Restarting games from scoring tool

[Maverick_GO]

As a lower level (ddk) player I find that the end of game scoring can unfairly influence games. I admit that I’ve been on the bad side of it even if you review my games (sorry opponent X).

In the instance below, my opponent did not restart the game, and because I had won I elected not to play it out either.

The issue as I see it is this. 2 players pass and both approve of the board, but then Kata Go says ‘that group is dead, territory unsettled’. In most cases the player that benefits from this info restarts the game and kills said group and swings the game. This is a bummer I think.

I don’t have an immediate solution, as I think making the scoring tool dumber would negatively impact higher level players but I thought it was an interesting topic.

Maybe during end of game scoring mark groups less aggressively, and leave it up to the players to contest them?

[GreenAsJade]

It’s been talked about before, but I find it actually an interesting question to explore:

What is the purpose of the scoring tool during the scoring phase?

Initially (and in the documentation) it was “to avoid the players having to manually mark all the stones, an assistance in counting”.

In recent times it has become “the thing that tells us precisely what the score is”.

It’s the latter “feature creep” that causes this problem…

[txwolf]

Did not get the question, it’s always players responsibility to score. In this case, tool just does the initial marking, players need to correct and agree to conclude the game.

In the case when the tool is not complete, it should take players seconds in my experience to mark the dead, confirm the board.

In the case players don’t agree, call the mod aka referee, no argument needed.

In this particular board, left side, if black is unsure that the two white is dead, black should add a move before pass, thus lose a point in certain ruleset.

If black is unsure about this area, but sure about its advantage is more than a point, it can simply add a move.

Now when black passes by mistake, white has an opportunity to win the game if it knows it. That’s fair game.

Then in this particular case again, if black passed earlier, white restarted the game, black should be able to defend it. Otherwise it was a bad pass to start. A little unlucky white realized later and took advantage of it, but black had the mistake first and sort of deserve it,

This is in theory though, I’m not seeing why black could not defend the left side. If black can not, well…

At last, the tool may frustrate and confuse beginners, should never trouble high level players. Maybe I’m missing the point here.

[GreenAsJade]

The point is that the scoring tool shows players area that are vulnerable that they did not realise are vulnerable, so players return to play to try to take advantage of that, after having seen the scoring tool’s appraisal.

[txwolf]

That I understand. So it is a problem for beginners, not high level ones. I think it is better off to continue to play out to see if the restart is valid. It’s a little lack of character to take advantage of the tool, but like I said earlier, the other player did not know better to create this situation. If it is an invalid play, but black does not know how to defend, black bears some responsibility too. White could’ve played before the pass anyway. So it’s not that “unfair” or unlucky.

Having this played out and learn the lesson is more valuable than a “faire” outcome of one game in my opinion. We make mistakes all the time, for example, to resign presumably. Not that big of a deal.

If one is really concerned about this rare scenario, maybe only play Chinese ruleset or alike, so one can feel free to add more defensive moves within its own territory. I don’t know. Regardless, this does not happen often, nothing to worry about too much.

[DVbS78rkR7NVe]

We use KataGo for score suggestions already?

[teapoweredrobot]

@txwolf

if black passed earlier, white restarted the game, black should be able to defend it.

I would like to know how a restart is handled. Is it always the other person who plays first? I.e. in this case black would mark the two stones as dead, white restarts the game and it’s black to play? Or can it be that white restarts and plays? Say if black passed last and the moves continue from there regardless of who asked for the restart.

[txwolf]

Interesting good point, I actually did not know

[txwolf]

I know now. If I remember correctly, When we both pass, it goes into scoring phase. At this point, both players can restart the game if they disagree, they can resume.

We can do a test to verify

[DVbS78rkR7NVe]

If you need to score, you didn’t fight hard enough!

[teapoweredrobot]

I’m not sure what you mean. I think it’s that either player can restart the game. There is a notification for “game restarted from scoring phase” so this must be something that can happen without one player’s input.

But then who plays? The person who didn’t restart or the person who’s turn it would have been?

For me if it’s the former then there’s not much problem (at least in this example.) But if the latter then there could be issues with playing on with new information.

[teapoweredrobot]

And what about correspondance games? Is it possible that one player passes and then the other passes, sees the score, restarts, it’s the other player’s turn but they have not seen the score state so have no idea where is unsettled or not and maybe pass again…

[Lys]

@S_Alexander

If you need to score, you didn’t fight hard enough!

We need a new meme: “scoring is for pussies”.
I guess what picture could be appropriate.

((actually I am that kind of pussy: I hardly resign a game))

[shinuito]

@txwolf

At last, the tool may frustrate and confuse beginners, should never trouble high level players. Maybe I’m missing the point here.

I think you completely missed the point alright.

@txwolf

not high level ones.

It’s a problem for every level below professional and nearly professional players I’d say. If katago is hinting there’s a weakness somewhere by marking intersections as neutral or groups as dead, strong players especially might be able to use this info and figure it out that there’s a seki somewhere etc.

[shinuito]

Unless it becomes OGS policy to have Katago in the score estimator essentially tell players how to finish the game, I feel like games like these should just be annulled. Not this game necessarily of course, but games where players gain new information and then resume.

This is the problem I alluded to a long time ago actually in effect, much more clearly than I imagined even.

Use KataGo for Scoring

Is the following an issue? In principle when the current system does it’s initial territory estimation it’s in an intelligent way in that it kind of knows which stones are alive and dead according to the rules, seki can be tricky etc. (It does some Monte Carlo playouts or something does it?) When katago does it’s territory count and saying which stones are live and dead it’s probably correct sure with some superhuman ability. What if at the end of the game katago declares some corner a seki o…

[teapoweredrobot]

@shinuito

I feel like games like these should just be annulled.

This seems not really the right solution though. Surely it should be that the indication of dead stones/settled areas be limited somehow to avoid this. Eg only stones or groups of stones in atari are marked dead or some such cut off in certainty that leaves more for players to mark themselves but not leaving it completely unassisted?

[shinuito]

I mean more so in the short term. I can imagine people being quite frustrated if say a person gets to scoring and thinking they’ve won, and then their opponent decides to go back and play again after seeing katago’s suggestions.

I don’t think it would be unreasonable for someone to ask for such a game to be annulled unless it genuinely is to become site policy that it’s ok. It’s very similar to getting to the scoring phase, having a friend message you to say “Hey the game isn’t finished you can capture theses stones” and then you going back to continue the game.

In any case yes some other long term should be implemented. @Vsotvep mentioned some ideas for example on how the usage of Katago could work

AI based score estimation and auto-scoring is now live

So, I mean the following algorithm: KataGo checks the position with White having next move KataGo checks the position with Black having next move Any discrepancy between these two is marked as “unsettled”, the rest is marked as “settled”. Chains of unsettled area (open intersections + unsettled stones) are considered as a whole. If a chain has empty intersections bordering on settled stones of both colours, this is a dame Else, if the number of unsettled Black stones is larger than the number …

I tried a suggestion as well

AI based score estimation and auto-scoring is now live

I think unsettled area during the game is a bit vague, and it’s probably fine if an estimator can give some points in areas that are a little uncertain. (Of course you don’t want that if you’re using this kind of thing in ranked games where you don’t want a strong ai opinion of what will likely be your territory.) On the other hand at the end of the game “what seems reasonable to both players” is also a bit vague but I do have a suggestion: Let Katago decide what stones are alive and dead ba…

AI based score estimation and auto-scoring is now live

In @pdg137’s example here’s some screenshots assuming pdg didn’t realise they could break in and pass. Black connects [image] White passes unlike the game [image] Black passes [image] I think the unsettled stones, namely G9, F8, F7, G7 and F6 should all be marked alive since they’re not pass alive, otherwise one runs into the problem of trying to distinguish “dead” stones in a territory from actually dead stones which assumes katago played on from the position. It would mean that the…

but I think it still would just give too much info away that there’s still moves to be made in the area. So likely something similar to Vsotvep’s algorithm is better.

I think though when you know it’s katago giving an opinion, it might still be hard to avoid some situations where it gives away a hint that there’s some aji. I guess if this is kept to a minimum it’s better than no change.

@teapoweredrobot

Eg only stones or groups of stones in atari

It makes sense but maybe a bit too limited, but maybe that’s how it might need to be, I don’t know

[Conrad_Melville]

@teapoweredrobot

I would like to know how a restart is handled. Is it always the other person who plays first?

When a game restarts, the person who plays first is the one who would have played next had the game not been interrupted. No one ever gets two turns in a row.

This thread addresses a significant problem that is much more common than one might think. I know this because I watch a lot of games, especially among DDKs. It also has the potential to affect higher-level games since KataGo is stronger than anyone on this site and might notice a weakness that even dans have overlooked.

When both players overlook an unsettled position that could be decided by the next move, and the autoscore reveals the defect, then acting on that information is the same as botting.

I see only two possible solutions: either prevent return to play, or annul the game if an unsettled position has the potential to change the outcome. Perhaps add a grace period of a few minutes before the game is officially scored, during which someone could file a mod report if they wanted the game adjudicated. The filing of a report might then freeze the game until adjudication occurs.

[GreenAsJade]

… or as the OP said, dumb down the scoring tool, so that it really only helps with obvious areas (if that is possible).

[yebellz]

@Conrad_Melville

When a game restarts, the person who plays first is the one who would have played next had the game not been interrupted. No one ever gets two turns in a row.

I believe this is the case for the various rulesets that use area scoring (e.g., Chinese, AGA, New Zealand, etc.).

However, this should not be the case under Japanese rules, which explicitly states that when a person asks to resume the game, their opponent gets the next move.

This quirk of the Japanese rules can even lead to both players losing.

Odd Cases in the Japanese Rules

Both Players Lose The Japanese rules state: Article 13. Both players lose 1. After the game stops according to Article 9, if the players find an effective move, which would affect the result of the game, and therefore cannot agree to end the game, both players lose. Note that article 9 (about resuming the game) also applies: Article 9. End of the game … 3. If a player requests resumption of a stopped game, his opponent must oblige and has the right to play first. In the official comment…

@Conrad_Melville

annul the game if an unsettled position has the potential to change the outcome

Technically, the Japanese rules specifically prescribes that both players should lose. However, pragmatically, it seems reasonable to annul the game to avoid the ambiguous effect on the rating system.