“Decision: case of using computer assistance in League A”

[Bill Spight]

Reading over some of the posts in this topic it seems that we expect someone to use a computer for much of the game.

It turns out that that has been, and still is, a pattern of cheating in chess. You have players whose every move is, according to top chess engines, neither a blunder nor a mistake, and who play only a few inaccuracies per game. Their rating never topped 2200 and suddenly they are playing like super grandmasters.

Such moves will typically be among the top three choices of any given chess engine, which is where I suppose the match one of the top three choices heuristic came from.

[Fenring]

If i am not wrong Carlo just won against a european 6d in World Amateur GO Championship.
He was table 3 of Italy, now represent his country in WAGC and win against a 6d.

[Uberdude]

If i am not wrong Carlo just won against a european 6d in World Amateur GO Championship.
He was table 3 of Italy, now represent his country in WAGC and win against a 6d.

Yes, he beat Ben DG 6d, the current French champion, a good win. Ben also lost to a 4d from Thailand so maybe he's jetlagged . I will be following Carlo in the thread about the WAGC. (maybe people missed it in the usually quiet Tournaments subforum)

[Fenring]

Benjamin just say he missed the first round because the alarm clock was not powerful enough.
I didnt missed the tournaments forum,but think its related to this topic too.

[carlozero]

I think bannishing Carlo was mistake.

I like to play with Leela and i learn a lot with it.
When i am playing with Leela i use analysis windows to choose one of the best move. But it is my choice not Leela choice.

Playing Go is expression of my spirit, sometimes i prefer to play moves that is slightly under Leela best choice, because i want to be human not superbot.

So i think Carlo is an artist, and we have to let him express his art with Leela tools.

Carlo is much more in advance than normal go players, he already understand that human is nothing with no AI

[OferZ]

So the WAGC has just ended.

I think the results tend to agree with the decision.
Carlo has won 3/8 games. One of them, surprisingly (considering his current rank), against Oscar Vazques (5d) - but that's just a one rank difference.
Among his losses:
Amir Fragman 5d(the Israeli team captain, btw, who's recently lost to Reem Ben-David - the "famous" opponent in the said game).
Juri Kuronen 6d from Finland
Vietnamese 5d
and Willem Koenraad Pomstra 5d from the Netherlands

Some of the games (rounds 3 and 6) were recorded and can be found here:
https://www.nihonkiin.or.jp/event/amaki ... cords.html

(sadly the game with France didn't take place, and the other win was against a 1d/2d from Iceland)

I didn't really explore the game records, although the result against Korea in the opening was interesting...

[Javaness2]

It is probably not a good tournament to examine his overall performance with. Given that he has just been accused of cheating, it can't have been the happiest of times, nor the best of preparations for such an event. For me it is more interesting to analyze the individual games themselves. It is always interesting to analyze Go games(!) but here we would be able to see if his PGETC style was holding good, or if he was playing in a different way, or generally just make some banal observations on the games.

[Bonobo]

So … you really registered for this comment, and with that nickname?

I think bannishing Carlo was mistake.

I like to play with Leela and i learn a lot with it.
When i am playing with Leela i use analysis windows to choose one of the best move. But it is my choice not Leela choice.

Playing Go is expression of my spirit, sometimes i prefer to play moves that is slightly under Leela best choice, because i want to be human not superbot.

So i think Carlo is an artist, and we have to let him express his art with Leela tools.

Carlo is much more in advance than normal go players, he already understand that human is nothing with no AI

Now, if THAT ain’t bad taste, then I don’t know.

[Jan.van.Rongen]

I am still very much surprised that nobody is really building up a defense for Carlo here. Let's look at the situation after move 70. Before that move all engines tell you B was behind, but 70 was a mistake that brought B a little bit in the lead. Black only temporaril;y increased his lead afetr another slack move at 84, but around move 124 and move 140 the evaluation is almost ewual again. Then white makes another error and loses a group. So white's loss is mailny caused by his own mistakes, not by an opponent who continuesly chose a top move from leela as the accusation was.

Evaluation was done with AQ, which is (~one stone) stronger than Leela 0.11. I aslo used leela zero for this - same result. That was in april, when AQ was still better. Leela Zero is now at least 2 stones stronger than AQ.

[hyperpape]

I am still very much surprised that nobody is really building up a defense for Carlo here. Let's look at the situation after move 70. Before that move all engines tell you B was behind, but 70 was a mistake that brought B a little bit in the lead. Black only temporaril;y increased his lead afetr another slack move at 84, but around move 124 and move 140 the evaluation is almost ewual again. Then white makes another error and loses a group. So white's loss is mailny caused by his own mistakes, not by an opponent who continuesly chose a top move from leela as the accusation was.

I'm not sure I follow what you're trying to say.

To pick up one thread, I'm not sure who was ahead really matters. Here's a counterfactual scenario: suppose I use Leela to pick all my moves, but unknown to me, my opponent is really strong against Leela, because they've learned some pattern it can't handle. I lose. Nonetheless, I cheated.

[Uberdude]

I am still very much surprised that nobody is really building up a defense for Carlo here.

Hmmm? I've done quite some analysis myself which has thrown considerable doubt on the statistical methods used, Bill has written a lot about the weakness of confirmatory evidence and compared to Regan's work in chess. Plenty of others have chimed in and I'd say the general mood here is sceptical of the conviction.

Or do you mean something more formal to the appeals committee rather than internet chit-chat? People may have taken actions not publicised here.

[Uberdude]

The appeal was successful.

The Italian appeal has demonstrated the weaknesses of the analysis and
rightly questioned its conclusions. With the available tools it is hard if
not impossible to convince someone of using an outside help in the
internet games. Thus, we do not think that it was proven without a
reasonable doubt that Carlo Metta broke any rules and the decision made by
the league manager should be reversed.

http://pandanet-igs.com/communities/euroteamchamps/409

[Javaness2]

I'm pleasantly surprised to see such a quick decision, the EGC will be a happier place now the chief referee has found his innocence.

[dfan]

I am also pleasantly surprised to see that the statement is so clear and non-defensive. It can be very hard for people to backtrack from decisions.

[Bonobo]

The appeal was successful.

[…] Thus, we do not think that it was proven without a reasonable doubt that Carlo Metta broke any rules, and the decision made by the league manager should be reversed.

I’m not a native speaker of English, but I think the comma I added here is essential, otherwise it could also be interpreted as follows:

we do not think that [it was proven […] and] the decision […] should be reversed.

Am I mistaken?