Hacker News new | past | comments | ask | show | jobs | submit login
Two-Headed Go (jefftk.com)
71 points by luu on Dec 18, 2019 | hide | past | favorite | 18 comments

Interesting! I would have guessed going 2:1 against a stronger player wouldn't help much. Maybe I'm digging in, but I'd expect that it stops being true soon -- I'd be surprised if a 1-kyu didn't beat two collaborating 6-kyus most of the time.

(I used to play a lot but don't anymore, was last at 1-dan AGA.)

I'm 1 kyu. I don't have any problem winning against two 6 kyus (I saw some experiments like that in the past.) There are simply too many things they overlook. And two 1 kyus (or a room of them) won't win against a 4 dan. We're playing different games.

What happens is that the weaker players become a little better at don't making reading mistakes and at finding possible moves. Unfortunately they don't get any better at evaluating moves and the unknown unknowns are still unknown. The better player finds moves that none of the weaker ones would think about or be able to explain.

I can imagine that the knowledge about the game of two 11 kyus have a smaller intersection than the ones of 6 kyus. This means they become much better together.

However two 1 kyu have about the same knowledge, so they could be at most 1 dan together, probably still 1 kyu.

1k as well here.

I don't see it as much about knowledge of the game, more about avoiding mistakes and less pressure that lets you look at a game more in-depth. That, and two people will probably coalesce towards something like honte anyway.

Add to this the player opposing is under pressure because he HEARS the discussion. Can he take advantage of knowing plans? Should he?

I'd say this is a nice learning tool, and honestly, anything that keeps you interested and makes learning and playing a bit more fun is fantastic. I've played 5-in-a-row go, pair go, 3-player go, 1-colour go, 3D go and tried different board sizes like 38x38 or 19x19 infinite (sides 'joined'). I've never run across this variant! Seems fun!

Making guesses about what a 2-player team's strength is is a losing proposition, though.

> Add to this the player opposing is under pressure because he HEARS the discussion. Can he take advantage of knowing plans? Should he?

Since we knew the opposing player was listening we mostly didn't discuss plans, at least not in a way that would have hurt the plans. Mostly we suggested moves to each other, and pointed out problems with the other person's suggestions.

This is super interesting!

I think 1kyu/1dan level is taken to mean that you've ironed out any major dejects in your game – as in, from here on out you're adding game lines:

A 1dan playing against two 5kyus will win if they combined still have overlapping defects, depends on the overlap. It'll be closer than one 1dan versus one 5kyu but I'd still expect the 1dan to win.

I would have expected a 6kyu to defeat combined 11kyus, so this is surprising. Would a 6dan defeat combined 1dans?

We should set up two-headed games with varying strength differences and find out, would be super interesting to see what the win outcome rates are.

This is roughly like starting with 50 extra points. A good lead, but you can still mess it up.

Plus, the 11kyus might be underestimating their rank. You can get better pretty quick, and might not realize it.

> the 11kyus might be underestimating their rank. You can get better pretty quick, and might not realize it

We've both been around our current level for years. My dad plays somewhat regularly with a small go club, while I mostly play my dad a few times a year.

I do think it's likely that if I either studied some theory or played a bunch of games against people slightly better than me that I would get better quickly. I'm a relatively non-central example of someone with my approximate level: I've been playing for about 20y, but almost all of my games have been against my dad.

Reminds me of Gary Kasparov vs The World, which surprisingly turned out to be a better game of chess than you might have guessed.


It sounds (from the comment on that page) like the biggest benefit they got was from being able to consult with each other to align on strategy and avoid blunders, which apparently isn't standard in Pair Go. Without that, I wouldn't really expect them to see qutie such significant benefits.

https://en.wikipedia.org/wiki/Go_variants#Rengo https://senseis.xmp.net/?PairGo

Unlike the OP game, pair Go is actually a handicap for the paired up side rather than an advantage. You only get one brain per move, and that brain might be forced into using a strategy it doesn't like by previous moves.

Common wisdom is that in a pair go game everybody has three opponents :-)

Remember that it's not allowed any talking, except asking who's to move, maybe who's to take a ko, and proposing to resign to the mate.

Agreed. The biggest issue for players probably below ~5k tends to be making big blunders, and so talking through moves with a partner would make a large difference at that level but wouldn't be replicable in a normal pair go game.

This is a pretty common format for playing go. Usually called pair go. There's even fairly large tournaments for it here and there.

(To be clear I'm only talking about the two players on a team part, it's usually 2:2, not 2:1)

Personally I hate it and am absolutely awful at it, but to each their own.

Normal pair go explicitly disallows nontrivial communication of any sort. Very different.

Pair baduk can be pretty fun. In my local club we do 2:1 and 2:2 pretty regularly. There is strong diminishing returns above two players though. 3:1 and 3:2 and 3:3 are all usually too much consultation for the game play to be fun if you don't have an opinionated leader on each side.

Just a picture? Who won? Any game record?

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact