Humans versus Jack (1)

The Computer Plays Bridge, I Don't Know Whether to Laugh or Cry

(published in Dutch Bridge Magazine IMP June 2005)
"...but a computer cannot play chess and will NEVER be able to play chess. Or at least not the first 2000 year. Why do I have to repeat myself again and again?"

With these words the Dutch international chess grandmaster Hein Donner emphasized his opinion in 1981. By now we have reached a new millennium and every chess player knows Donner was completely wrong. Donner, who died in 1988, didn't have to swallow his words because it took about another ten years before the supremacy of 'that thing' over the human mind was established. By now, almost any chess player prepares and analyses his games with the chess program Fritz or one of Fritz's nephews.

Compared to computer chess, computer bridge has only just been born. But peering over the fence at other intellectual games we can wonder what awaits us. How long will it be before computer beats Man at the bridge table?

That question is not easy to answer. There are fewer people programming computers to play bridge than chess. More significantly, chess is a game of complete information, but bridge is not. A chess player sees all the pieces on the board, his own and his opponent's, but a bridge player sees only his own hand during the auction and only his own cards and dummy's during the play. The information obtained from the bidding usually provides only a sketchy picture of the two or three unseen hands. A bridge player must act on inferences from the calls and plays of others to form assumptions about their hands; the better the player, the better his inferences and assumptions (paradoxically, the better the opponents, the more reliable these inferences may be!). Based on these assumptions, a bridge player makes his 'moves'. This makes it hard to predict how well computers will eventually play bridge.

Instead of guessing, why not have computers and people play against each other? After Jack won four consecutive world computer bridge championships, his developers invited some top Dutch players to play against him. The following pairs accepted the invitation:
Ton Bakkeren-Huub Bertens
Jan van Cleeff-Vincent Kroes
Paul Felten-Erik van Valen
Erik Janssen-Jeroen Top
Bart Nab-Gert Jan Paulissen
Anton Maas-Bep Vriend
Bauke Muller-Simon de Wijs
Vincent Ramondt-Berry Westra

And, by special invitation, Jack-specialists:

Just vd Kam-Hanneke Kreijns

Most of the players have represented or still represent The Netherlands in international championships. Perhaps you think that we overmatched Jack, but we have observed that players below 'Meesterklasse' (the highest league in Dutch competition) simply make too many errors to provide a good test for Jack. A 'first Division' player (second league) can play reasonably well but will always slip and make careless mistakes, and Jack beats second Division pairs very easily indeed.

In the Netherlands we have about 100,000 bridge federation members. The Meesterklasse consists of twelve teams. Crediting each team with three pairs gives us 72 top players. There are about 150 first Division players and 300 second Division players. All together, there are slightly more than 500 players skilful enough to compete against Jack on roughly equal terms, slightly more than half of one per cent. The other 99½% have no chance whatsoever in a match against Jack. Here's a typical deal to illustrate my point.

East deals
Both vul
J 9 7 6 3
A 7 4 2
K J 10
J
K 10 4
K 10 9 8
8
10 9 7 5 2
N
WE
S
-
6 3
A Q 9 7 6 4 3 2
K 6 3
  A Q 8 5 2
Q J 5
5
A Q 8 4
West
Jack
North
human
East
Jack
South
human
--33
pass4passpass
pass

It is about half past eleven at night and we are playing Board 26. Only two boards to go! This information is totally unimportant for Jack, but after three hours of play his human opponents are tired. After East's preemptive opponing, North and South manage to bid their game and the declarer is very satisfied when he sees dummy. East wins the first trick with Q and returns 2. Declarer prefers not to trump, and throws a loser. Well done... but wait a moment: instead of a heart he throws a club. West ruffs and returns a club. Declarer takes two clubs, ruffs a club and takes the A. After eliminating clubs, he throws West in with a spade to the king.

Not good enough. West simply exits in hearts and gets a heart trick later. One down. Of course throwing a club is very bad play, below the normal standard of this player. At another time and place he would not have made this error. But experience has shown that at other times and places, other things will go wrong. People tend to make mistakes and it is very difficult to avoid them.

Of course this is not the only advantage a computer has. To name a few of the others:

  • A computer never misunderstands the calls of other computers programmed with the same instructions, never forgets a convention, and is always on the same 'wavelength' as his electronic partner.
  • A computer neither loses track of the played cards nor miscounts a hand.
  • A computer never dwells on past errors or unsuccessful outcomes.
  • A computer never quarrels with his partner.

But let's be fair. Humans also have their advantages. The most important one is no doubt the ability to be flexible. Thinking in bits and bytes makes Jack very rigid. If someone opens a 15-17 notrump, until otherwise shown, Jack counts him for 15-17 HCP, no more and no less. Humans, however, sometimes deviate slightly from their supposed requirements for calls or more to the point, the more gifted humans can not only judge when to deviate but can also recognize when their opponents might be deviating.

All in all, pitting Jack against top human pairs figures to be very interesting. The Jack team will try to keep you informed. It is hard to predict the outcome, but I don't think that Jack will be blown away. I don't think it will take long before computers acquire the same standing they already have in the world of chess world: as aids in analysing and improving human play. I must admit that even now, when I write bridge articles, I no longer check only my spelling, but I also check my analyses with Jack.

Wim Heemskerk
translation by Danny Kleinman