Not too long ago an occupational survey of top bridge players would have revealed that more of our experts had been trained in law than in any other profession. Today a similar survey would almost surely show that the list is headed by computer scientists.
Does this mean that a computer might play better bridge than a mere human being? Hardly. You can feed the thing with data but you can't make it think. More than 30 years ago a genius named William Patzer taught a machine to defend a single hand so that it would always defeat the logical game contract, no matter what line of play the declarer chose. In later years, with far more sophisticated equipment at their command, computer experts calculated the winning way to play blackjack (twenty-one), but about the best they were able to do with bridge was to teach a computer to deal hands. Which reminds me of a letter Ely Culbertson once received: "I have mastered the shuffle and deal. Now what?"
A more recent attempt, reported in the September 1971 issue of Psychology Today, was made by a San Diego computer consultant named Chris N. Napjus, who claimed to have successfully programmed a computer to "learn" to play bridge. After 800 hands, the computer attained a level that would be considered fairly good for a beginner.
The uncanny element of this achievement was that the computer, programmed to play only as the declarer, seemed to learn as it played, but in time its improvement rate dropped. It got into trouble when faced with a problem like the one on the deal shown here. As an example, put yourself in the South seat and try your luck at a four-heart contract.
A computer—and not a few of the human experts holding the West cards—would be programmed not to make a takeout double of a bid of one major suit without having four cards in the other and thus would miss, as West did here, any possibility of finding a good save at five diamonds, which goes down only one because the defenders cannot make more than one heart and two diamond tricks. (Five clubs would prove unprofitable because North could lead his singleton diamond; South would give his partner a ruff and get back in with the heart ace for a second diamond ruff that would give them 500 points.)
Since East-West did not sacrifice, your task is to make four hearts. First, count your losers after ruffing the second club lead. You will find none in the red suits, so your job is simply to avoid losing three spade tricks. Any robot could be taught to draw a round of trumps and then ruff out South's two losing diamonds and dummy's one remaining club. It could also be programmed to lead a spade from dummy and duck the trick on the theory that West might be forced to win it, in which case he would then have to make a return favorable to the declarer. But as the cards lie, East would hold the first spade trick and his spade return would give the defense two more tricks to sink the contract.
Thinking produces better results. After ruffing the second club, you lead a trump to dummy, ruff the last club, get back to dummy with another trump and take an "unnecessary" finesse in diamonds, losing a trick in that suit but getting back two. West, in with the king of diamonds, cannot make a winning return. A spade lead sets up your king and you can then discard another of dummy's spades on your ace of diamonds. A club lead lets you ruff in your hand while discarding a spade from dummy and a second spade goes on the diamond ace. And a diamond return into your ace-jack gives you two spade pitches from dummy.
Did you find the winning play? Or did you play like a computer?