Good point. We also (at least from the info in this thread) don’t have the full story from the AGA’s perspective, the full story from Remi’s perspective, or the full story from the 1ks perspective.
It’s entirely possible that more is going on here than just one game; the 1k may have just been unlucky enough to get caught in cheating prevention reforms.
I believe people tend to play safe when they are ahead. However when behind, most people don’t play safe, but instead start fights (even if they are a little unreasonable), to create opportunities for the opponent to mess up.
It makes perfect sense to me that from your perspective, 2D+ players play very safe, because presumably they get a steady lead early on.
At the same time, it makes perfect sense to me that a 6D falls apart and plays crazy when they are in a losing position and unable to catch up. That’s usually how it goes when you play against an entity that is much stronger than you.
Ok, I finished reading the AI suggestions on OGS and can honestly say I could’ve played all black moves and better. White was 80% lead around move 108.
At this point, black decided to invade top middle, which is normal in my view, but the problem is the 6D was unable to defend his territory.
Finally, white played horrible yose in the end games based on AI suggestions, many I can see cruel moves that I can play, but not high D players.
Example: move 142, -19%
144, -19%
146, -15%
148, -21%. what the heck is this move? I did not get it at all
Note, all black move preceding those white moves are negative double digit moves, aka, cruel moves. Then a 6D responded every single time.
You ask yourself this question: with an AI on hand, why does black play those cruel moves?
Without any evidence other than “I don’t think it’s right” I don’t see how, to be honest.
As I said, for changes to happen, he needs to do his part with more than “I lost, sus”.
He named names.
I would appreciate more a discussion if he pushed for webcams without bringing another player (who was cleared by the organizers) into it.
No, it’s not about win or lose. A high D player plays totally different way than us k players naturally. They play far few cruel moves. They know the value of typical yose and know to tenuki without thinking. Now I know he is indeed a 6D, I am convinced he was drunk or sleepy, or doing something else on the side.
You can look at the game and tell me if you as a 4D would play like that in a normal day.
To add, OGS AI indicates it is quite an even game with up and downs of course till the end game. The 6D never played anything aggressive to “catch up”, rightly so because he did not need to. My thoughts is he just assumed he would win the game and played safe to respond every black’s move, then found out in the end …
Look at this picture, I would not get myself into white situation if I played someone 5 levels below me, with a risky invasion, four key stones captured, unless I was drunk, sleepy, or watching EPL …
Actually, that’s the part where the 6D managed to turn the game around because black backed off (when black was winning the scary capturing race). Typically, when a stronger player comes back from a losing position, it’s with sequences like this where you need strong reading to maintain the lead. After that, the game was super even and I don’t see any large endgame mistakes on either side. If black was using an AI, I would expect them to just extend at N9 for move 105, because there would be no way to be sure even katago could win the endgame after Q11. This isn’t really strong evidence either way.
Here is Katago’s graph. Green is winning %, purple is difference of points.
White was behind by 11 points, so made a crazy invasion at move 92. At move 105, Black let White live, hence the drop in the graph. Other than that, and except for suboptimal endgame moves, most of the game was a smooth cruise towards victory for Black.
Understanding that sort of question fully would be the start of an automated cheat-detection system, in my opinion. I can’t answer it, but maybe some more knowledgeable people can. They may not actually want to in order to keep any developing systems secret and thus harder to fool.
Indeed. I think some work was done on that a couple years ago when there was another scandal but I can’t recall and the bots back then (like elf, Leela and LZ) didn’t agree much with each other. From watching pro games with commentators using AI, I feel like 100 perfect moves never happen.