For clarity, as you said, this is what is currently implemented.
However, it also makes people uneasy wondering if they are playing a person or an AI.
We like to know that when we lose, it was a “deserved loss” - a person more skillful than us defeated us. We do not like the idea that we lost “because that person felt the universe owed them a win so they decided to cheat”.
For this reason, we continually review the AI detection methods available to us, and deploy them the best we can, so that our users can feel that “what can be done is being done to prevent this”.
I don’t think we have an “obective measure of how big of a problem it is”. We certainly receive a flow of complaints about it - accusations that people are doing it - and regularly warn and ban users as a result of analysis. This is not a “rare occurrence”.
This doesn’t tell you “how big the problem is” but it does tell you that “the problem is real”.
That’s fair. I never said that you should not address cheaters and make clear that what they are doing is wrong. I want to note that I didn’t argue with the quality of the recognition. I even assumed that you can be 100% sure, but argued that even then the current procedure is very fair and should stay. Furthermore, what I did say is that you should be calm and not too harsh when approaching this issue. My point is that it is not reasonable to socially outcast someone because they made a mistake that ultimately only has stolen some time and nerves but did not impose any serious damage. I think that it is fair to first look if someone is able too reflect and tries to change before making something public.
I will say that worse things happened to me on OGS (e.g. a certain troll reregistering and threatening me with physical violence multiple times). If I suspect someone to have cheated, I block the account, maybe report it if I have more than a feeling and move on.
Cheating is not always completely obvious, I’ve had strong suspicions in about 1 game out of 1000 among my games. I guess that about 1% of players are cheaters at all levels. I’ve seen players weaker than 10k who found many dan-level moves in the same game.
Sometimes it’s not about just the first move but the follow up.
Examples might be like in Tsumego, some people can just get lucky and guess the first move for the “right” or “wrong” reasons, but then also getting the right response each time, isn’t always as obvious.
So the probability of getting any one move might not be unrealistic, but
becomes less and less likely in the same game, for example.
Still, that would be the best holistic analysis that players could do by themselves, without additional tools. I believe as well that some people, go teachers, pro players etc could be quite good at it, when they have a lot of experience of seeing players at certain levels make the same kinds of mistakes or have the same kinds of blind spots.
I had a clever idea that you might ask a number of (European) professionals around the world to play against players suspected of being used in the AL. What do you think?
You thought about the idea of asking a number of pitchers to develop new pitchers to reach different levels for the rest of the team to play against players still playing in the AL to tell the managers but there are things they can do who am I telling them what to do
I’m at least trying to say something good about this discussion here. I’m happy if in the future it’s important that there be a robot to help them locate the players. Remember what I said here. Thank you and have a good day.
My answer was serious, I think pros have better things to do as regulating the online cheating.
Now it’s not so obvious to detect cheating in a sure and firm way so this is quite a problem which I dunno how to help myself. I trust that all go servers are interested in ways to make things better.
On a positive note, I stand by my previous position that, the harder cheating is to detect (i.e. the more it looks like a human player), the less harmful it is for the opponent.
I can only tell you that you can definitely detect cheating if a player makes 100 to 80 percent accuracy against an equal or higher level player. If he uses one move, he makes one move. The AI does. You get to 80 to 68 percent accuracy for a whole game. That’s about it.
For a proper objective judgement with few false positives and few false negatives, I think there would need to exist thoroughly tested and standardised procedures.
I think obscure hand-wavy arbitrage procedures won’t be good enough to be accepted by players, referees and sponsors of high profile online tournament games.
How would you calculate those percentages? Are those the percentages of moves where the player plays the blue move, or perhaps top 3 or top 5 moves? Which AI do you use for this reference? Which version, number of visits, other settings? Or would you reference a selection of AI in some sort of voting scheme?
Would you use the same citeria for games that ended during the middle game versus games that ended during the endgame? How about time control: correspondence, live, blitz?
Does the established level of the player matter? Like pro, amateur dan, SDK or DDK player?
I think a lot of testing would need to be done to answer those questions and converge on a highly reliable procedure (with few false positives and few false negatives). I think some group of experts would have to invest a serious amount of time (and money) to reach that goal. It’s not something that can be done by just anyone in a single afternoon.
If organisers, players and fans are not convinced that the procedure is highly trustworthy, they probably won’t submit to it, or they will reject it again as soon as it leads to some controversial decision.
I was trying to give some input on what I thought this topic was about (automated cheating detection?), but I suppose I’m just not understanding what this topic is actually about, which point you’re trying to make, and what you’re proposing. Perhaps that also has to do with some language barrier.
You’re right, you said what you think is right to say. My point is that I would like something that understands the game. Make software that aims to analyze games in real time and if it detects cheating and can stop telling the administrators that there is something suspicious on the site, it’s worth it for them to see the same game that it stops.