Go Variant Server

Published 19 Aug 2020 by antti (last edited 19 Aug 2020)

As we know, recently the prevalence of go ai software has made it hard to organise tournaments online. If you organise a tournament with no surveillance, ai-users will inevitably pop up; and if you want to organise a prestigious and serious tournament, you have to do something like the 2020 Korea Prime Minister’s Cup, where all participants are required to have a video call showing their playing profile. Even this is not a foolproof solution, so additionally there is a team of anti-cheating staff checking the games for anomalies.

At the ngd, we used to organise small competitions with prizes, but in recent years we have stopped these because they incentivise cheating. As the main reason to participate in a league such as ours is learning, we have carefully tried to make sure that there is as little incentive to cheat as possible – although the flip side of this is that the students have to find their own motivation for studying.

More recently, I came up with a kind of solution that qualifies as go activity, re-enables small competitions, and also works as community-boosting activity. Enter seequ, one of our league participants, who has recently started creating a go variant server.

Go players love to create variants of the base game. There are pair go and rengo; there are atari go, one-colour go, toroid go, and kill-all go; and the list just goes on with strange board sizes and shapes and increasingly convoluted variants such as phantom go, tetris go, and ‘4+1’ (where, if you form exactly four stones in a row, you get an extra move). To get an idea of the number of possibilities, see this list of variants on Sensei’s Library. Most of these are closely enough related to go that stronger go players do better at them, while at the same time regular go ai are (at least for now) of little use. Thanks to this quality, as well as the fact that many variants are also considerably fun, I see go variant tournaments as a promising way for sparking interest in the game online.

The go variant server that seequ has started creating is for now at an early stage, but you can already play standard go, pair go, three-colour go, and pixel go. Note that, currently, any games played may disappear when the server is updated, and that there is no way to review played games (although undoing is possible, and you can explore freely by claiming both/all player seats).

Here is a showcase of some of the variants currently included.

Three-colour go

This is a fairly standard variant of regular go, played with three players. Although several interpretations exist, on seequ’s server the game is played by area scoring – in other words, you don’t get extra points by opportunistically capturing the opponents’ stones, and only the number of living stones matters in the end.

Multiplayer go is highly political in that it is easy for losing players to decide to ‘gang up’ on the leading player, quickly reversing the situation – in other words, thick positions are extremely valuable. For this reason, it might be fair to hide which player is playing which colour, as otherwise stronger players are more likely to get targeted (I have been on the receiving end of this more than once).

It turns out that, among other things, seki can be a really complex mechanic in multiplayer go. Consider the following diagram.

At first glance, this looks like a three-way seki (provided that the surrounding black and white groups live). But is this really the case?

Black might decide to sacrifice her stones, complicating the position...
...or White...
...or Blue.

In the above fashion, there are three possible follow-up scenarios, all of which depend heavily on the move order (we use Black→White→Blue).

Black suicides, and the result seems like it should probably be a seki between White and Blue.
White could technically consider capturing Blue instead, too, but that would not help the upper white group.
White suicides, but Blue captures Black instead. The result seems like it should probably again be a seki between White and Blue.
Blue could consider capturing White instead, too, but that would not help Blue’s own situation.
Blue suicides, but Black (probably) captures White instead. Now it seems like the black corner stones survive and the blue stones are fated to die.
Black could well consider capturing Blue instead, too.

In the above three diagrams, only the suicide by White benefits its player, so it should be the eventual outcome – meaning that, even though the original position looks like seki, it is nothing of the kind.

Pixel go

This is personally my favourite go variant, which Mikkgo came up with when we were studying go in the Experience Go in China 2009 program in Beijing (Jeff was there as a teacher). Mikkgo recalls having thought, ‘what if there was a way to play between lines in regular go’, which led to the idea of ‘increasing the resolution’ of the go board. Hence, in this variant the players play 2×2 clumps of stones, and you can try to fit your move partly on top of existing stones or over the sides. The 19×19 board in pixel go, in other words, roughly translates to a 9½×9½ board, which is just beautifully imperfect!

This variant can be played with either area or territory scoring, although area scoring makes for the simpler interpretation. Territory scoring creates an extra layer of depth in that you want to avoid filling your future territories any more than necessary – but not at the expense of leaving behind weaknesses.

This 1–2 is a standard pixel go opening, closely resembling a popular 9×9 opening where both players start on the 5-4 point.

Black invades!
A pixel go joseki?

It is possible to ‘stick’ to existing stones, meaning that the full 2×2 clump does not get played. For example black 3, the ‘shoulder-hit attachment’, is a very valid technique. 4–7 shows one way the situation might develop.

Three-colour pixel go

One of the best features of seequ’s go variant server is that it is possible to combine different variants – in other words, we can also have three-colour pixel go!

Good game?

Quickly playing around with this variant, I already found a mind-boggling scenario, which is no doubt just the first of a myriad interesting possibilities.

This is a bit of a tsumego! It appears that the a stones will probably live in a seki.
If Black tries to capture Blue, then White will be able to link up his stones; and if White tries the same, vice versa happens (unless Blue decides to interfere and gang up on Black). It is best for Black and White to leave the blue stones untouched – that is, unless all the black and white stones are alive and Blue is winning the game.

Talk about a complicated game! And with this, the variant server is only getting started.


This post has no comments so far. Please log in to comment!