Weak score estimator and Japanese rules

Again, I get that you’re fond of your creation, and I’m not asking what you like about your creation.

I’m asking for what you think is the optimal tool that should be given to players. What should it do, what is the criterion by which you measure how well it is doing its job, why is each of these uses (if more than one) necessary to implement?

I told already

3 Likes

I think the current estimator should be removed, so players cannot use it during a game to cheat (cheating is still possible in other ways that cannot easily be plugged). Instead, there should be a new better, AI-based scoring button available optionally during reviews, during analysis when not playing a game, and for advising a player automatically when resignation is overdue by several moves. Notice: just my opinion, and I am only 10k.

1 Like

Don’t we already have this?

I wouldn’t really want that in game, though.

1 Like

We kind of do, except that it is not available to all players like the Estimate function is.

Then it should only be done for unranked or beginning players. The problem this is meant to solve is described above: beginning players sometimes don’t resign, but instead keep playing useless dead stones.

How can you cheat with the score estimator?

I don’t think there is an AI that exists that can tell a 10k player when to resign. It would have to identify not just live and dead groups, but groups the actual players know how to defend or kill.

2 Likes

its not about cheating, its about not wasting time on boring counting. We just need estimator that gives 0 Go knowledge and calculates area in purely geometric sense. I gave examples how to do it in posts above.

Such estimator cannot be used to precisely estimate current score, but it perfect to know when to resign.

1 Like

When it is your turn to play, enable Analysis mode, make any simplifying moves that occur to you, then click on Estimate Score. This gives you a clue, especially if you are a beginner. Then return to Game mode. Of course, this procedure won’t work in Blitz games, but it should work in the other two kinds of game lengths.

The AI can detect when a player makes a move that has only one liberty and is later captured. When such sacrifice plays happen in N consecutive moves, the AI can inform the beginning player that resignation is customary at the end of a game when one is losing.

I think that case is better handled by a human, either their opponent or a third party, so they can learn how to end a game.

1 Like

Who is the human that would do this in practice? It’s not exactly polite to tell your opponent to resign, and I don’t think playing while (significantly) behind is a reportable offense

1 Like

Not only would it be inaccurate in estimating score, it would be inaccurate in determining group status. I think especially the latter issue would make a simple algorithm useless as a resignation advisor.

3 Likes

its just not doing it. User need to click group himself to make it dead - its fast and user not gets additional Go knowledge that way. After this its pretty accurate.

There are some cases, such as endgame trolling, where it is reportable, but I think we’re primarily talking about beginners who don’t know better, and a mod can be called in that case to help explain scoring if you don’t feel confident in it. It’s hard to find a balance, but I do think one should attempt, when playing someone who may never have played a game to scoring before, to offer a bit of advice on how to close up the last few boundaries and pass.

2 Likes

Most players who play while significantly behind, even massively behind, are weak kyus who do so by playing self-atari moves or infilling. Those are both reportable offenses on OGS.

1 Like

There is no reason to burden either a player or a volunteer with the task of teaching how a go game ends, when the OGS software can take care of it. Surely you agree with this point.

I don’t think of it as a burden, helping someone get started in the game, and an explanation from a human you can ask questions of can be a lot easier to understand than other alternatives.

I think this is a tricky matter. When teaching beginners IRL in the context of (say) a beginners course, yes that is what would usually happen and I think it works well in most cases.

But online may be a different story (maybe especially so in ranked, “competitive” games). Opponents are usually complete strangers. There are players from different languages and from different cultures. There is no body language to support communication. All of this makes is more likely to accidentally offend newcomers with well intended but unsollicited advice, let alone giving them (subtle or not so subtle) hints to resign.

2 Likes

I don’t either, and I used to do a lot of go teaching to beginners. But now I like to play quick games, do a fast analysis, and move on. I no longer wish to tutor people who make useless moves. You and I will have to agree to disagree on this one point in our discussion. If you are willing to take a poll of OGS players (I believe this forum supports polls), then I would concede this point only if the poll shows that I am in the minority. I think most single-digit kyu players or above want to focus on games, not tutoring.

Therefore, I stick to what I suggested in my OP. Thanks for the discussion.

Mmh, I just noticed at some point that I don’t enjoy teaching complete (stranger) beginners because basically everything they do is wrong and I doubt anyone benefits much from being told just that.

Their time is better spent learning about which moves are legal and fooling around with other complete beginners. When they mess up equally often, they tend to enjoy the mutual fails.

The next step is doing lots of 1-3 move tsumego to grasp the very basics of life and death.

But it’s not just absolute beginners. Sometimes I gloss over something when discussing a game with ~10k players simply because explaining why something they did or thought was wrong would require me to explain something else first, so effectively it would turn into an exercise in yak shaving.

1 Like

I’m not talking about teaching games, but just not dragging out the endgame with pointless invasions. I agree that a teaching game doesn’t serve much point, and a complete beginner would be much better off playing games and, if they have the motivation, doing some (level appropriate) tsumego.