Oliver Roeder has been writing about games since August 2014 when he first landed a gig writing for FiveThirtyEight. His first article for the site was about Nigel Richards, the eccentric New Zealander who was comfortably the best Scrabble player in the world at the time.
Roeder’s book Seven Games has just hit shelves and contains a chapter on Scrabble. Reading the chapter, it turns out that Richards is still comfortably the best player in the world with perhaps one exception: Quackle, a piece of Scrabble playing software.
In 2014, when Roeder first wrote about Richards, the Elo ratings between Richards and Quackle suggested Richards was the better player by a solid margin. In Seven Games, Roeder is much less sure about Richards’s chances (Roeder must speculate since Richards never plays computers).
Quackle has benefitted from the extra half-dozen years of tinkering by programmers. Richards, however, is only human.
I chatted with Roeder earlier in the week about the process of writing Seven Games and what he’d learned from chasing AI from game-to-game.
The shape of things
Each chapter of Seven Games discusses a different game. When strung together, those seven games — in order: chequers, chess, Go, backgammon, poker, Scrabble, bridge — form a bildungsroman of artificial intelligence. The real game the book is interested in is the game of writing computer code that can beat the best humans in a given field.
When I spoke to Roeder earlier in the week, my first question was about where the idea for that arc had initially come from.
“A lot of it stemmed from reporting that I did for FiveThirtyEight,” Roeder explained. “I got started in journalism because I played competitive tournament Scrabble, and I have an interest in games. So I convinced the editors to let games be one of my beats. I covered the chess world championship, Scrabble, crosswords, a little video game stuff — and one thing you can’t help but notice when you write about this stuff is the influence of the computer. A.I. players or other technological influences and this just came up over and over again. There didn’t seem to be a proper cultural/technological history about games, and given my reporting, I felt well-positioned to do it myself.”
Each game in the book presents a new cast of characters with a new problem.
These programmers, players, and academics struggle in turn with depth in chequers, breadth in chess, the scale of calculation in Go, chance in backgammon, hidden information in poker, temporal tradeoffs in Scrabble, and co-operation and language in bridge. Each chapter feels integral to the overarching story, but Roeder still had to fight to get the book into the shape he wanted.
”It was going to be six chapters,” he told me. “But I convinced my editor to let me do a chapter on Scrabble — which is the game in the book I’m personally best at — so I said I could write a really good chapter on that in a couple of weeks, though of course, it took much much longer than that.
“The biggest debate I had was actually whether we should talk at any length about video games. The argument for that is that they have received very interesting computer science attention. And by some of the same companies — DeepMind did Go, but they also did Starcraft.”
The field of AI is a fast-moving one, and I wondered if there were any developments since the book went to the printers that Roeder wished he could have included.
“One of the questions I grapple with in the book is that you have all these great game-playing AIs. But so what? Can they do something in the real world? I think the answer to that is increasingly becoming ‘Yes.’ Though even when I finished the manuscript the answer would have been much more No.’
“I mean DeepBlue is a stark example, they took it apart after it beat Kasparov and it didn’t even play chess again. But the modern neural net stuff seems more adaptable. I’d have liked to explore the transition from the game board to the real world a bit more.”
Roeder doesn’t play poker as much as he used to, though he did pony up $1,500 to play the 2019 WSOP Monster Stack for the poker chapter in this book. When I asked him if researching the book had given him any thoughts on how AI might impact poker, his response was perhaps the most pessimistic of the conversation.
“Personally, I would be very hesitant to play online, because I don’t understand why anyone wouldn’t be using some kind of assistance. If you’re comfortable with the existing technology and other people are too, go ahead and play poker online, but I would be very nervous about it at high stakes.”
His feelings about the other games in the book, where cheating is far less profitable a pastime, is that technology has been a boon, with only a few exceptions.
“I do think that AI takes some of the magic out of these games,” he explains. “But I think that is limited to the highest levels of these games. These games are still available for people like me, an amateur hobbyist. The games are exactly as fun as they’ve always been. AI has approximately the same influence on my chess game as a racecar has on my going jogging.
“Take the chess example in particular — as play becomes more perfect, play becomes more boring. Many suspect that perfectly played chess is a draw. What was the most exciting game of the world championship? It was game six. For the sole reason that the players were running out of time and making mistakes. Mistakes are extremely exciting to watch.
“The other thing that happens is you see unique styles of play dissolve because everyone is using the same piece of software and trying to asymptotically approach whatever the software is doing. You see this in poker to some extent and in the other games as well.”
The more they’re played the better…
But the main thing Roeder wanted me to take away from both the book and our conversation was that these games remain available to all of us as sources of entertainment, edification, and aesthetic pleasure.
“The impact of AI has been very subtle unless you’re among the top ten guys,” he insists. “Among the 99.9% of us who aren’t those guys, computers have largely been a good thing. The biggest part of that is the dissemination of skill. You can get better really quickly. Anywhere in the world you can play a million hands of poker in a year. Whereas before you had to live near a casino or a good home game. If you wanted to get good at backgammon you had to live in New York or London where there were clubs. But now, anyone can become good. Getting a little bit good makes a game far more interesting, so that effect is really positive.”
I mentioned to him near the end of the interview that since reading Seven Games I have been playing on Zyzzyva (anagramming software designed for Scrabble training) and working through bridge tutorials online. That seemed to please him.
“It should come as no surprise that I’m a huge booster for games, I think the more they’re played the better, more or less without limit. When someone sends me an email about an article I’ve written and they say that they’ve started playing chess again or picked up Scrabble, it’s always satisfying.
“That’s journalistic impact as far as I’m concerned.”
Seven Games is on sale now.