Computer bridge

src: i.ytimg.com

The computer bridge is a game contract bridge game using computer software. After years of limited advances, since about the end of the 20th century the field of computer bridges has made great progress. In 1996, the American Contract Bridge League (ACBL) established the World Computer Bridge Championships, which will be held annually along with major bridge events. The first championship took place in 1997 at the North American Bridge Championships in Albuquerque. Since 1999 the event has been conducted as a joint activity of the American Contract Bridge League and World Bridge Federation. Alvin Levy, a member of the ACBL Board, initiated this championship and has coordinated events each year from the start. Event history, articles and publications, analysis, and play notes can be found on the official website.

Video Computer bridge

World Computer Bridge Championship

Kejuaraan Bridge biasanya dimainkan Computer Dunia dengan diikuti sebagai round robin knockout ANTARA EMPAT kontestan teratas. Pemenang ACARA tahunan adalah:

• 1997 Jembatan Baron
• 1998 GIB
• 1999 GIB
• 2000 Jembatan meadowlark
• 2001 Jack
• 2002 Jack
• 2003 Jack
• 2004 Jack
• 2005 Wbridge5
• 2006 Jack
• 2007 Wbridge5
• 2008 Wbridge5
• 2009 Jack
• 2010 Jack
• 2011 Jembatan Hiu
• 2012 Jack
• 2013 Jack
• 2014 Jembatan Hiu
• 2015 Jack
• 2016 Wbridge5
• 2017 Wbridge5

Maps Computer bridge

Computer manusia versus

In Zia Mahmood's book, Bridge, My Way (1992), Zia offers a £ 1 million bet that no four-person team she chooses will be beaten by a computer. A few years later the GIB bridge program, the brainchild of American computer scientist Matthew Ginsberg, proved capable of playing expert declarers like winkle squeezed in a play test. In 1996, Zia withdrew the stakes. Two years later, GIB became world champion in computer bridge, and also had 12th place score (11210) in declarer game compared to 34 top human in Par Contest 1998 (including Zia Mahmood). However, such a par contest measures technical bridge analysis skills only, and in 1999 Zia defeated various computer programs, including GIB , in individual round robin matches.

Further advances in the field of computer bridges have resulted in stronger bridge playing programs, including Jack and Wbridge5 . These programs have been ranked highly in national bridge rankings. A series of articles published in 2005 and 2006 in the Dutch bridge magazine IMP explained the match between the five-time computer bridge world champion Jack and seven top Dutch pairs including the Bermuda Bowl winner and two champions Europe in power. A total of 196 plays are played. Jack defeated three of the seven pairs (including the European champions). Overall, the program lost out to a small margin (359 versus 385 IMP).

src: cdn.vox-cdn.com

Cardplay algorithm

The bridge poses a challenge for the players who are different from board games like chess and leave. In particular, the bridge is an incomplete game of stochastic information. At the beginning of the transaction, the information available to each player is limited only to his card. During the next bidding and playback, more information is available through the bidding of the other three players on the table, declarer partner cards (put) laid open on the table, and cards played in every trick. However, only at the end of the game, complete information is obtained.

The top-level bridge program currently deals with this probabilistic nature by producing many samples representing unknown hands. Each sample is generated randomly, but is limited to compatible with all information available so far from offers and games. Furthermore, results from various game lines were tested against the optimal defense for each sample. This test is performed using so-called "double dummy solvers" that use an extensive search algorithm to determine the optimal game line for both parties. The game line that produces the best score averaged over all samples is chosen as the optimal game.

An efficient double-dummy splitter is the key to a successful bridge-playing program. Also, as the number of computations increases with the sample size, techniques such as sampling are important to produce a set of samples that have a minimum size but are still representative.

src: ardupilot.org

Comparison with other strategy games

While the bridge is an incomplete information game, a double dummy solver analyzes a simplified version of the game where there is perfect information; the offer is ignored, the contract (trump suit and declarer) is given, and all players are assumed to know all the cards from the start. Therefore solver can use many game search techniques of the game which is usually used in completing the game win/win/sweepstick two perfect players like chess, go and reversi. However, there are some significant differences.

• Although dual dummy bridges in practice are competition between two common players, each "player" controls two hands and the card must be played in the correct order reflecting the four players. (That makes the difference which of the four hands wins the trick and should lead the next trick.)
• Double dummy bridges are not just win/lose/series and are not exactly zero-sum, but the number remains because the two parties are playing compete for 13 tricks. It's a trivial thing to turn the game of constant numbers into a zero-sum game. In addition, the objectives (and risk management strategies) in general contract bridges depend not only on the contract but also on the form of the tournament. However, since the double dummy version is deterministic, the goal is simple: one can without losing a general purpose to maximize the number of tricks taken.
• The bridge gradually scores; each playing a trick contributing irreversibly to the last "score" in terms of winning or losing tricks. This is different from the game where the final result is more or less open until the game ends. On the bridge, a predetermined trick provides the natural lower and upper bounds for alpha-beta pruning, and the interval shrinks naturally as the search deepens. Other games usually require an artificial evaluation function to enable alpha-beta pruning at a limited depth, or having to look to leaf nodes before pruning is possible.
• Relatively inexpensive to count "sure winners" in multiple positions in double dummy solvers. This information improves pruning. This can be thought of as some sort of evaluation function, but while the latter in the other game is the approximate value of the position, the first is the definitive bottom line at the position value.
• During the search process of a dubble game tree, one can form an equivalence class consisting of cards of similarly apparent value in a particular position. Only one card of each equivalence class needs to be considered in a subtree search, and furthermore, when using transposition tables, equivalence classes can be exploited to increase hit rates. This has been described as partition search by Matthew Ginsberg.

• Many strategy games have proved difficult in complex classes, which means that any problem within the class of complexity can be reduced in polynomial time to that problem. For example, the generalizations of x ÃÆ'â € "x have been proven EXPSPACE -complete (both in EXPSPACE and EXPSPACE - hard), which effectively means that this is one of the most difficult problems in EXPSPACE . However, since there is no natural structure to be exploited in double dummy bridges toward evidence of violence or incompetence, unlike in board games, the problem of violence persists.

src: i.pinimg.com

The future

Compared to computer chess, computer bridges have not reached the world-class level, but the top robots have shown consistent levels of play. To demonstrate this, see the analysis of the last few years playing at www.computerbridge.com. However, while computer chess has taught programmers a bit about building machines that offer intelligence like humans, more intuitive and probabilistic games such as bridges can provide a better testing ground.

The question whether a bridge playing program will reach a world class level in the future is not easy to answer. Computer bridges do not interest near chess computers. On the other hand, researchers working in the field have achieved most of the current progress in the last decade.

Regardless of the level of the bridge robot, the computer bridge has changed the game analysis. The commercially available dual dummy program solves the bridge problem in which all four hands are known, usually within a fraction of a second. These days, some book and magazine editors will only rely on humans to analyze bridge problems before publication. Also, more players and bridge trainers take advantage of computer analysis in post-mortem matches.

src: i.ytimg.com

See also

• The Computer Olympics
• Monte Carlo Method
• Search the Monte Carlo tree
• Import sampling

src: img.diytrade.com

References

src: s-media-cache-ak0.pinimg.com

External links

• BridgeGuys. "Participating the Computer Bridge Software for the World Computer Bridge Championship and the History of the Software Program" (PDF) .
• Ginsberg, Matthew L. "GIB: Steps to a Bridge-Expert-Level Program". CiteSeerXÃ, 10.1.1.52.2188 .
• Bethe, Paul M (January 14, 2010). "State of Automated Bridge Play" (PDF) .
• Philippe Pionchon (1984). "Artificial Intelligence and Bridge Games". Le Bridgeur Review . Classical AI analysis applied to bridge.

Source of the article : Wikipedia