After the game ends in which there were so many players that you never even got a chance to play a move, what result would be displayed in your profile?
Sounds like an edge case but there are many open rengo challenges aiming for over 500 participants. There’s even one going for 2k participants on 9x9.
1 Like
It would be displayed that you participated in a Rengo game where the team won.
I see. It’s just strange because you never even interacted with the game and yet it’s on your profile. Maybe that’s the best way to handle it anyway.
We put “canceled” games from non-rengo matches on your profile too. I suppose one could make the argument that one’s presence in the game is interaction 
I can rationalise it in all sorts of ways (for example “you interacted by committing to play if your turn came up”) and I can see all sorts of problems if it were otherwise. How would it even show? Under what conditions? etc.
But of course the real reason that it is this way is because it’s the simplest thing.
“Keep it simple” and all that.
4 Likes
I did not manage to remember it, but I am sure enough about general setup.
The problem: given the set $a_i$, choose signs +/- for each variable to minimize $|\sum(sign_i * a_i)|$.
The heuristic:
- Use divide-and-conquer: first, split the set A into two subsets, A_1 and A_2 and choose signs to minimize value of |sum(A1)| and |sum(A2)|. Now we have two options to consider: A1 + A2 and A1 - A2. Choose the better one.
- ???
- Combine these two steps to get efficient algorithm.
I think what would work well and easy to implement would be the following variation:
- Sort $a_i$ by absolute value: |a_i| <= |a_i+1|
- Divide A into 2 parts, A_1 and A_2.
- Choose signs in A_2 to minimize sum(A_2) (recursively).
- Substitute whole set A_2 with sum(A_2), i.e. solve the problem for A_1 and new element x=sum(A_2).
- To get signs of individual elements from A_2, we would need to take sign of x into account.
- When set is small enough, we may use bruteforce algorithm to find out best signs (i.e. if A has <= 3 elements we need to handle it not by this recursive routine.
So it is almost n log n algo, should be fast.
I don’t know if this is intentional, but I noticed that only the active player (and not the one who misclicked) can ask an undo. Especially in correspondence games it might be difficult for the other player to notice it before that the opponent plays, and players don’t always read/understand the game’s chat.
I think the undo should be asked from the player who played the wrong move, not from the following player.
3 Likes
Another game in which the wrong team’s time is running out. This seems even more strange to me because no move has been played. Maybe it is related to what happens during the blue timer?
3 Likes
I tried to reproduce it here.
After the first player lost by timeout, time continued to run normally. When the player of the other team resigned, the wrong timer started to run. Playing normal moves solved it (which also happened in the game I reported earlier, white played).
If it only happens in the first few moves it’s almost useless to fix it, I’ll try again now.
Edit: With the same sequence, it also happens later. In this game, move 8. Rengos are always unraked, but some people might be unhappy with a wrong result they couldn’t change.
2 Likes
PopcornCzar resigned, another player timed out and it’s PopcornCzar’s turn again
(It was already reported, it’s just another example)
Do you think we can reproduce this “on demand” some time when I’m looking at it, or instructions for me to force it to happen?
I suppose resignation+timeout (in this order) in the same turn by players of the same team makes it happen. Now I’ll try.
No 
Here Sofiam2 couldn’t play after that it resigned and Sofiam timed out (I don’t have much imagination with nicknames)
I tried also with 4 players and less than 6 moves (here) but the game ended.
I can’t find anything in common between those two games I didn’t consider, apart from the fact that they were by correspondence and the ones I tried weren’t.
I realised where I was wrong, the player who resigns must not be the active one. I managed to do it here.
The steps to reproduce it (with 1= active player and 2= bench player) are:
2 resigns
1 times out
2 comes back in the game
This is at the top of my list to look at. Though I don’t have an immediate opportunity, I will try to soon.
1 Like
Auto handicap, why did we get 9 stones?
Averages are 15.2k against 7.6k.
(also, ranks below 25k are displayed correctly before move 1, I saw a 27k that was shown as 25k after it).
Supporters can’t start the AI review in some games they played, but not all of them. The error they see is “To begin an AI review, you must be a site supporter”.
Example:
(Reported privately, twice, by 哲學PhiloSoulMatePMme. Moderators can fix it)
1 Like
I think that was something that was on the to-do list, or well the “something we’d like list”. It could be a something like only the first two players to play can run a review if they’re supporters or something.
1 Like
i do not know what you are talking about
be able to re-balance teams (before move 1?)
it’s ok if the current player cannot change team
or add an extra step before auto-start to be able to balance.
The “big” problem that is not resolved is when players time out on their first move. You can end up being left alone against 5 before any move has been played.
maybe, to be able to change teams until move N, N is the total amount of players
Thanks - I can reproduce this problem now, I’m digging into it.
Edit: fix submitted.
Note: this ought to fix “wrong person’s clock running”.
I’m not sure about the other symptom (“person resigned but they are still playing”).
The bug I fixded might be able to cause that, and therefore now be fixed, but I never saw “wrong person playing” myself, so I can’t be sure.
2 Likes
I think this one pauses on weekends, but hasn’t paused. Others have.