Mmm - since rengo games aren’t ranked anyhow, I have a feeling that this is one that might just need to get out into the main server and see how it goes…
I started an auto handicap 3 people rengo, can you (@benjito and @jlt) join, please?
Me + benjito against jlt should get 2 stones, I think
Edit: Seems to work well, benjito and I got two handicap stones.
Thanks for joining it 
2 Likes
and there is no time to choose who is in which team?
Is there auto-balance system that is not handicap?
2 Likes
When bots are online on beta? I’d like to get a rank to test too!
Does it use humble rank for [?] ? 
1 Like
Obviously, right? That’s what “it starts when the last player joins” means.
Of course, you can adjust up to the last person
My feeling is that if you care about balance, it’s probably a small game and you don’t need auto start.
If you have a big game, the random join order is probably roughly balanced.
This is not to say that some sort of auto balance is not possible. I’m just saying that you have a good set of tools now already
auto start is bad idea for those who care about balance. But, they can try auto start anyway, and then dislike the result.
If there will be system that chooses team for everyone, so average rank is roughly same, then “dislike the result” part will not happen.
So auto-balance is important when there is auto-start. With it there will be less sad people than without it.
1 Like
If i am not wrong, @stone.defender means some auto pairing, like separating the extreme levels between the teams and then permuting to reduce handicap (if necessary), something like that
3d 18k 12k 1k
extreme 3d 1k / 12k 18k
dispatch 3d 12k / 1k 18k
permuting 3d 18k / 1k 12k
1 Like
From experience, the team’s “average” in Rengo is not the arithmetic mean though, as the weaker player definitely weighs stronger.
And this makes sense if we consider that the stronger player already knows that he is locked out of applying complicated ideas because they might backfire greatly if they rely on every move in a sequence being the correct one.
Rengo after all requires the stronger player to adapt to his weaker team mate to form harmonic teamwork.
However, it seems hard to determine the weight distribution, even just what kind of distribution it should be. I have no concrete numbers for you 
.
2 Likes
Maybe auto-balance is not the most urgent thing to implement but the simplest way would be to order players by strength P1,…, Pn.
Then reorder like this: P1, Pn, P2 P(n-1), P3, P(n-2),…
Denote by Q1, Q2, Q3,… the reordered sequence.
Write n=4k+j for j=0,1,2 or 3.
First, put Q(4i+1), Q(4i+2) for i=1,…, k-1 in team 1 and Q(4i+3),Q(4i+4) for i=1,…, k-1 in team 2.
If j>0, add Q(4k+1) in team 1.
If j=2 then add Q(4k+2) in team 2.
If j=3 then add Q(4k+2) in team 1 and Q(4k+3) in team 2.
The result won’t be perfect but will avoid severe imbalances.
2 Likes
Should come together with autostart or it will result in a quantity of uninteresting games.
4 Likes
I’ve played handicap rengo also and it’s worked more or less fine.
One could use an arithmetic mean and points system like
Or previously we’d used an arithmetic mean of ranks and pandanet handicaps with an extra six points of reverse komi per half rank beyond 9 stones.
http://pandanet-igs.com/communities/pandanet/60#handi
I think it’s actually much more subtle than just
There’s also the fact that the players can have wildly different styles and that can completely throw of their combine game plan.
So there’s much more to consider than just weak+strong vs weak +strong. There’s also weak+weak vs strong + strong.
I can imagine with more than two players per team, the weighting on any individual players rank probably won’t matter that much. Your instinct might say that the lowest ranked player might make the biggest
mistake, maybe that could be true on average, but since you literally can’t make a plan two moves ahead it’s quite likely anyone faced with a difficult position will make a sizeable mistake.
I think the point is, what is it practically? Is it half a stone lower than the arithmetic mean, is it the min of the two ranks? I don’t think it’s anything like the second one. We played a rengo where it was 2v2 with two stronger players giving handicap to two weaker players (at least 10 ranks difference), but one of them couldn’t play on at one point and a stronger player than us took over, and you could feel the impact on the game, especially the endgame.
If you want something implemented I think the mean is a good starting point, at least until there’s a clearer starting point to calculate such a correct weighting.
Except people might want imbalances in some cases. Some people, friends for example might want to be on the same team despite the fact it wouldn’t be the most balanced. I think if we always balance the team it might not even be a handicap more than 0.5 Komi we might not even need an auto handicap function half the time
1 Like
A simple arithmetic determination of the handicap has the advantage to be understood and then recognized (checked too) by players. No need to send them to a wiki page somewhere with complex grids or mathematics.
The automatic handicap can be proposed as a default value for not casual games, and the creator then can modify it.
2 Likes
In the mean time there is auto-handicap …
1 Like
True. auto balance still:
-can reduce the handicap needed.
-can push the difference of level in a same team
Which is what i expect in a automatized casual.
1 Like
When you folk work out what “auto balance” should do, then I reckon well look at putting it in
4 Likes
Just written the post before, for casual games with autostart autohandicap. (See my post 531 for more)
For simple rengo, everything can be just by default, like an help with different options (like weaks vs strongs mentionned before). Even to have no autobalance is fine to me in simpke rengo.