Let’s keep this thread narrowly tailored to the move threshold issue. I’m only interested in whether we should increase the value from two, and whether the new value would be four or six.
2 Likes
The difference is that here a single player repeatedly abandons games too early and without reason, which makes them a sandbagger by their own actions.
In the case that @mark5000 mentions, the new player is inadvertently airbagging since their opponents resign. The problem of enforcing this by asking the opponents to keep playing is that the resigning opponent generally is not a repeat offender. Moreover it has a huge impact on the rank of the new player. It’s different people each time they play against, and many of those different people that resign don’t have any precedence of doing that, since an average 5k doesn’t usually encounter complete beginners when they match with some other “5k”.
Like I said, neither player wants to play the game, I think it should be possible to annul those games anyways just to improve the playing experience for both players.
1 Like
I see that one of the opponents is developing a “disconnection” habit when behind. Maybe it just happened (I hope this is not considered “naming and shaming”, if it is I will of course remove it).
The other person seems to resign a lot in the beginning, even when moves are not as weird as the linked game. Either it’s their personal thing or they don’t know lost games with two moves are annulled. I won’t speculate because I don’t know.
I think this game was a match made in weirdness. Both have some weird pattern behaviors (to my untrained eyes), I guess who got 2 with 1 shot? 
1 Like
I don’t oppose the suggestion that games with less than 6 moves should be annulled automatically, as games like that can hardly be used for establishing a ranking (although maybe not on 9x9, idk). But I agree that there could be unwanted consequences, like others have mentioned, that people resign when they dislike the opening of their opponent.
So I would suggest to put countermeasures in place. Some other form of penalty. Maybe if a player resigns / timeouts many games before move X (depending on board size), they temporarily can’t start a new game.
2 Likes
I think neither four nor six is going to be much help with these kind of problems. I’ve encountered many of these cases where the strong player resigns only after the new player completely messes up the first encounter.
It will help a bit, but the problem will persist.
1 Like
you talk about new accounts, then do it to [?] accounts only, then it will be possible to allow cancel even after 10 moves.
I think if opponent resigned, but Kata thinks you lost, you should have “annul” button, without waiting for mod.
This seems like a pattern. But on my floor, I don’t know if it’s that unusual (I will have you know I won (technically) 2 games in a row last week, I won the second and lost the first in a winning position because I timed out (lost in chat), and I was about to resign to both of them, thinking I was far behind. I mean, I’m stupid like this, but statistically I can’t be the only one).
I believe the system should absolutely flag patterns, but annulling should be more finetuned, or remain in human hands.
And in mass annulments the player should be notified, because the system is not foolproof. There are people who don’t check their rank and/ or their game history, and all their opponents will form ideas by seeing a bunch on annulled games in their profile, no? At least they should have a chance to appeal.
1 Like
Am I supposed to only play moves that my opponent likes? This is getting a little silly. If it’s cheese, then it should be easy to beat it. If you play a move, you’ve accepted the game. Any number of moves above this is arbitrary.
6 Likes
I firmly believe the strong player is in the wrong here. If they don’t think their opponent is worth their time, they should take the loss. Or do detective work on each opponent, if they only care to cross their hands with worthy opponents. It doesn’t matter the reason, if you don’t stay to play the game, you have consequences.
If you remove the “but the one player is so strong wow” then this isn’t an issue. And such measures as discussed here should be about the players as a whole not looking fondly only to some of them.
3 Likes
This isn’t the problem here, nor do I think the strong player is in the right; the problem is that winning one of your first games has a huge effect on the rank of the new player: their [?] makes it so that after two wins they are practically assigned dan level.
This gives a lot of trouble down the road, with opponents thinking the new player is trolling, even more games being resigned, etc. All the while the rank of the new player does not get immediately adjusted, since that’s just what rank does.
It’s not that the opponents are doing good work by resigning, but it’s impossible to enforce this, because generally the resigning player only does it once (and the bad effects are immediately in effect). Banning the resigning player is unreasonable and doesn’t help, warning them is slightly more reasonable but also doesn’t help.
The problem is: new players get ranked up by opponents resigning one of their first games, which messes up the estimated rank of the new player and incorrectly matches them with sdk’s.
How can we prevent these new players from accidentally ranking up?
Either we prevent opponents from resigning the games. This requires changing the mentality of a huge part of our player base (apparently, since this is a common problem), and needs to send this message to people before they encounter a new player and resign. How do we do that?
Or we prevent these resigned games from being counted towards rank. Seems a lot easier to me.
1 Like
tl;dr
I like 4 moves. You can play one move by accident/ you clicked on the wrong challenge/ you forgot to check the time etc. Once you play a second move, you commit. After 4 stones, it’s a legitimate game and whatever happens, happens.
That’s my 2 lollipops on the subject.
1 Like
Hey, glad you’re still here! And thanks for your input.
2 Likes
From my conversations, I’ve had the impression that many people think 3–6 moves is the optimal range.
I remember someone saying that Fox uses five, which is what Mark suggests. I favour three – etc. etc.
The ranking threshold certainly must be less than ten moves.
3 Likes
HUGELY OT, I JUST DON’T WANT TO LEAVE THE QUESTION HANGING, BUT I’LL RESPECT MARK’S WISHES AND NOT RESPOND ANY MORE ABOUT THIS
You can’t. Let the system do its thing, it adjusts (remember when I was a 13k for a week because of chance? fun times). You are trying to micromanage it. Ranks go up, ranks go down, ranking system gets a freshen up now and then.
If enough SDKs are not too proud to play with someone beneath them and actually stay on the board and beat the newbie, then the newbie will get their true rank soon enough. If it takes 2 games to reach Dan, does it take much more to fall down again? Probably not. If SDKs don’t want to “change their mentality” then they will be stuck with random opponents that nobody bothered to actually play so they will get their actual rank.
Or create a pool of players accessible to [?] for their first 5 games. Or make the first game obligatory against a bot. You’ll say all these are too restricting for the nature of this server. Well, the same goes for wanting to prevent absolutely any inaccuracy to a program when the thing involves 2 people.
As I said, players who really, really, feel strongly about it, let them do the detective work. I doubt an SDK will see 1 game of their opponent and not know if they want to play them or not.
If I resign from a game I was winning ACCORDING TO KATAGO, I lost ACCORDING TO ME. I was not good enough to read the board correctly and tell I was winning, therefore I WASN’T GOOD ENOUGH TO WIN THE GAME. Enough with the rank accuracy and that algorithm, let people do what they want with their games.
I’m not here to make the algorithm look good, I’m hear to play.
I’m tired of being asked to be what an effing AI wants me to be, I’m the one playing.
Sometimes I wish we could do away with KataGo and its assessments and perfect plays, honestly, it’s exhausting.
6 Likes
who invented the 19, 13, 9 rule initially, pretty genius I have to say. 
1 Like
I’d be fine with 3 or 4 moves as the next minimum step. 3 if game creator is white and 4 if they are black. I’d want to monitor complaints that people are just resigning unusual openings. And if it could be limited to ? accounts then so much the better. But fundamentally I’m with @Gia really.
The solution seems to be to at least show new accounts as 20k or something so that initial matching is with others around that rank. I think the problem is more nuanced isn’t it? I.e. new players getting ranked up by SDK opponents resigning. I guess a TPK opponent resigning against a ? player doesn’t shoot them up the rankings.
A better solution might be to go back to letting new accounts specify a strength. Or asking how many games they’ve already played or what accounts they have on other servers/IRL ratings and making an (approximate) ranking from that type of information.
1 Like
Therefore such short games should definitely not be counted towards ranking, right?
How about a cancellation cutoff at some low number, like 2 moves. Separately from this, let’s have a ranked cutoff at a higher number, like 20.
If you abandon the game
- before the cancellation cutoff, the button is labelled “Cancel” and the result will be “B/W+ Cancellation, ranked, annulled”;
- before the ranked cutoff, the button is labelled “Resign” and the result will be “B/W+ Resignation, ranked, annulled”;
- afterwards, it will be “B/W+ Resignation, ranked” as normal.
2 Likes
The trouble with a high number is that players see they are at a disadvantage and they cancel—thereby avoiding a ranked loss. This is what happened with the old 19-move standard in a 19x19 game. As I recall, people complained, and the standard was changed. There is probably an old thread about it.
2 Likes
Yeah I agree, it’s a terrible idea to let people figure out they’ve played an opening badly and be able to just escape out of the game without a loss especially like 20 moves in.
I think the best thing from lightly reading this discussion would be that if one is thinking to use move number or katago to decide on whether a game should be adjusted (annulled etc) would be that instead it just auto flags the game rather than making a judgement.
Edit:
Ok one might worry then that it’ll make too much work for mods, if it flags hundreds of games, but then one should design the auto-flagging system very precisely.
For example one should flag games that say katago thinks neither player has a significant lead (less than X points X~5-10 points, whatever you imagine some opening blunder to be) and the game is less than Y moves (Y~20) or something like this. Then the mod can decide if it’s worth annulling the game.
Edit2:
Maybe even multiple ways it would autoflag, the previous case to catch people who just resign because they don’t like the opening or maybe had connection issues etc. One could also have another flag that the winning player resigns while being X points ahead (X~10+) with less than Y moves played (Y~20 moves) to catch players who resign because they think there opponent is a new player etc.
1 Like
I think the cut-off limit of 2 moves makes the most sense, since after both players have placed a stone it usually means that both players are willing to play and finish the game. The old “n moves on nxn board” had the problem on how it didnt actually scale evenly on smaller boards, 18 moves on 19x19 is only about 5% of the board, while 8 moves on 9x9 is almost 10% of the entire board. You can definetly screw up your winning chances with just the first 4 stones you place on 9x9 board >___>
I think thats the problem for having any fixed limit, 6 moves on 9x9 is lot more than 6 moves on 19x19 board.
Also one thing which i - as an mostly correpondence player - really like about the current 2-move limit is that it prevents people escaping early.
Back when we had the old system in place, i always felt the pressure on playing the first n moves very quickly when the game started. I had many occasions where i created an open challenge, someone took it, we played few (<n) moves, and then my opponent just stopped playing. It was “free” for them to let their time run out, so they really often did just that.
It still happens quite often that someone accepts an open challenge or clicks automatched correspondence, and just cancels it immediately, but i dont find that as annoying as playing almost 10 stones and then leaving for the game be annulled.
I guess 6 moves would be better treshold than 19 moves, but i’m afraid it would still give some people the option to start games, put stones on empty corners and then abandoning it for free.
3 Likes