Right now, Chess is having a doping problem.
CW Drugs, but also no, not drugs.
This isn’t about people using chemical stimulants to improve their performance in game situations and in tournaments. Fans of The Queen’s Gambit might be already familiar with the idea of chess players getting some variety of high or stoned and focusing their way through games, but this isn’t what I’m talking about. Besides, Chess, as a game that’s been played by, largely, privileged and wealthy people for centuries, so back in the 18th century, just having coffee was doing atypical stimulants.
It isn’t that, that’s not the problem, no no no. I mean, I don’t think that at this point, people are trying to drug test competitors at chess tournaments. Maybe they are, I’m not familiar with those floor rules.
Doping is the term we typically use to refer to athletes doing things generally that involve artificially altering their biology to make them better at a sport. The term is pretty broad, now, because one of the things with professional sports and cheating at them is that there’s a bit of an arms race going on about types of cheating. It’s why rules about what not to do are either incredibly specific and actively patrolled, or kinda general.
There’s this idea, goodhart’s law, which summarises as ‘when a measure becomes a target, it ceases to be a good measure.’ If you measure people’s performance and skill based on some criteria, that can give you information about well they’re doing. But if people know that’s what you’re measuring, and want to present good results, they can structure their behaviour to present those results. This is usually used as an example of why you don’t let people know testing outcome parameters, because then they’ll be reshaping their behaviour to the test.
I don’t have a good name for it, but I am familiar with an alternate form of Goodhart’s Law, so I guess we’ll call it Trahdoog’s Law, where defining the limits of the rules results in people doing exactly as much as the rules permit. In a social moderation system, if you list the bannable offenses and don’t include a catch-all ‘we can make judgment calls’ addition, you will be argued around that infractions aren’t technically infractions. It doesn’t matter how obviously something should be seen as against the rules, if the rules define these boundaries technically and specifically, then people will always find ways to circumvent them that are ‘technically not breaking the rules.’
You can often get away with that in some kind of play environments, where you let people break the rules, then address the rules when they’re broken, but as you scale up, this becomes untenable. You need people to learn a general social ruleset, and you need to do that by removing people who aren’t willing to play nice. And that’s the problem with referring to cheating with specific terms.
One of the common forms of cheating in the Olympics lately has been ‘blood doping,’ where the subject increases the amount of red blood cells in their blood; not necessarily through the use of drugs. Sometimes it’s just giving blood at time A, waiting for the blood to replenish, then reinfusing your own blood at time B. This isn’t obviously ‘performance enhancing drugs’ – after all, it’s your own blood! – but it’s definitely a form of performance enhancement that requires some medical expertise. The points of human limits in high-level competitive sport like the Olympics are pretty much this constant arms race between ‘well technically’ and ‘no, really,’ and it’s interesting, but also means that the Olympics are, well
I don’t think much of the olympics.
Anyway, but what, are people injecting blood to get better at chess? No.
The quandrary is pretty much a 2020 phenomenon, where for obvious reasons a large amount of chess play has moved from face-to-face play in physical spaces on physical boards to online play. What’s been happening is there’s a rise of chess players who are playing online chess in online tournaments, with chess applications or AI in another window that they can feed the board state into and get ‘optimal’ play.
What’s wild about this is, because the players are definitionally in a remote location and unobservable, there are only so many things you can make as a requirement of engaging with the tournament that can give this behaviour away. The more boundaries you put to prevent this kind of cheating, the fewer people you’re going to get engaging with your game – just because it’s harder to.
This means there’s a sort of witchcraft of trying to spot people doing this, and it isn’t just hard, it’s hard in a way that only appears in aggregate. Just sitting there and constantly making optimal plays according to an AI is something that you can track by running the same AI and comparing the moves side-by side, right? You can feed the game state into an AI, then track how often the moves of the game reflect what an AI would do. But that is going to catch someone who just happens to be copying an AI move for move – what about if they’re only using it in short periods in difficult periods, running the AI during knotty game sections and otherwise playing naturally? Do you watch for long pauses for processing of moves that may be handing the information over to the AI? What about times when a player acts really fast?
Also, what about false positives? What if a player just plays slowly? Managing your time is a skill in Chess, and it’s natural to think about complex moves a lot. What if a player pauses, thinks hard about a move, and makes the best move? That’s kind of what chess players want to do all the time!
And then, if you do deduce cheaters in a tournament over the course of the tournament, what do you do with every win they’ve gotten? Do you contact all the players they eliminated and let them know they’re back in? What about the tournament structure can be made to handle this? Are eliminations or resistance the way to go?
By the way, I don’t have a solution to these problems. I just think the problem is, itself, fascinating.