When AlphaGo beat the greatest human player the world has ever seen at the complex board game Go in 2015, it was seen as a landmark point in AI development.
Over the decade after this achievement, we’ve continually been told AI is the answer to everything, whether it’s planning your next vacation, curing disease, or ending civilization as we know it.
Sick of hearing AI is the infallible, multipurpose tool we should all be using so we don’t fall behind? Rejoice then, because another board game is proving AI isn’t as smart as it is, and the billionaires want us to think.
The board game? Scrabble.
Related
How a K-pop love song could normalize AI companions, digital affection, and cyberlove itself
Love is in the algorithm
AlphaGo vs. Lee Sedol
Move 37
Without getting too deep into AlphaGo’s 2015 success, it’s important to recognize the achievement.
Go is a strategy game like chess, but a larger board and the number of possible moves each turn make it even more complex to master.
When AlphaGo played world champion Lee Sedol, at least 200 million people watched the stream, highlighting the match’s interest and eventual cultural impact.
AlphaGo’s famous game-winning “move 37” made Sedol “think about Go in a new light,” and it was estimated that a human had a 1-in-10,000 chance of making the same move.
Demis Hassabis, co-founder of Google DeepMind and “father” of AlphaGo, called the win a “watershed moment” for AI.
Dominating chess and Go
But not Scrabble
Computers had been beating the world’s best chess players since the late 90s, and when AI beat Sedol at Go, it was easy to think that our feeble, squidgy brains couldn’t hope to beat specially developed software at even the hardest games out there anymore.
Since then, AI has been pushed to the forefront of every major tech announcement. Companies have pivoted entirely to developing AI, and there’s probably not a single job left that doesn’t use an element of it every day.
While thinking about what I’d do when AI makes me redundant, I watched a random YouTube video about the world’s best Scrabble player. The story is about Nigel Richards and is fascinating, but one thing in particular surprised me.
In the video, and according to data from the World English-Language Scrabble Players Association (WESPA), it’s shown that during the endgame, where no letters are left in the bag to use, in tracked matches, Richards had a cumulative error rate of 1.2%. When analyzing the same games, the top Scrabble AI had an error rate of 4.9%.
Richards is shown to be an astonishing player, and someone who the current Scrabble AI simply can’t beat.
Not so clever now, are you!
AI still can’t consistently win
Scrabble has been around since the 30s and became a household name in the 50s. Various AI programs are designed to analyze and play Scrabble, including Quackle and BestBot, and they can beat the world’s best players. Just not all the time.
Richards’ lifetime win rate is estimated to be 75%, meaning he wins three out of every four championship games he plays. What’s more, he’s not alone, and a handful of other elite Scrabble players have a similar win rate.
Outside the best-of-the-best, WESPA puts the average win rate for its players at an even 50%.
Scrabble AI engine Quackle is estimated to have a 50% to 55% win rate, while BestBot increases this to around 60%. For context, the best chess AI systems, such as Stockfish, almost never lose to a human.
Nigel Richards and Scrabble are AI’s true gaming nemesis, and they offer a glimmer of hope for all of us.
It’s good, but it’s not the best
Computers lack the skills, for now
The numbers above show that Scrabble AI engines still aren’t quite as proficient as the world’s best Scrabble players. This isn’t to say they’re bad, far from it. In a series of 500 games against Scrabble grandmaster Kenji Matsumoto, Quackle won 248 to 252.
Although impressive, this is far from dominant. The AI makes mistakes and can’t think as creatively as a player like Matsumoto or Richards.
As an example, in a video by Will Anderson, a Scrabble champion, it’s explained how the Quackle engine didn’t see a series of audacious, game-winning moves from Richards, and instead, its approved moves would have resulted in a draw.
In a fascinating Medium post from 2022, BestBot creator César del Solar called Scrabble a far from “solved” game, and included a quote from grandmaster Matsumoto, who said:
I have no doubt that some day Scrabble computers will achieve superhuman strength. But as of now, computers in Scrabble are far from reaching the levels of excellence that they have reached in other games.
Responding to a post on the Hacker News forum in 2024, del Solar also offered up a reason why AI isn’t dominating Scrabble yet. He wrote:
It’s that Scrabble is more than just about finding the top-scoring word. There’s a lot of board shape considerations, inferences, volatility, etc. that current engines, including mine, don’t yet take into account.
In 2025, Solar said in an interview:
My hope is that we can get something that can beat Nigel Richards, convincingly. It sounds hard, but I think it’s possible. Give us a few more months or years, and we’ll get there.
The figures don’t matter
Humans matter
Falling down the Scrabble rabbit hole and starting to understand how special the top players in the world are has been fascinating. This isn’t the place to go into depth, and it’s not my area of expertise, but I recommend you look into it more if you’re interested.
What I wanted to highlight is that while two games which are often associated with incredibly clever people — chess and Go — have been conquered by computers and AI, a game that is likely (and probably wrongly) seen as less cerebral is still ruled by humans at a championship level.
It takes creativity, knowledge, and both language and mathematical skills to excel. While the creators of Scrabble AI keep making more capable engines, I’d rather they don’t succeed.
If Nigel Richards retires from the game, I also hope another exceptional player is ready to take his place and keep AI from the top spot once again.
If, like me, you sigh at yet another instance where AI is heralded as the answer, remember it still can’t beat the best human players in the world at Scrabble, a game we’ve probably all played around a family table. I find the thought quite uplifting.