Practice makes perfect but do statistics make swiftest?
Oct. 11, 2021
From sports to music, anyone can improve with practice—or a keen sense of statistics. Data analysis in science, finance or elections is nothing new, and even baseball and other sports make extensive use of statistics for optimizing training and teams. But Takeru Matsuda may be the first to dissect game play in competitive karuta using the power of statistics. In this arcane Japanese game, players must learn poems on pairs of cards, memorize the cards’ placement and snatch up the right card as soon as they hear the matching poem. The best karuta players can tell in a fraction of a second, from just the first syllable, which poem is being recited, and the game involves frenzied slapping of cards.
For Matsuda, karuta is both a hobby and an example of practical statistics. While he’s been playing for over 10 years, he is also the leader of the Statistical Mathematics Unit at RIKEN CBS, so naturally he decided to do a study on karuta. “I was interested in how fast we can discriminate the card, and the speed of course is related to dominance in karuta,” says Matsuda. During the match, “the player hears the voice and infers which card is being read. So I made a mathematical model of this process, basically giving an AI the voice data and asking which card it is.” Most people cannot make these discriminations quickly but, like the strongest human players, Matsuda’s model can tell which poem is in question from the first syllable. Even when many first syllables sound essentially identical, with statistics you can pull out unique knowledge from data, says Matsuda. “In a computer, images, video or audio are just arrays of ones and zeroes. It’s nonsense, but with statistics, we can get meaning out of it. Statistics is a viewpoint for accomplishing data-driven understanding or prediction or control”, whether of complex game-play in the case of karuta or in other domains, like making sense of the behaviors created by the brain.
“In karuta, players listen to a spoken poem
and match it to the corresponding card as fast as possible.”
Matsuda is particularly keen to apply his statistical methods to data from the brain. “Each brain has its own internal world. Data from other fields don’t have another level to them,” he says, “but the brain thinks or feels things itself. Deciphering what is computed through data from the brain requires another level of abstraction, which makes the brain special compared to other applications of statistics.”
Without a neuroscience background, Matsuda approaches studying the brain according to the motto ‘think like an amateur, do as an expert’, coined by Carnegie Mellon professor Takeo Kanade. “You start from an amateur idea, then tackle a problem using sophisticated thinking,” says Matsuda. “I also want to start from the beginning and try to develop my own ideas, for example addressing the interplay between mathematical elegance and practical, real-world applications.”