On Oct. 12, in the basement of a Unitarian church on the town green in Lexington, Mass., a carpenter named Michael Cresta scored 830 points in a game of Scrabble. His opponent, Wayne Yorra, who works at a supermarket deli counter, totaled 490 points. The two men set three records for sanctioned Scrabble in North America: the most points in a game by one player (830), the most total points in a game (1,320), and the most points on a single turn (365, for Cresta's play of QUIXOTRY).
In the community of competitive Scrabble, of which I am a tile-carrying member, the game has been heralded as the anagrammatic equivalent of Wilt Chamberlain's 100-point game in 1962 or Don Larsen's perfect game in the 1956 World Series: a remarkable, wildly aberrational event with potential staying power. Cresta's 830 shattered a 13-year-old record, 770 points, which had been threatened only infrequently.
Since virtually all sports involve variable conditions, comparing one performance to another is technically imperfect. Consider the absence of black players in Babe Ruth's day, or the presence of steroids in the Barry Bonds era. On its face, the new Scrabble records seem to avoid such problems. No one's juicing in Scrabble. Points in a game are just points in a game, and Michael Cresta scored 830 of them. On Scrabble's members-only list-serve, Crossword Games-Pro, most players have hailed this harmonic convergence of vowels and consonants as a triumphal moment. But the record-worthiness of the shot heard 'round the Scrabble world is more complicated than it might look.
Let's begin with the fact that Cresta and Yorra aren't expert-level players. They know the basics—like the 101 two-letter and most of the 1,015 three-letter words—but they're both rated in the bottom third of tournament players. In Lexington, where the record was set during the club's regular Thursday-night session,Yorra is known for trying implausible words and hoping they're in the Official Tournament and Club Word List. Cresta has memorized thousands of obscure words (like those ending in WOOD or starting with OVEN) by reading, writing down, and tape-recording pages from the Official Scrabble Players Dictionary. But he doesn't study the highly probable wordsthat are essential for climbing the competitive ranks. "These are not guys who have low ratings because they haven't played in many tournaments," Mike Wolfberg, the Lexington club's statistician, told me. "They have low ratings because they aren't very good."
So, how did they break the all-time Scrabble scoring record, set during a tournament by two experts, one of whom has been known ever since as Mr. 770? The simple answer is that Cresta-Yorra was a fluke. Given that Scrabble is played in more than 200 clubs and there are more than 200 tournaments a year in North America, the thinking goes, it was inevitable that Mr. 770's record would fall, especially with the growth of serious study and an increase in words in the Scrabble dictionary.
But there's more to it than that. To understand how Cresta and Yorra broke the record, let's take a closer look at the game. (For the full play-by-play, click here.) Yorra opened with JOUSTED, a "bingo"—Scrabble lingo for using all seven tiles, which earns you an extra 50 points—worth 96 points. Cresta then traded in all seven of his tiles in the hope of getting more-playable letters, not an unusual move. Yorra bingoed again, very nicely, with LADYLIKE for 73 points and a 169-0 lead. The first L in LADYLIKE landed between two triple-word-score squares, giving Cresta a shot at Scrabble's holy grail—a "triple-triple," covering two triple-word scores with one word. That's worth nine times the value of the word, plus the 50-point bonus for using all seven letters.
Triple-triples are rare in Scrabble—I've played no more than a dozen in a thousand or more games—because they require a confluence of mathematically improbable events. Cresta's play, FLATFISH, for 239 points, was especially unusual because it contains infrequently occurring letters (two F's and an H) and isn't a common word. Many good players would have missed it. Cresta didn't because he had studied words beginning with F.
Yorra challenged FLATFISH, a reasonable move given the word and its score, but it was in the official word list, so he lost his turn. Cresta exchanged tiles on three of his next four turns, while Yorra bingoed again, this time with SCAMsTER. (The lowercase letter represents one of the game's two blank tiles.) Yorra told me he had no idea whether the word was legitimate. (It is.) SCAMsTER was simply the first possible bingo he saw. That put another letter, the R, in a triple-triple lane. Cresta, who held I, O, Q, U, and X,recognized he was three-quarters of the way toward a really huge triple-triple: QUIXOTRY. (He had studied words starting with Q.) He exchanged two letters from his rack in hopes of drawing the needed T and Y. From Cresta's vantage, 56 tiles were unseen, including three T's and one Y. The probability of pulling one of each was 1 in 513. *
Cresta beat the odds. And when Yorra didn't block the open R—because he played his fourth bingo, UNDERDOG, for 72 points—Cresta laid down his 365-point QUIXOTRY (a quixotic action or thought).
After making just three plays, Cresta had an amazing 614 points. The rest of the game was pedestrian. Neither player bingoed again, though Cresta played the recently added word ZA (short for pizza) for 66 points. When he laid down VROW, a Dutch woman, Cresta passed 770. (For a cell-phone-camera image of the final board, click here
Looking at the game as a whole, it's clear that a lack of expertise created the conditions for the record. The play that enabled QUIXOTRY, for one, was a clear mistake. When Yorra played SCAMsTER, which scored 65 points, there were eight other bingos available worth 72 points or more that wouldn't have dangled a letter in a triple-triple alley. Among them were several common words, including the 94-point dEMOCRAT. Most players would have taken a few extra moments to search for one of those moves.
I asked Jason Katz-Brown, a Massachusetts Institute of Technology junior ranked 10th in North America, to analyze the game. Unlike most players mid-level and higher, Cresta and Yorra didn't keep track of the letters they drew on each turn, so it's impossible to fully examine their possible moves. But we do know what letters they played on each turn. When Katz-Brown input those into a Scrabble-playing computer program he co-wrote called Quackle, he found that Cresta and Yorra had better moves on 14 of their 22 nonbingo turns. One example: Cresta scored just 30 points using the second blank when he could have held it and tried for another bingo.
Technically, Cresta's strategy was unsound. Fishing for a once-in-a-lifetime play might be understandable in a casual game, where winning is less urgent. But in competitive play—even in a club setting, where there's less on the line than in a rated tournament—exchanging letters three times, as Cresta did, to enhance some combination of Q, U, I, and X is unorthodox at best, suicidal at worst. (The strategically correct move was to dump the cumbersome Q and move on.) In Scrabble, the player who waits for the miracle word usually loses. The implication: Cresta wasn't terribly worried about whether he won or lost.
"If they weren't really trying to win," an intermediate-level player named Mike Eldeiry wrote on the Crossword Games-Pro message board, "then can we really consider it our record? Fun, yeah. Neat, sure. Promotable, why not? But record, ummmmmmmm, I don't know." Eldeiry told me the game reminded him of a 600-foot batting-practice home run. If experts always shot for the moon, he said, "I think they'd have cracked 850 by now. But they'd have lost a lot of games in the process."
Most CGP posters defended Cresta and Yorra. Lexington-club regulars said they just played differently than Joe Expert might have. The democratic Scrabbling message: Even someone who doesn't study word lists for hours on end can achieve greatness. "Non-experts often make suboptimum plays," wrote Rod MacNeil, a top-100 player who witnessed the game. "This time that resulted in some pretty eye-popping plays. But they found them." Another expert, John Van Pelt, said, "When faced with the possibility of playing a Q-X triple-triple, they see it as a good opportunity to advance their winning chances. So they go for it."
Cresta, who is 43 years old, didn't start playing Scrabble competitively until a couple of years ago. He told me he loves learning and playing unusual words; at carpentry jobs he sometimes transcribes dictionary pages onto walls or sawhorses. In the record game, Cresta said he went fishing as soon as he drew Q, U, I, and X. "I wanted to get QUIXOTE down bad, or QUIXOTIC." When SCAMsTER hit the board, he immediately spotted the possibility of QUIXOTRY. But he also realized that those other words were possible. "I like to gamble," Cresta said. "I'm trying to win the game, but I'm trying to get that word down, too." Strategy wasn't a big concern. "I'm not playing a top player."
The difficulty posed by this game, and by games in general, is judging the role of circumstances in the commission of records. In this case, the sensible moves would have been just another set of moves in just another game. The wrong moves produced history. But is that enough? If 830—or any record—happens as a result of boneheaded play, tactical ignorance, or the pursuit of a good time, should it count? Or should records be reserved for those who have earned the right to set them, and who set them in expert fashion?
Here's what I think: Michael Cresta holds the record for club play, while Mr. 770 keeps his tournament mark. And here's what Michael Cresta thinks: "It's really not that big of a deal because I'm really not that great of a player. If you get two experts together, that game's not going to happen."
Thanks to Jason Katz-Brown, John O'Laughlin, Seth Lipkin, and Keith Smith.
Correction, Oct. 30, 2006: This story incorrectly stated that 57 tiles were unseen when Michael Cresta exchanged two tiles in hopes of drawing a T and a Y. There were 56 unseen tiles at the time. As a result of this error, the probability of Cresta drawing a T and a Y was misstated, both in the original piece and in a previous correction. The correct probability is 1 in 513. (Return to the corrected sentence.)