Cover Story / Ken Regan In Regan’s Honda Accord, we talk about
how his chess work has spawned non- chess-related ideas, from how to use computers to grade massive open online courses, to how to think about the future economy. Tyler Cowen, Regan’s childhood friend and an economics professor at George Mason University, is the author of Average is Over, which came out in 2013, and Cowen fills a chapter with predictions extrapolated from Regan’s research. Cowen reports how freestyle (human-computer) chess teams play stronger than computers do on their own and argues that the future economy will consist of high-perform ing human- computer teams in all aspects of society. Regan takes pride in playing a prom inent part in his friend’s book. Randomness affects all aspects of
Regan’s life. His wallet oozes scraps of pa - per that contain names, num bers, and reminders. He doesn’t own a smartphone. When we enter his office, unopened boxes crowd the floor, and spewed across every shelf and workspace lie papers, stacks of books, piles of notebooks, an ancient monitor, a 90’s-era radio, and milk crates full of ephemera. A few months earlier, Regan moved to a new build ing construct - ed by the university and he claims he hasn’t had time to unpack. A clean spot the size of two cafeteria trays makes room for a monitor and keyboard. On another small clearing, con spic uously placed a - cross from us, sits the only item in the room besides the com puter equipment to have received Regan’s apparent care: a framed por trait of his wife. A tab on Regan’s browser is open to a
fantasy baseball site. He loves baseball, and he was watching the 2006 baseball playoffs and logged into
PlayChess.com, an online chess server, when he first heard about the Kramnik forfeit. Regan feels a responsibility to do for
professional chess what steroid testing has done for professional baseball. The Mitchell Report was commissioned in 2006 to investigate performance enhancing drug (PED) abuse in the major leagues, around the time Regan began his anti-cheating work. While baseball enters its post-PED era, FIDE has yet to put a single perfor - mance enhancing device—the chess world’s PED—regulation into place. It wasn’t until mid-2013 that the Association of Chess Professionals (ACP) and FIDE organized a joint anti-cheating committee, of which Regan is a prominent member. In mid-2014, the committee plans to ratify a protocol about how to evaluate evidence and execute punishment. Regan clicks a few times on his mouse
and then turns his monitor so I can view his test results from the German Bundesliga. His face turns to disgust. “Again, there’s no physical evidence, no behavioral evi - dence,” he says. “I’m just seeing the
24 June 2014 | Chess Life
Figure 1
0
0.85
1
On a standard 25-question multiple-choice exam, if a student answered 22 questions correctly and three questions incorrectly (shown here), then such a result would be equivalent to stacking 22 points at location 1 and three points at location 0. The number 0.85 represents the average or “best fit” location that summarizes the student’s overall score.
Figure 2
Partial Credit
Drop Off From Best Answer
Instead of stacking points at location 0 and location 1 on a number line, Regan distributes partial credit over a two-dimensional plot.
num bers. I’ll tell you, people are doing it.” Regan is 53. His hair has turned white. What remains of it, billows up in wild tufts that make him look the professor. When Regan acts surprised his thick, jet-black eyebrows rise like little boomerangs that return a hint of his youth. His enthusiasm for work never wanes; his voice merely shifts modes of erudition that make him sound the professor.
TO CATCH AN ALLEGED CHEATER, Regan takes a set of chess positions played by a single player—ideally 200 or more but his analysis can work with as few as 20—and treats each position like a ques tion on a multiple-choice exam. The score on this exam translates to an Elo rating, a score Regan calls an Intrinsic Perfor mance Rating (IPR). There are, however, three main differences between a standard multiple-choice exam and Regan’s anti-cheating exam. First, on a standard exam each question has a fixed number of answers, usually four or five choices; on Regan’s exam, the number of answers for each position equals the number of legal moves. Second, on a stand ard exam, one answer per ques tion receives full credit, while the other answers receive zero credit; on Regan’s exam, every legal move is given partial credit in proportion to how good it is relative to the engine’s top choice. (Partial credit falls off as a complicated nonlinear relationship based on the engine’s evaluations. Credit also abides by the constraint that all moves taken together for a position must sum to full credit.) The third difference is the scoring method. (See Figure 1) A standard multiple- choice exam is scored by dividing the number of correct answers by the total number
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76