On March 2, Levy Rozman, in a pink sweater and round glasses, was streaming his Chess.com matches to 12,000 viewers over Twitch. “All right, this looks like a cheater,” he said without pause, as he clicked on the pigeon icon of his opponent, Dewa_Kipas. Rozman, an international chess master, scrolled through the profile, disbelieving. His opponent had climbed nearly a thousand points in the span of a month, ranking 2,300 to Rozman’s 2,431. And his profile didn’t include the sort of title—“FIDE master,” “national master”—that the ranking implied. In fact, Rozman would later discover, Dewa_Kipas, or “fan god,” was a bird-feed seller in Indonesia.
Chat echoed back: “LMAOOOO,” “CHEATER.” “Let’s see if we can get some content here,” said Rozman.
Rozman has been playing in chess tournaments since he was seven. In 2011, he attained National Master status, and in 2018, International Master. Now 25, he’s known not only on the chess circuit; like other top players, he has developed a large following on Twitch, on YouTube, and on Twitter. High-level chess has experienced an unprecedented online boom due to the pandemic. An average of 895 people watched chess streamers on Twitch on March 1, 2020; a year later that cumulative audience expanded to 21,491.
In this majority-digital world for the 1,500-year-old game, it’s tempting to trust that every truly high-level player would by now be a known quantity—whether through International Chess Federation ratings or social media. Nobody could just come out of nowhere and dethrone a chess king, right?
Rozman knew that if his opponent was cheating it would be a strange game; algorithms often make choices that most humans simply wouldn’t. Still, small things baffled him. At Rozman’s level, obvious moves typically play out in a couple of seconds. Dewa_Kipas regularly took between seven and 10 seconds to make any move against the chess master, even when he had just one or two options. Highlighting Dewa_Kipas’s knight on his stream, Rozman said he was worried that his opponent might move it to A5. “I would expect this, although it is a computer, so I also kind of expect some weird pawn play,” he said. The knight moved to A5.
In their attempts to climb the competitive ladder, cheaters refer to AI-powered chess engines to inform their moves. And as chess has moved more and more online, cheating allegations have skyrocketed, according to top online chess site Chess.com. Rozman has himself turned cheat-hunting into a bit. On YouTube, where his thumbnails are full of silly faces and chess boards, Rozman ran a “catching cheaters” series, which he spun into a similarly themed Discord channel. It’s an entertaining break from the big-brain chess plays and long, thinky matches—lighthearted morsels in the rapidly ballooning content economy of online chess.
Twelve thousand viewers cheered Rozman on as he played against Dewa_Kipas. About 10 minutes in, his opponent’s rook had blocked in Rozman’s king. Black won. “REPORT HIM!” “POLICE!” yelled chat.
Rozman again perused his opponent’s stats. Dewa_Kipas’ accuracy, or similarity to how a chess engine would play, during the match was 94 percent; Rozman’s was 76. Over its last 10 games, the account’s accuracy never dipped below 80. It hit over 99 percent in two of those. Rozman reported him that day. Hours later, Chess.com banned the account.
That night hateful messages, many written in Indonesian, began to fill Rozman’s social feeds. Threats, even. His girlfriend received the same: “hey bitch, we will kill u soon haha, from indonesia :).” Rozman was getting tagged on Twitter, YouTube, Instagram, with accusations that he had leveraged his notoriety to ban a legitimate player. What sparked the onslaught, he would learn, was a post on an anime superfan’s Facebook page.
Ali Akbar, a 24-year-old whose Facebook banner is a sprawl of pink-haired shoujo manga, had written a post about Rozman from his home in Bandung, Indonesia, to his 3,500 followers. According to Akbar, Dewa_Kipas was a 60-year-old retired tournament player named Dadang Subur who had only recently become addicted to Chess.com. Also, he wrote, Subur was his dad. And he most certainly did beat chess master Rozman.
In terse bullet points above an endless scroll of anime memes, Akbar told a much different version of the match: how his father had played a famous Twitch streamer, won, and was then mass-reported by Rozman’s huge fan base. “My father's account was blocked,” he wrote in Indonesian. “I swear I don't accept it … because public figures can block people's accounts at will.”
Akbar’s post went viral. His allegation—that the word of a celebrity chess master is weighed more heavily than an average player’s—resonated with thousands of people, including both lesser-known chess enthusiasts and very online Indonesians fed up with Western internet celebrity culture. Memes flooded Rozman’s private inboxes, some demanding a rematch. A post accusing Rozman’s fans of mass-reporting Subur, a “retired professional chess player,” rocketed to the top of the Chess subreddit with thousands of upvotes. (The thread was locked, moderators wrote underneath, because it was being “heavily brigaded by outside groups” and “unsubstantiated.”) Indonesian news media covered the controversy too, painting Rozman as an entitled celebrity. Rozman was receiving several messages, comments, and tweets every minute in Indonesian.
Rozman tried to explain on Twitter that Chess.com wouldn’t ban somebody just because he or his fans had reported them, that there was a whole process involved to weed out alleged cheaters, but it was no use. He locked down his accounts.
“I didn’t know if I should double down or just apologize to reel it in,” says Rozman. “I knew I was in the right, but sometimes you just take the L. It’s an insane situation.”
At 7 am Eastern time on March 3, Rozman messaged Akbar to reconcile. “Hope we can discuss things peacefully,” he wrote. Akbar said he was sorry that his post sparked harassment, and he offered to explain all of the abnormalities with his father’s account. He listed out the most common questions, and then supplied answers: Why had Subur only played Blitz, or fast chess, a couple of times, when most 2,300-level players have high scores in that discipline as well as Chess Puzzles? (He can’t move that quickly, especially on his old Android.) Why was his rank in Puzzle so bad? (Akbar himself had played it, on his father’s account, “on a whim.”) Why did he only play Rapid chess, in which matches have a 60-minute limit? (He is technologically challenged, and it’s the only traditional chess mode he can access from the app’s main menu without changing settings.) Why did his ranking increase so dramatically? (He played nonstop for hours.) Why were his moves so similar to a bot’s? (Because he trained against bots.) Akbar recorded a 20-minute video with his father in which Subur claims he beat Shredder, a commercial chess engine, on a grandmaster-level setting four times in a row. They buckled down.
Through his son, Subur agreed to an audio interview with WIRED over WhatsApp, later translated from Indonesian to English. Subur says that he started playing chess in junior high. He was self-taught and owns a lot of chess books. He joined several local chess clubs, attended local tournaments, and even trained under his neighbor, the late chess grandmaster Herman Suradiradja. (WIRED could not confirm this.) He likes the Caro-Kann Defense, the Sicilian Defense, and Gambit. He began to play against Shredder, he says, when he got an iPhone 3GS. He believes character and memorization are key to high-level chess play, but tempo, or pressuring your opponents in an optimal number of moves, is the most important. “I mastered the tempo, and when you master the tempo, it doesn’t matter who your opponent is.”
Subur maintains that the game was legitimate. But he condemns the harassment, saying those responsible for it “don’t understand the situation.”
“He is a kind man,” he says of Rozman. “I don’t feel better than him. The play was in my favor at the time. That is all.”
His son, Akbar, shared a Google Drive with WIRED containing a 2005 chess certificate congratulating Subur on first place at the Singkawang City Chess Championship, a photograph of his father holding up a “Jasindo Chess Tournament 2012” banner with several other men, and several videos of a hand flipping through a book of hand-written chess notations from games against Shredder. He also included a message for Chess.com, stating his father has no FIDE rating because he was, before he retired five years ago, just an office worker who was passionate about chess.
“We operate in the realm of statistics,” says Danny Rensch, Chess.com’s COO. “And in statistics, there is never certainty-certainty. There’s always the realm of anomalies.” It’s always possible, for example, that someone could play the best chess game of their life while they’re drunk off their ass. It’s just so very unlikely.
The controversy was already in motion when Rensch entered the fray—“one of the most crazy” controversies online chess had ever seen. “It hasn’t gone away.”
An international chess master himself, Rensch wanted to see Chess.com’s report on Dewa_Kipas so he could assess the site’s decision to ban him. Chess.com has its own seven-person Fair Play team, less like moderators and more like data scientists, which Rensch oversees. The case of Dewa_Kipas got a lot of eyes at Chess.com, in part because of the brigading Rozman experienced after his very public, very confident report. The site hears about alleged cheaters both from player reports and alerts from algorithms, which the Fair Play team uses to help make banning decisions. Internally, they’re called “cheat detectives.” They close thousands of Chess.com accounts every day.
Chess.com’s great challenge is determining what cheating in chess looks like over the internet: Is it a second monitor with a chess AI running? Is it a Discord chat with a grandmaster? Is it a straight-up bot?
The site’s cheat-catching algorithm measures how closely a player’s moves resemble a chess engine’s. As there are a lot of those to choose from, Chess.com pools several together on top of Stockfish, an open source chess engine. What happens next is like the International Olympic Committee’s anti-doping measures, says Rensch. They know what the best athletes are capable of when their white blood cell count and oxygen levels are both normal and elevated. If something strays outside of that lane, he says, it looks like guilt. Winning streaks, other in-browser behavior like excessive tabbing, and whether they’re a FIDE-recognized or title player are all factors that the algorithm and moderators weigh.
“We operate in the context of knowing who a lot of the best players in the world are already,” says Rensch. Rensch points to Alireza Firouzja, an Iranian kid who attained grandmaster status at age 14. “This kid grew up on my site,” he says. “I’ve been watching him play like a monster since he was 11.” Rensch would not say how many people have accused Firouzja of cheating. It’s a lot, though.
Like many online platforms, lesser-known users often accuse Chess.com of protecting its stars. Onlookers scrutinizing Rozman’s stream drew a direct line between his report and Chess.com’s ban. Privilege. But Rensch says nothing could be further from the truth. They’ve even banned chess masters over statistical anomalies. “We are legally ready to go to court with every closure we make. And we act on data and evidence,” he says. “Did [Rozman] actually have influence over any of the final decisions? Zero.”
Rensch says Dewa_Kipas was an “absolute, absolute certain case” of cheating. Over dozens of games, Chess.com determined that Dewa_Kipas’ moves in chess games matched a chess engine at a rate that is “not reasonably possible for a human player,” higher even than the top-ranked Indonesian chess player, grandmaster Susanto Megaranto: 95.3 to 94.4.
Rensch always tries to remind top players that they’re human. They’re emotional. They’re competitors. The most neutral actor here is the algorithm. There’s a famous saying by Emanuel Lasker, the longest-running World Chess Champion in history: “On the chessboard, lies and hypocrisy do not survive long.” And the algorithm, while not perfect, cuts down on that survival rate. “We’re acting with no emotion and no investment other than to protect the integrity of the game,” Rensch says.
For Rozman, the nasty YouTube comments are down from 10 every minute to one every few minutes. More supportive messages in Indonesian have appeared in his Twitter mentions, some expressing shame for the brigading and harassment. He’s scared, but interpreting the controversy as a lesson on the dangers of misinformation. “It’s really scary,” he says. “It would be nice to just get the word out that something as insane as this could happen.”
Back in Bandung, Akbar and Subur say they are attempting to get an Indonesian chess master to their house, so Subur can play them live on camera for a national news audience. Subur feels confident that he’ll win.