the traders who submitted them). The buyer with the highest bid is designated as the holder of the current bid and the seller with the lowest ask is designated as the holder of the current ask. So for example, in a game with 2 buyers and 2 sellers, if B1 submitted a bid of 30 and B2 submitted a bid of 20, B1 would have the higher bid and would thus be designated as the holder of the current bid price, which would be 30. If S1 submitted an ask of 100 and S2 an ask of 40, S2 would have the lower ask and would thus be designated at the holder of the current ask, which would be 40. Buy-Sell Step: Only the buyer holding the current bid and the seller holding the current ask can trade with each other during a buy-sell step. This is also a simultaneous move game, in that the buyer and seller have a short period of time (e.g. 1 second) to decide whether either wants to accept the others bid or ask. If both decide to accept each other's bid and ask, one unit is exchanged at a price equal to the average of the bid and ask prices they submitted. Otherwise if only one party accepts the other's bid or ask, they get to buy or sell at the price named by the other party. Thus, continuing the example above, if S2 and B1 decided to accept each others' bids and asks, then they would exchange one token at the average price of 35=(40+30)/2. If B1 accepts S2's ask, but not vice versa, then B1 would pay S2's full ask price of 40. Similarly if S2 accepted B1's bid, but not vice versa, then S2 would pay B1's bid price fo 30. Once a trader holds a current bid or ask, they are compelled to trade it if the other side accepts. The only way no trade will happen during a buy-sell step is if neither party accepts the other's bid or ask. The Santa Fe double auction continues on this way in a finite number of these alternating bid-ask and buy-sell steps. The fact that each bid-ask or buy-sell step lasts a short duration of time, enables humans with their slower response times (but perhaps greater intelligence) to compete on a more equal playing field. Otherwise, if the game were done in continuous time, the computerized traders could submit hundreds of bids and asks in a period of milliseconds, completing trading with each other before the human traders even had a chance to see any bids or asks and hit the submit button to place their own first bid or ask. Further details on the the operation of the Santa Fe Double auction market and rules for creating computer programs to trade in it are in the Participant's Manual for the 1990 Santa Fe Double Auction Tournament. The initial version of the Santa Fe Double auction was programmed in C/C++ and human trading was done via a simpler character-based terminal interface, at the early stages of the development of the Internet. Since then, the code has been rewritten in Java which has enabled us to develop a nicer graphical based trading interface. This makes it much easier and more fun to play as a human trader in the Santa Fe Double auction.

What is the Santa Fe Double Auction Tournament?

This was a tournament, conceived, designed, and held in 1990 at the Santa Fe Institute, that created a large scale competition between program traders playing the roles of buyers and sellers in a computerized version of the double auction market known as the Santa Fe Double Auction Market.

The Santa Fe Double Auction Market is a web-based, computerized version of the double auction played in a finite number of discrete time steps to enable human traders and computerized traders to play on a more equal footing. The organizers of the Santa Fe Double auction tournament wanted to enable humans to trade against program traders prior to the tournament in order to get experience in how the market actually works, and to be able to observe in real time how their own strategies as well as strategies of other contestants perform in training runs prior to the actual tournament.

It is important to have some brief delay or discretization of time steps, otherwise traders who can execute and communicate more quickly have an inherent advantage. For example, if trading were done in continuous time the computerized traders would have an advantage by being able to submit bids and asks much more quickly than any human trader, given the relatively slow response times of humans playing the game from computer terminals connected to the Santa Fe double auction market.

In the Santa Fe double auction, a fictional commondity called a token is traded between buyers and sellers. Trading occurs in a finite number of alternating bid-ask and buy-sell steps.

Bid-Ask Step: During a bid-ask steps all traders have a finite period of time when they can submit bids (an price offer to buy a single unit by a buyer), and asks (a price offer to sell a single unit by a seller). No trader can see the other bids and asks until the end of the bid-ask step, at which point all traders are shown all bids and asks (and the identities of the traders who submitted them). The buyer with the highest bid is designated as the holder of the current bid and the seller with the lowest ask is designated as the holder of the current ask. So for example, in a game with 2 buyers and 2 sellers, if B1 submitted a bid of 30 and B2 submitted a bid of 20, B1 would have the higher bid and would thus be designated as the holder of the current bid price, which would be 30. If S1 submitted an ask of 100 and S2 an ask of 40, S2 would have the lower ask and would thus be designated at the holder of the current ask, which would be 40.

Buy-Sell Step: Only the buyer holding the current bid and the seller holding the current ask can trade with each other during a buy-sell step. This is also a simultaneous move game, in that the buyer and seller have a short period of time (e.g. 1 second) to decide whether either wants to accept the others bid or ask. If both decide to accept each other's bid and ask, one unit is exchanged at a price equal to the average of the bid and ask prices they submitted. Otherwise if only one party accepts the other's bid or ask, they get to buy or sell at the price named by the other party. Thus, continuing the example above, if S2 and B1 decided to accept each others' bids and asks, then they would exchange one token at the average price of 35=(40+30)/2. If B1 accepts S2's ask, but not vice versa, then B1 would pay S2's full ask price of 40. Similarly if S2 accepted B1's bid, but not vice versa, then S2 would pay B1's bid price fo 30. Once a trader holds a current bid or ask, they are compelled to trade it if the other side accepts. The only way no trade will happen during a buy-sell step is if neither party accepts the other's bid or ask.

The Santa Fe double auction continues on this way in a finite number of these alternating bid-ask and buy-sell steps. The fact that each bid-ask or buy-sell step lasts a short duration of time, enables humans with their slower response times (but perhaps greater intelligence) to compete on a more equal playing field. Otherwise, if the game were done in continuous time, the computerized traders could submit hundreds of bids and asks in a period of milliseconds, completing trading with each other before the human traders even had a chance to see any bids or asks and hit the submit button to place their own first bid or ask.

Further details on the the operation of the Santa Fe Double auction market and rules for creating computer programs to trade in it are in the Participant's Manual for the 1990 Santa Fe Double Auction Tournament.

The initial version of the Santa Fe Double auction was programmed in C/C++ and human trading was done via a simpler character-based terminal interface, at the early stages of the development of the Internet. Since then, the code has been rewritten in Java which has enabled us to develop a nicer graphical based trading interface. This makes it much easier and more fun to play as a human trader in the Santa Fe Double auction.

Back to Main Page	Why a Tournament?