Controversial Go opinions

[benjito]

It would reduce to perfect play surely.

[square.defender]

or to random play that ends in jigo

[Vsotvep]

And, the devil player would need the ability to see the future as well. Consider a normal player, who flips a coin at positions to decide their move, surely this would need the devil player to play perfectly as well, since he can’t know beforehand which choice the normal player will make.

Or take a normal player who flips a coin to decide whether to continue reading a move out further. Then at no point can the devil be sure that the normal opponent will manage to read out the bad sequence or not.

---

Thus a devil player would need physically implausible assumptions to function, while a god player only needs enough space and time, but not necessarily a physical impossibility.

[gennan]

@Vsotvep

I also do not see how to formulate a strategy that encapsulates the idea of playing for a bigger lead on the cost of needing your opponent to make a mistake. How to determine that a move is “difficult to read”? The devil would need information outside the game of Go to be able to formulate such a strategy, such as information about the strategy of the opponent (i.e. common mistakes). It cannot be a universal strategy therefore.

Well, the god would use no other information than the current game state and a full tablebase (full minimax annotated game tree of go, including handicap games). It would not use any knowledge about its opponent.

The devil would use that and all other information available in the universe. If we feel that it’s cheating by using more information than the god, then maybe we should formulate a rule to prevent this (and OGS should also hide names, ranks and any other information about opponents).

[benjito]

@Vsotvep

And, the devil player would need the ability to see the future as well.

Yes, I think I said this:

@benjito

I think the devil has mind-reading/predictive capabilities as well.

But yeah, this is physically implausible. That’s why I said this is a physics/engineering problem. (Although one might way the same of a God-perfect player as well)

[tonybe]

Given a perfectly spherical Go player moving on a frictionless goban…

[benjito]

@gennan

and OGS should also hide names, ranks and any other information about opponents

OGS Rules and Regulations

…

394.5.1 No universal simulations (computational or otherwise).

By the way, I think it’s funny that this topic has become a discussion on “white devils” >.< Where my Gwailou at?

[antonTobi]

@Vsotvep

I also do not see how to formulate a strategy that encapsulates the idea of playing for a bigger lead on the cost of needing your opponent to make a mistake. How to determine that a move is “difficult to read”? The devil would need information outside the game of Go to be able to formulate such a strategy, such as information about the strategy of the opponent (i.e. common mistakes). It cannot be a universal strategy therefore.

I agree that this is hard (possibly impossible) to formalize, which is why I want to keep ideas like this separate from the idea of perfect play. It’s something quite different, but we can’t deny that such play exists in reality. There are no mindreading devils, but I could most likely crush a 20k by a much larger score margin than a player who just plays an arbitrary perfect move in each position, because I have some limited understanding of their understanding of the game.

[tonybe]

As long as we’re performing abstract Gedankenexperiments with mind-reading / theory-of-mind, I feel like we should include Perfectly Rational Pirates and Philosophical Zombies. Now that would be a fun game of Go!

[benjito]

And now we need Idiots and Rational/Devil Pirates.

Rational pirate A gets thrown overboard for taking 98 coins. Devil pirate B survives by only taking 10 coins and letting the idiots split the rest.

EDIT: Devil can probably do better than that just by splitting the pot with one idiot.

[ababab61]

Considering GO was originated in China, perhaps “Wei Qi” would be more appropriate than “Baduk” (which appears to be a Korean term).

[Samraku]

@mekriff

@le_4TC

Would it be practically possible to make the 0-point completely random play?

Possibly… but it would require a couple things: 1) a random play player that is anchored to the 0 point, 2) that this random player would have to make a notable effect on the ratings pool (by playing against other players/bots), and 3) that the entry points are properly spaced away from this 0 point

@le_4TC

Yeah I’m not aware of any human rating pool which is anchored at random play, it’s probably a bad idea in practice (despite its aesthetic appeal).

A compromise could be to decide on some rank more in the middle as the “real” anchor (could use some bot to avoid having to pick human anchors as KGS did (does?)), but then estimate the distance to random play empirically, and offset the ranks such that 0 approximately lines up with it.

The Arimaa Gameroom Rating. And yes, it does implement this by actually tying to 1000, not 0. A variety of weak bots ranging from random play to ArimaaScoreP1 (1000 rating) were made to play eachother until their ratings stabilized relative to eachother, it was then noted that ArimaaScoreP1 was roughly 1000 elo, and also around the level of a complete human beginner to the game, making it a good candidate to anchor the rating system.

[dragon-devourer]

And now, the ultimate controversial opinion for this forum…

DGS is better than OGS

(And now, let the backlash commence, LOL)

[benjito]

@dragon-devourer

DGS is better than OGS

Obviously the person who eats dragons is going to prefer a Dragon server

[dragon-devourer]

And the ultimate controversial Go opinion:

Go - it’s just a game…

No! I take it back! May the Gods of Go have mercy on me!

[su2]

Going back to a perfect devil story. I was playing 6h games against Kata-bot here on ogs and 6h games (as white). It lead me to realize that it is good to think about the board state by two numbers: score of perfect play and complexity. So good handicap player uses some points (wrt to perfect play) to change the complexity of the boardstate in the direction it makes sense for him. I expect there are usually more than one move in the perfect play, so sometimes it is not even needed to play suboptimally to choose the more complex play.
And I would try to define the complexity as how many (almost) optimal moves there are…

[bugcat]

I sighted this comment on Sensei’s Library, by the KGS 2d user Calvin, very early in nachtrabe’s blog. Controversial part in bold.

Among top amateurs (9d KGS) and pros who teach a lot, I haven’t seen much difference in the type of advice they give. Somewhere below amateur 3d or so you hear all kinds of crazy stuff that stronger players would disagree with, so that can be confusing. You can learn from them, too, but Guo Juan gives this advice: when asking a stronger player to review a game, ask one and no more than one question, and do not let them digress too much, because to them, almost every move you make is wrong, and if you let them get into that, you’ll get very confused.

[bugcat]

Comment from the general Review page on SL:

Getting a stronger player to review your games can be very helpful, since he will see possibilities and problems that neither players ever looked for. A DDK should look for someone 10 stones stronger, and a SDK should look for someone 5 stones stronger. This ensures that your reviewer will be able to find plenty of useful suggestions and variations about which they are absolutely confident, and which they can explain to you easily. Someone three stones stronger than you may have insights to share, especially in their favorite area of the game, but is better regarded as a peer than a teacher.

To map that onto a conversion chart:

Player rank	Mininum suggested reviewer rank
1k	5d
2k	4d
3k	3d
4k	2d
5k	1d
6k	1k
7k	2k
8k	3k
9k	4k
10k	1d (note this bizarre retrograde jump!)
11k	1k
12k	2k
13k	3k
14k	4k
15k	5k
16tk	6k
17k	7k
18k	8k
19k	9k
20k	10k

[Atorrante]

Why that jump?

[gennan]

@bugcat

A DDK should look for someone 10 stones stronger, and a SDK should look for someone 5 stones stronger.

One could interpret those 2 data points differently. For example:

Player rank	Difference	Mininum suggested reviewer rank
6d	2	8d
5d	3	8d
4d	3	7d
3d	3	6d
2d	3	5d
1d	3	4d
1k	4	4d
2k	4	3d
3k	4	2d
4k	4	1d
5k	5	1d
6k	5	1k
7k	6	1k
8k	6	2k
9k	7	2k
10k	7	3k
11k	8	3k
12k	8	4k
13k	9	4k
14k	9	5k
15k	10	5k
16k	10	6k
17k	11	6k
18k	11	7k
19k	12	7k
20k	12	8k