reviewing SL articles using LZ and criticism

Knotwilg · Post by **Knotwilg** » Sun Dec 15, 2019 7:28 am

Hi John

First of all, I'm happy that, when I present the material more cautiously, you seem more appreciative of it. Even if you didn't necessarily shout at me personally, I think most of the heat in the debate comes from suggesting that "the pros are proven wrong by the bots", while it's more "our interpretation of pro wisdom is proven doubtful by the bots". You keep emphasizing that pro thinking is demonstrably closer to bot evaluation than our vigorous attacks on conventional wisdom imply and I for one will honor that corrective reminder. So, when using "conventional wisdom", I mean "the amalgamate of direct advice by professionals up to Go Seigen, advice distilled into books by ghost writers and teachings by Western amateurs of variable reliability". That's a mouthful

John Fairbairn wrote: The usual Japanese proverb doesn't even mention two stones - it just says you become weaker. But the Japanese proverb begins "joseki oboete", and if you look up oboeru in a good J-E dictionary such as the Green Goddess this is how it begins: "commit to memory; fix in one's mind; remember; know; get by heart; memorise." Nelson begins: "remember; memorise". So where does the sense of "learn" (= understand) come from?

Can we take the original Japanese proverb and turn it into a better English proverb so that we can get rid of the bad proverb? Any suggestions?

Bill Spight · Post by **Bill Spight** » Sun Dec 15, 2019 7:46 am

John Fairbairn wrote:The usual Japanese proverb doesn't even mention two stones - it just says you become weaker.

The Japanese Wikipedia article ( https://ja.wikipedia.org/wiki/定石 ) has the two stones version, as do other Japanese sites.

But the Japanese proverb begins "joseki oboete", and if you look up oboeru in a good J-E dictionary such as the Green Goddess this is how it begins: "commit to memory; fix in one's mind; remember; know; get by heart; memorise." Nelson begins: "remember; memorise". So where does the sense of "learn" (= understand) come from?

Well, yeah, memorize is the first meaning, but there are others, including learn. But I agree that memorize is what the proverb intends, certainly not understand.

Knotwilg wrote:Can we take the original Japanese proverb and turn it into a better English proverb so that we can get rid of the bad proverb? Any suggestions?

"Memorize joseki and get two stones weaker" or the like would be a better translation.

Kirby · Post by **Kirby** » Sun Dec 15, 2019 8:10 am

I agree that oboeru can mean “learn” in some contexts; learn can even mean memorize in english (e.g. learn something by heart).

As usual, context matters. And here, I think it’s obvious that the meaning leans more toward “memorize”. Why would knowing some reasonable sequences make you weaker? Memorizing a sequence and repeating it without thinking is obviously what’s dangerous.

And that’s not just for joseki - it applies to the new stuff we learn from AI, too.

Learn what pros think; learn what AI thinks; then think for yourself!

Knotwilg · Post by **Knotwilg** » Sun Dec 15, 2019 8:27 am

(Sorry for probably going too fast in publishing, but since Bill has already read the first chunk of analysis, I want to continue on a number of variations - should we make a new thread for the position itself?)

Click Here To Show Diagram Code: [go]$$W Playing actively $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . a . . . | $$ | . . X O O O . . . . . . . O . . b . . | $$ | . X . X . . X . . , X . . . . X . . . | $$ | . . . X O O O . . . . . . O . . . . . | $$ | . . . . X X . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

As mentioned on the SL page too, the real "sin" is probably to exchange A for B. If we omit the exchange, then LZ seems to like the idea of leaning and/or capping much more.

Click Here To Show Diagram Code: [go]$$W Turn and play elsewhere $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . a . . . | $$ | . . X O O O . . . . . . . O . . b . . | $$ | . X . X . . X . . , X . . . . X . . . | $$ | . . . X O O O . . . . . . O . . . . . | $$ | . . . . X X 1 . . . . . . . . X . . . | $$ | . . . . . . . . . . 2 . . . . . . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 3 . . | $$ | . . . O . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

This is LZ's first idea. The top remains pretty hot but first the lower left will be settled, or rather they will settle simultaneously.

Click Here To Show Diagram Code: [go]$$W Playing actively $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . 6 . . . . . . . a . . . | $$ | . . X O O O . . . . . . . O . . b . . | $$ | . X . X . . X . 2 4 X . . . . X . . . | $$ | . . . X O O O . . 3 . . . O . . . . . | $$ | . . . . X X . . . . 1 . . . . X . . . | $$ | . . . . . . . . . . . . . . 5 . . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

The second idea is to cap immediately.

is a leaning maneuver more typical of bots (a shoulder hit) fixing a weakness in the surrounding stones, which apparently urges Black to live with

Click Here To Show Diagram Code: [go]$$W Playing actively $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . a . . . | $$ | . . X O O O . . . . . . . O . . b . . | $$ | . X . X . . X . . , X . . . . X . . . | $$ | . . . X O O O . . . . . . O . . . . . | $$ | . . . . X X . . . . 3 . . . . X . . . | $$ | . . . . . . . . . . . . . . 1 2 . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

This inspires LZ (if I may say so) to explore the shoulder hit as a starting move, with a similar result.

%%%%

These three moves get the following metrics
- the turn: 18k 55,3%
- the cap: 6k 55,2%
- the shoulder hit: 3k 55,6%

These are the main moves LZ explores on her own account. The slide starting the "catenaccio joseki" is discarded after 250 playouts, giving 53%

Bill Spight · Post by **Bill Spight** » Sun Dec 15, 2019 8:34 am

Kirby wrote:I agree that oboeru can mean “learn” in some contexts; learn can even mean memorize in english (e.g. learn something by heart).

As usual, context matters. And here, I think it’s obvious that the meaning leans more toward “memorize”. Why would knowing some reasonable sequences make you weaker? Memorizing a sequence and repeating it without thinking is obviously what’s dangerous.

And that’s not just for joseki - it applies to the new stuff we learn from AI, too.

Learn what pros think; learn what AI thinks; then think for yourself!

I grew up avoiding rote memorization, if possible. However, humans have a long history of memorization, going back before civilization and certainly before writing.

It has its place, I think. Not in memorizing joseki, because we don't know if joseki are correct, and certainly the rest of the board matters. But I think that memorizing life and death, including all variations and relevant stones, can be useful.

Kirby · Post by **Kirby** » Sun Dec 15, 2019 8:48 am

Probably, memorization isn’t completely distinct from the thinking process. It reminds me a little bit of that “chunking” concept.

xela · Post by **xela** » Mon Dec 16, 2019 12:51 am

Knotwilg wrote: Of course we can say, what works for Fujisawa works for me, or we shouldn't care about the crazy things LZ says, we can still learn from conventional wisdom, carefully developed knowledge by professionals, crystallized into clear guidelines. And if we play by yesterday's pros standards, we are still at least 6d today.

First of all Knotwilg, many thanks for the interesting example and great analysis!

I find myself wondering if LZ is identifying the next level of trick play here. I can sometimes find moves that I know are wrong but are more effective than the "proper" moves against a 10k opponent. And high dan players have no trouble doing the same to me in handicap games. I guess the top pros are still fine-tuning their judgement based on feedback of the form "this worked against a human" without any way of knowing for sure that they've found the "best" move. (And I'm not saying that LZ knows best either, just that it's giving us a different perspective.)

Gomoto · Post by **Gomoto** » Mon Dec 16, 2019 6:48 am

I always have a bad feeling when I play a mediocre move and I know already my opponent will answer in a suboptimal way. But perhaps this is just part of go.

Life In 19x19

reviewing SL articles using LZ and criticism

Re: reviewing SL articles using LZ and criticism

Re: reviewing SL articles using LZ and criticism

Re: reviewing SL articles using LZ and criticism

Re: reviewing SL articles using LZ and criticism

Re: reviewing SL articles using LZ and criticism

Re: reviewing SL articles using LZ and criticism

Re: reviewing SL articles using LZ and criticism

Re: reviewing SL articles using LZ and criticism