And the AGA fought tooth and nail to keep it’s win in the Transatlantic Tournament when the European team had internet issues despite a proctor being in the room and confirming that he did play his move on time. 
At this point they really should publish an in depth explanation and put measures in place to insure it won’t re-happen.
I think you really have to compare these graphs to those of other games, preferably lots of them. If you leave the big blunder of move 105 aside for now, it is really a matter of one side consistently playing better moves than the other. This is something that I see in games where one player simply is better than the other. This is not something where one player has a bad day or such. Bad days result in blunders and blind spots. When there is a sloped curve, it indicates something that I nowadays correlate with »actual strength difference«.
This means that as far as I can tell, the black player was at least one or two stones stronger than the white player. If white is a 6 dan, then black is a 7 dan or 8 dan at minimum.
This means that it must be reproducible. Let him play another game, but in a controlled environment, let’s say, some proctor. If he is really this strong, he can still play at this level, examination stress nonwithstanding. He doesn’t have to win or lose, he has to demonstrate the curve. If this is not practical, just get the record of any future live game. Time will tell.
I must say that the little time used is not a good sign. It seems to me that it is consistent with entering a move on a different screen, choosing something from there, and entering it in the game. Of course, it is also consistent with a pro player making his rounds in a simultaneous game. Not saying it was one or the other.
As for the strategy used to »dampen the obviousness«, in the case that the player cannot demonstrate his strength, I’d guess that this might be a downgraded AI (such as »Bender« at leela-1-playout), which might explain the few blunders.
One move that I really found astonishing is Black’s 113. Even if it occurred to me that I could play here before living on the side, I’d spend a lot of time making sure not to overlook something. Beautiful move.
According to a message on Facebook, Black spent 2 seconds before playing move 113.
Well, the issue is that the likelihood of making a mistake is a function of both the player’s strength and the complexity of the board. I would guess that the stronger the players are, the more complex the board becomes. This means that any system for detecting AI cheating should probably work with many games from the same player, hopefully playing against a broad range of strengths. It may be possible to adjust for the complexity of the game in some way using the outputs of a NN.
There is only one way to settle this.
We form two teams with opposing views and fight it out on the board.
I’m in. Let me open Katago and I’ll be back in a few minutes.
Savage, lmao
Yoonyoung Kim - I joined the Pandanet AGA (American Go... | Facebookft=top_level_post_id.3806064726152092%3Acontent_owner_id_new.100002458068454%3Aphoto_attachments_list.%5B3806057212819510%2C3806057306152834%5D%3Aphoto_id.3806057212819510%3Astory_location.6%3Astory_attachment_style.album&refid=18&tn=%2C%3B
Open FB posts are for us 
A 8p participates the league …
Of course it is allowed by rule, thus not cheating. But looks very much open robbery for the prize money to me and when the plan did not work out …
I have no idea how these leagues work, I’m interested.
It’s an open tournament, with 3 players from the city per team. Whoever wins 2 matchs out of three win and there are promotion to a higher tier (with higher prizes) when your team finish first in it group (and demotion if you are last). Basically the same system as the Chinese city leagues except online.
Robert Tirak’s finding in that thread look pretty incriminating if they can be confirmed. Thanks for digging that up.
Are there comments in the FB post? I’m on my phone, FB comments don’t show on Android afaik
Yes including one were a poster claim to reproduce all of black’s move in this game and 2 previous games from last year in the same league by picking katrain’s move with the most playouts (after 2000 total playouts). I haven’t reproduced that because I neither have katrain nor the other games’s kifu but I’m sure someone will shortly.
Those are the two games mentioned above.
I did look at his 2019 games, obvious k level.
Starting to think he really did cheat.
And that, my friends, is why I believe a “big stink” was justified! If the organizers don’t address an issue like that before the event, then what are the chances they are going to handle it correctly without some external pressure?
The win against jk2015 was played in March 2020, jk2015 is AGA 5 dan. It wasn’t a landslide like the game against the 6d but the level of play was clearly not kyu.
Sadly even if he does turn out to be a cheater I still can’t support the public accusations. I have seen a lot of people have there life turned upside down because of things like this. If there is some chance you will be wrong and ruin someone else entire life for it. Then you should be prepared to bet your own.
The unfortunate part is lets say Remi turns out to be wrong (likely I think he is right) but lets say he isn’t the backlash on him won’t be as bad as the back lash will be on the 1kyu even if it turns out he wasn’t cheating.
Basically the idea is that the players involved in accusing will be safe from the harsher back lash that the 1kyu will now 100% receive regardless of weather he actually was cheating or not.
I get what you are saying though that in order to make something happen it had to be drastic. Its just sad that the person accusing always has less to put on the table.
Like if you are a fake victim you will always have the happier ending. Even if you don’t actually win anything as long as you have incriminated the fake villian you simply win.
(Don’t misunderstand of course not saying there won’t be backlash if the player turned out to not cheat Remi would still receive back lash. But, it would mean 10x less then what the cheating player would have received even if it turned out he wasn’t cheating. “Why” you might ask. Because the fake victim has only incriminated somebody but did not actually do any real criminal act. And “Why” would the 1kyu still get backlash if he is cleared of suspicions. Because the people who incriminated him will likely still incriminate him and not clear up the accusations of there own. So there will always be “somebody” out there who says “They are right that 1kyu is a cheater so when I see him I am gonna call him a cheater” or if the 1kyu ever reaches a high ranking he will be called a fake.)
