Hacker News Reader

Comment by benchaney | Hacker News Reader

I disagree about memorization. That may be true for chess, but the go board is too large and the game to long for that to be viable. In the opening, bots will generally recommend several places per move that lose 0 or -0.1 points compared to its preferred move. You would have to memorize thousands of opening just to have a chance of still being in book for the entire opening, and even then your opponent can take your out of your preparations at a cost of less than 1 point.

View on HN

Go already required memorizing 'joseki' or standard patterns that crop up again and again (particularly in the opening). The usual mantra is 'don't learn joseki, learn _from_ joseki'. But the truth is - like openings in chess - if you don't know them and you play someone who does, you're often at a disadvantage.

To add to your point, Fischer was complaining about chess openings going stale 20 years before computers came on the scene

After he fell out of the pro scene he became way more interested in chess variants than vanilla chess (more specifically his Fischer Random Chess). I can't imagine being at the level he was at, where you needed to hundreds or thousands of hours memorizing and practicing openings... so not surprised he appreciated a game that valued more on-the-fly decision making than heavy preparation.

You don't actually spend 100,000s of hours memorizing openings. You study the basic opening principles and key variants, but for the most part it comes very naturally. You play 100,000s of hours of chess, and EVERY chess game has an opening.

> You play 100,000s of hours of chess

100k hours is 11.4 years straight.

Just sayin’...

They said "hundreds or thousands" not "hundreds of thousands"

I can't find many good resources, but at least the joseki explorer on OGS has been listing some pretty heavy favoring for particular moves on a given joseki. Couldn't we see specific joseki lines get more entrenched and go further?

AI have certainly been changing which joseki are popular and which aren't, but this isn't the same as increasing the amount of memorization in the game since joseki aren't new.

The AI favored lines are likely to get heavily entrenched. I wouldn't expect them to go further because they generally end when becomes better to play elsewhere. There are a few fighting joseki where this isn't the case, but AI doesn't seem to like them much, although there are a few exceptions.

> There are a few fighting joseki where this isn't the case, but AI doesn't seem to like them much

Do we know if those moves are actually weaker overall, or is there some other sort of selection bias at play?

I think some of the AI are trained by forcing them to play random moves every so often. That might bias them towards larger, more flexible openings where there are large regions of the board that make sense to play in rather than specific move sequences.

But if I were guessing, the AI probably doesn't like fighting joseki because it knows how the fight will end. AIs are much, much better than humans at working out who won a fight.

Reply on news.ycombinator.com