REPORT ON THE 5^TH ACBL WORLD COMPUTER-BRIDGE CHAMPIONSHIP

Toronto, Canada

July 23-28, 2001

Alvin Levy[1]

Introduction

The 2001 FunBridge.com World Computer-Bridge Championship, the 5^th annual ACBL World Computer-Bridge Championship conducted by the American Contract Bridge League, was held in Toronto, ON, July 23 to 28, 2001, as part of the ACBL’s summer North American Bridge Championships.

In 1996 the ACBL established a World Computer-Bridge Championship, to be run annually in conjunction with a major bridge activity. At its inception I was appointed by the ACBL as agent to conduct this event, and have acted in that capacity up to the present time. The 1997 inaugural Championship was held in Albuquerque, NM during the ACBL’s summer NABC. The second annual Championship was held in Chicago, IL during the 1998 summer NABC. The 3rd championship was held as part of the WBF’s 1999 Bermuda Bowl/Venice Cup, held in Bermuda, January 2000, in conjunction with the 50th anniversary of the Bermuda Bowl. The 4th Championship[2] was help in Maastricht, The Netherlands, in conjunction with the WBF’s 2000 Bridge Olympiad. In addition, various demonstrations, exhibition matches, and promotional activities have taken place at these events.

The Contestants

This year’s event, sponsored by FunBridge.com, pitted six computer-bridge programs, representing six countries, against each other. Details are shown in Table 1.

Program	Programmers	Country	Website
Blue Chip Bridge	Ian Trackman and Mike Whittaker	UK	www.bluechipbridge.com
Bridge Baron	Stephen Smith and George Yanakiev	USA	www.bridgebaron.com
Jack	Hans Kuijf	The Netherlands	www.jackware.nl
Micro Bridge	Tomio and Yumiko Uchida	Japan
Q-Plus	Hans Leber	Germany	www.q-plus.com
Wbridge5	Yves Costel	France	Ourworld.compuserve.com/homepages/yvescostel

Table 1: The Contestants.

Last year’s championship, held in Maastricht in conjunction with the WBF’s World Olympiad, had nine entries. Unfortunately, the winners of the previous three championships (Meadowlark Bridge, 2000; GIB, 1998 and 1999) were unable to participate this year. Bridge Buff, Oxford Bridge, and newcomers Sabrina and M2000 could not get the new Table Manager interface to work properly with their programs. This problem should be resolved well before next year’s competition. Eleven different computer-bridge programs have competed over the five-year history of the event and there are at least three more programs ready to compete in the near future.

The Conditions of Contest

Preparation for this year’s event started in March 2000, with the contestants, potential contestants, and some invited advisors take part in discussions that influence the Conditions of Contest. These conditions included allowable conventions, alertable bids, speed of play, number of boards played, and the use of a central server to monitor play. A description of the complete conditions of contest can be found at the official website members.aol.com/allevy/Toronto. Approximately one month before the start of the championship Convention Cards where exchanged between the participants.

This year an automated form of play was adopted. A table in play consisted of four computers (NESW) connected to a central server. This central server, or Table Manager (TM), handled all the play. At the start the opponents input each other Convention Cards. The central server was given all the hands for a match. Humans never touched or interacted with the play except when an alertable bid was made. The description of an alertable bid was manually input into the opponent’s program. After the bidding the dummy was exposed to each program, and the play began. The play was automated, and at the end of play, the next deal was started. After a round of 16 boards, the set of deals was played with the NS and EW reversed. The TM kept a record of all play and scored the matches. This requirement limited the final field to six, as some of the possible contestants had some difficulty adapting to the format.

The Competition

The competition consisted of a 32-board round robin, scored on a 20 VP scale. Jack was the winner of the round robin. Jack’s performance was impressive, winning all five round robin matches, and scoring 82 out of a possible 100 VPs (see Table 2). In second place, winning four of its five round robin matches, was Micro Bridge with 63 VPs. Jack and Micro Bridge then played a 64-board KO for the championship, with Jack starting with a 12 IMP carry-over (the lesser of the VP total of their individual match and the overall standing).

 

		1	2	3	4	5	6	Total pts
1	Blue Chip Bridge	X	2	8	7	4	0	21
2	Bridge Baron	18	X	6	2	0	2	28
3	Q-Plus Bridge	12	14	X	9	6	6	47
4	Micro Bridge	13	18	11	X	4	17	63
5	Jack	16	20	14	16	X	16	82
6	Wbridge5	20	18	14	3	4	X	59

Table 2: The standings at the end of the 32-board round robin scored on a 20 VP scale.

In the final KO match, Jack gained ground each quarter, and won convincingly, 136-84 (see Table 3), thus staking claim to the title of the 2001 World Computer-Bridge Champion.

 

	Carry-over	1-16	17-32	33-48	49-64
Jack	12	43	90	114	136
Micro Bridge		18	45	64	84

Table 3: Final 64-board KO match, scored in IMPs.

An Example of Play

A well declared hand from the final 64-board KO match between Jack and Micro Bridge follows.

  Vul: NS Dealer N

            S A K 4
            H 8
            D K 9 8 3 2
            C 10 8 7 4

            S 10 8 7 2
            H A Q 7 5
            D Q J 6
            C K J

1C = 2+ card
2C = inverted
2NT = 13-14 balanced, not forcing, but not a bad minimum
 

Lead: H10. Won by HQ. Jack played DQ and ducked the next two hearts. Since 3NT would be down if West had CA Jack later played up to CK. In the other room the bidding stopped at 1NT (Jack won 10 IMP).

The quality of play for each program was good at times and poor at other times. On the good side there were many well-played hands, which can be attributed, for the most part, to good “simulations”. On the bad side there were many badly bid hands and many obvious mistakes in play. In my estimation the overall level of play was at an Advanced/Intermediate level roughly equivalent to the average “club” player. The time horizon for the development of a program that plays at the expert level is uncertain. While programs can solve double dummy problems and sometimes perform well at single dummy problems, they have different performances when deductive logic is involved. They need simulation-based routines to make consistent expert decisions. The incorporation of concepts taken from game theory may be the direction needed to travel in reaching the expert level of play. Simple technological developments, such as bidding databases, improving speed, and single dummy solvers, should be included. Add to this the use of simulation techniques (a weak-link in human play) especially in competitive and high-level bidding situations and in many play-situations, and you may approach the artificial intelligence needed to beat the experts.

Exhibition Matches

In addition to the main competition there were exhibition matches pitting humans against computers. The two finalists, Jack and Micro Bridge, joined as team mates and played eight-board matches against two different teams of International stars. One team was Jacek Pszczola-Piotr Gawrys and Sam Lev-Pinhas Romik, who defeated the computer team by 18 IMPs. Another team was Lisa and David Berkowitz and Marty Fleisher-JoAnna Stansby, who defeated the computers by 8 IMPs. The players apparently enjoyed the experience, while the computers immediately showed a screen saver with the saying…“Wait 'til next year.” It was nice to see International star, Zia Mahmood, rooting for the computers. Zia has a great interest in computer play, having participated in many promotional activities in the past.

A hand from the Pszczola-Gawrys-Lev-Romik versus Jack - Jack - Micro Bridge - Micro Bridge match follows.

Vul: EW Dealer E

                S 10 6
                H 8 4
                D A Q J 10 4 2
                C J 7 3

S K 9 5 4                 S A 7
H A K 7 3               H Q J 10 9
D K                         D 9 7 5 3
C K Q 8 6                C 10 9 5

                S Q J 8 3 2
                H 6 5 2
                D 8 6
                C A 4 2

 

West	North	East	South
Pszczola	Jack	Gawrys	Jack
--	--	Pass	Pass
1C	3D	Pass	Pass
Dbl	Pass	4H	all Pass

At favorable vulnerability Jack showed his “new school” approach to preempting. That did not stop Pszczola-Gawrys, who were playing the Standard Yellow Card, from showing their winning style ... or maybe Gawrys thought he could take advantage of his opponents in the play. Of the four times the board was played they were the only pair to reach game. Five made easily and 10 Imps scored up.

Besides all the bridge, the contestants found time for two discussion groups. One technical, on the use of the Table Manager, and the other general, on future rules, regulations and conditions of contest. Chip Martel, world champion player and co-chair of the ACBL and WBF Law’s Commission, was gracious enough to lead the latter discussion group.

Information on the contestants, pictures, and other useful information, can be found at the official website members.aol.com/allevy/Toronto or http://www.computerbridge.com/

Editorial Note

In the refereeing procedure of this report there arose a discussion on the human reporter’s comment. Therefore the Editor consulted two prominent Bridge players, Anton Maas and Bep Vriend (a former World Champion), on the bidding performances and the comments given in the examples above. They provided the ICGA Journal with the following remarks.

On the example of play: the critique was that the hand was not well played, since North’s 2C bid is not an inverted raise; in particular opposite a hand that might contain only two clubs (as South’s does). The 2C bid should show 5 clubs. 1D seems a much more sensible response. The result is that NS get a truly awful 3 NT contract, and they were only able to make it because of a lucky lead and favourable distributions in three (!) suits.

Maas and Vriend: “It is true that 2C (inverted minor) is rather non standard. For a hand that passed in the first bidding round, five clubs is normal. So 1D is standard. The “lucky lead” is among others possible when the right hand plays, the opponents have to play towards your hearts. “Favourable distribution in three suits” is overdone. The diamonds should not be distributed extremely awkwardly, the hearts have to be 4-4; then the player should make a lucky choice by playing toward the correct club honneur (50 percent). If the hearts are 5-3, as is the case, then play to CK. In summary, the contract was (too) high, and somewhat luck is needed, but not too much. However, this is not Jack’s best performance.”

Levy: The referee says that NS get to a truly awful 3NT. Even with other distributions of the EW cards and other leads, the contract seems to be a reasonable IMP Vulnerable game (40%). Maas and Vriend point out that luck is needed, but not too much, thus I would conclude that they would put in the category of a reasonable vulnerable 3NT. I ran the NS hands through the bridge analyzer Deep Finesse and determined that 3NT makes 45% of the time. While this is not a definitive test it strongly suggests that 3NT is a reasonable vulnerable game at IMP scoring. In addition I believe that the play was excellent. By ducking hearts twice declarer found that West started with 5 hearts to the KT9 and if West had the CA, and one or both Qs, he might have overcalled 1H. So if hearts are 4-4 Declarer has to guess the location of the honors and if hearts are 5-3 declarer is likely to find the CA in East's hand. Of course, my comment that "the CA has to be in East's hand to make the contract once the hearts are found to be 5-3" was true but that was in keeping with Declarer's best overall play. Is it a reasonable vulnerable 3NT? I say yes. And was it well played? Absolutely!

On the exhibition game: the critique was that North’s 3D preempt actually pushed EW into the game, since it allows East to revalue his hand based on the lack of wasted value in diamonds. East’s 4H bid is virtually automatic in this sequence and with these cards, and probably had nothing to do with “taking advantage of his opponents in the play.” To describe the situation by saying that the preempt “did not stop” Pszczola and Gawrys from reaching the game is therefore misleading.

Maas and Vriend: “3D by Jack is a fine bid, a characteristic of the aggressive style of the current top players, by which they use the advantage of the vulnerability (the opposition is vulnerable and Jack not). Gawrys and Pszczola perform well thereafter; they should do so in this situation. However, weaker players would not have been successful here. The referee states that the 3D bid (unfortunate? or bad?) has helped the players to reach 4H, a contract that they should not have reached otherwise. We do not believe this. If Jack does not interrupt the bid sequence, any top-level pair would have reached 4H easily.”

Levy: The collective comments of Maas and Vriend agree with my comments. Jack's 3D is characteristic of the aggressive style of the current top players. Maas and Vriend also point out that lesser players (and that must include most of the bridge world) would have not bid 4H. I'm sure they didn't mean all lesser players, but it still leaves room for my comment that the preempt did not stop Pszczola-Gawrys from showing their winning style. My added comment that Gawrys may have bid 4H thinking that he could take advantage of his opponents in the play was meant in fun and in seriousness. Other top experts who sometimes play against computers, such as Zia, have sometimes found it to be a winning tactic to bid aggressively against their computer opponents. In time that may be found to be a losing tactic, but that is a discussion for another day.

It is my pleasure to thank the anonymous referee as well as Anton Maas and Bep Vriend for their constructive criticisms to make the report lively and raise the level of discussion. I look forward to the next Bridge competition (- Ed.).