Brow Beat

In Defense of Proper Names in Scrabble

The Scrabble community was rocked Tuesday by widely exaggerated news that an imminent rule change would make it acceptable to play proper nouns. The reports sparked a glut of hang-wringing over the alleged demise of this sacred pillar of board gaming. It was bad enough arguing over obscure medical terms without debating whether string theorist Baton Zwiebach is famous enough for his 27-point surname to count.

In fact, the new rule applies only to a European spinoff of the game aimed at a younger audience. But the most devoted Scrabble fanatics might want to think twice before celebrating. Had this news been true, it would have been the best thing to happen to human Scrabble players in almost 15 years.

Scrabble is what game theorists call an “imperfect information” system, meaning the players do not have the same data when making a move; like most card games, they know what’s in their rack but not in their opponent’s. (Chess or checkers, by contrast, are games of perfect information; both players can see every piece.) Master Scrabble players do not merely have large vocabularies. They can think ahead to what letters a new word leaves behind on the rack, how compatible they are, and what the other player is likely to be holding based on which letters have not yet been overturned.

It naturally follows that a computer has certain advantages in this game. In December of 1986, a program called MAVEN, designed by Brian Sheppard, went 8-2 in a match against Scrabble grandmasters, eventually finishing second in the tournament. That early version of the program, which Sheppard describes in a 2002 paper (PDF), had the official Scrabble dictionary memorized and could fine tune its strategy over time, developing a policy of “rack management” that could think ahead to future moves. MAVEN has continued to evolve, particularly in its ability to simulate games and play wisely when only a few letters remain.

Allowing proper names would be humanity’s revenge against the machine. The computer can learn its countries and state capitals, but how can it follow the torrents of news and culture that introduce new words and names on a daily basis? I’ll see your QUINSY and raise you XIUQUAN , Hal.