Life In 19x19
http://www.lifein19x19.com/

Man vs Machine article in wired
http://www.lifein19x19.com/viewtopic.php?f=18&t=10286		 Page 1 of 1
Author:  yoyoma [ Mon May 12, 2014 12:56 pm ]
Post subject:  Man vs Machine article in wired
http://www.wired.com/2014/05/the-world-of-computer-go

A nice article, centered around the recent Densei-sen event in Japan. Much more detailed that most of the similar ones I've seen in mainstream media.
Author:  Marcus [ Mon May 12, 2014 1:03 pm ]
Post subject:  Re: Man vs Machine article in wired
I caught this article through Hacker News (and noticed hyperpape commenting there, which reminded me that I still have an account here).

They did a pretty good job with this article.
Author:  hyperpape [ Mon May 12, 2014 2:11 pm ]
Post subject:  Re: Man vs Machine article in wired
A related question: how would you explain how big a four stone handicap is to a layman? When the New Yorker article was posted, there was a lot of confusion about how big a four stone handicap really is by people who don't know a lot about Go. As players, we have an intuitive understanding of four stones against a pro, and that it's both pretty good, but still a big difference. Yet I don't know how to explain that to outsiders, and it seemed like people ended up thinking it was "massive" (the term this article used) or quite a small difference.
Author:  RBerenguel [ Mon May 12, 2014 2:20 pm ]
Post subject:  Re: Man vs Machine article in wired
Michael Jordan playing with legs tied together.
Author:  EdLee [ Mon May 12, 2014 3:50 pm ]
Post subject: 
Hyperpape, analogies can help, depending on the audience.
If they know something about golf or chess, for instance,
you could use some handicap analogies there.
Otherwise, look for other shared backgrounds or experiences.
(The Wired article did exactly this: "the Eastern version of chess,"
"grandmaster" Norimoto Yoda, etc.)

For example, we can look at a photo of a Go Congress,
(say, the US Open), and, "See this entire section of people
at this table? Some of them have played for over 10-20 years,
and they are at 7 stones from pro. Over here, this group also has
people playing 10-20 years, and they have no chance
even with 9 stones against a pro. See this little kid here?
She needs only 2 stones from a pro. See this gentleman?
For the past 20 years, he has been 3 stones from pro.
(4 stones from pro at age 8 is vastly different than at age 48.)
So there is a huge range of level differences. :) "

Do we have any long-distance runners here?
A quick search on marathon times returns some
average time around 4.5 hours for men and around 5 hours for women:

Marathon times

If the record marathon times are around 2.5 hours,
how do people feel about a 4 stone handi vs. a pro
likened to a 2-hour head start in a marathon? :)

Similarly, we can probably use some swimming analogies.
In an 800-meter freestyle, look for some "average" time,
look at the world records, and find some head-start time
to give an idea of the 4-stone distance (pun intended :) ).

( I prefer something "slow and painful" like a marathon to
a 100-meter dash -- it seems easier to convey the long struggle
with a layperson, perhaps ? :) )

Of course, we always have the tortoise and the hare.
Author:  Tim C Koppang [ Mon May 12, 2014 4:30 pm ]
Post subject:  Re: Man vs Machine article in wired
I'm a distance runner, and your marathon analogy certainly resonated with me!
Author:  SmoothOper [ Mon May 12, 2014 6:28 pm ]
Post subject:  Re: Man vs Machine article in wired
It's like giving your opponent in chess four moves, except Go is a way better game because a chess game would be over.
Author:  John Fairbairn [ Tue May 13, 2014 1:29 am ]
Post subject:  Re: Man vs Machine article in wired
Quote:
how would you explain how big a four stone handicap is to a layman? When the New Yorker article was posted, there was a lot of confusion about how big a four stone handicap really is by people who don't know a lot about Go.


Although I personally don't like it, there is the popular line that if a top pro played against a go he would need four stones to win.

Quote:
As players, we have an intuitive understanding of four stones against a pro, and that it's both pretty good, but still a big difference.


Although I basically agree, I'd like to toss in the suggestion that amateur go players underestimate how big four stones is. I think the cause of the illusion is that we are generally used to a less rarified atmosphere in which moving up four stones with a bit of hard study is quite easy. On the slopes of Everest, however, moving up to the very top takes a lot, lot more than hard graft.

To use an analogy of my own, I have an impression that many amateurs who take four stones from a pro seem themselves as at the level of a Formula 3 driver against an F1 driver. I see them rather as closer to a good go-kart driver.
Author:  HermanHiddema [ Tue May 13, 2014 2:51 am ]
Post subject:  Re:
EdLee wrote:
If the record marathon times are around 2.5 hours,
how do people feel about a 4 stone handi vs. a pro
likened to a 2-hour head start in a marathon? :)


Well, I don't think anyone who can play a pro on four stones can be considered "average" :)

In go terms, "average" is SDK. Dan players are above average, DDK below average.

Here's a distribution of marathon finishing times: http://www.marathonguide.com/features/a ... shingTimes

From that: if you can run a marathon in under four hours, you are definitely an above average amateur runner.

So lets say that every stone on handicap is the equivalent of 15 minutes head start in a marathon. That means:

If you can beat a pro at 8 stones (2 hour head start), you are above average.

If you take 10-12 stones (2.5 to 3 hours head start), that's around average.

The best computers can take 4 stones (1 hour head start).

This also contains the component that John mentions, in that not all four stone handicaps are equal. It is easier to get your marathon time from 6 hours to 5 hours, than to get it from 4 hours to 3 hours (and much much harder to then get it anywhere near 2 hours).
Author:  Bill Spight [ Tue May 13, 2014 7:54 am ]
Post subject:  Re: Man vs Machine article in wired
hyperpape wrote:
A related question: how would you explain how big a four stone handicap is to a layman? When the New Yorker article was posted, there was a lot of confusion about how big a four stone handicap really is by people who don't know a lot about Go. As players, we have an intuitive understanding of four stones against a pro, and that it's both pretty good, but still a big difference. Yet I don't know how to explain that to outsiders, and it seemed like people ended up thinking it was "massive" (the term this article used) or quite a small difference.


In terms of territory, if you play someone the same strength as you, you expect each player to make about the same amount of territory. If you play someone four stones stronger, you expect them to make around three times as much territory as you.
Author:  skydyr [ Tue May 13, 2014 11:59 am ]
Post subject:  Re: Man vs Machine article in wired
Bill Spight wrote:
hyperpape wrote:
A related question: how would you explain how big a four stone handicap is to a layman? When the New Yorker article was posted, there was a lot of confusion about how big a four stone handicap really is by people who don't know a lot about Go. As players, we have an intuitive understanding of four stones against a pro, and that it's both pretty good, but still a big difference. Yet I don't know how to explain that to outsiders, and it seemed like people ended up thinking it was "massive" (the term this article used) or quite a small difference.


In terms of territory, if you play someone the same strength as you, you expect each player to make about the same amount of territory. If you play someone four stones stronger, you expect them to make around three times as much territory as you.


If I'm explaining handicapping, I usually try and work in a progression, like:

From the top pros, the best in the world, there are people who they can beat 50% of the time with 9 stones (around 1dan). There are people that those people can beat with 9 stones 50% of the time (9k). There are people THOSE people can beat with 9 stones 50% of the time (18k). Those players can still beat some people 50% of the time giving 9 stones(27k). And finally, those people will be able to give a few stones to a rank beginner.

Usually, people will get to 27k very quickly, within a couple weeks. Most will get to the next group within a month or two. Many will get to the group after that (9k) within maybe 6 months to a year. Some will get to 1d within a few years. Very few will ever get closer than 4-5 stones from a top pro, or 2-3 from the weakest pros, and even that is relatively rare.
Author:  hyperpape [ Tue May 13, 2014 12:52 pm ]
Post subject:  Re: Man vs Machine article in wired
Lots of nice ideas. I like the idea of an hour's head start in a marathon, in part because my brother just ran his first marathon, at age 39, and finished in 2:58, which is 53 minutes behind the world record. So I have a point of reference (that's the strength/weakness of particular analogies).

I was never a very fast runner, just like I am a mediocre go player. Even if I dedicated myself to it, I will never run a 3 hour marathon, or reach amateur 5 dan, even if I drastically ramp up my efforts. But I also have enough context to know that that amateur and my brother simply don't belong in a competition with someone who's world class.

David Foster Wallace has a nice bit about that in a piece on tennis. He idly wonders if he could play against a young man who is just breaking into the professionals, but almost instantly realizes otherwise when he sees him warming up. Later, he sees that kid play against Agassi, and sees a second huge gap in skill separating him from the top. But I'm even worse at tennis than I was as a runner, and I'm just taking that on testimony.
Author:  daal [ Tue May 13, 2014 11:42 pm ]
Post subject:  Re: Man vs Machine article in wired
Using my favorite metaphor for go, this is what it looks like when a pro goes up against a 4 stone handicap... even if they're not stones.
Author:  daal [ Thu May 15, 2014 4:43 am ]
Post subject:  Re: Man vs Machine article in wired
Here's an interesting tidbit from the article. Does anyone know anything more about this study?

Levinovitz in Wired wrote:
According to University of Sydney cognitive scientist and complex systems theorist Michael Harré, professional Go players behave in ways that are incredibly hard to predict. In a recent study, Harré analyzed Go players of various strengths, focusing on the predictability of their moves given a specific local configuration of stones. “The result was totally unexpected,” he says. “Moves became steadily more predictable until players reached near-professional level. But at that point, moves started getting less predictable, and we don’t know why. Our best guess is that information from the rest of the board started influencing decision-making in a unique way.”
Author:  RBerenguel [ Thu May 15, 2014 5:01 am ]
Post subject:  Re: Man vs Machine article in wired
daal wrote:
Here's an interesting tidbit from the article. Does anyone know anything more about this study?

Levinovitz in Wired wrote:
According to University of Sydney cognitive scientist and complex systems theorist Michael Harré, professional Go players behave in ways that are incredibly hard to predict. In a recent study, Harré analyzed Go players of various strengths, focusing on the predictability of their moves given a specific local configuration of stones. “The result was totally unexpected,” he says. “Moves became steadily more predictable until players reached near-professional level. But at that point, moves started getting less predictable, and we don’t know why. Our best guess is that information from the rest of the board started influencing decision-making in a unique way.”


I'm checking his Google Scholar and can't really find it :/
Author:  Shenoute [ Thu May 15, 2014 6:53 am ]
Post subject:  Re: Man vs Machine article in wired
RBerenguel wrote:
daal wrote:
Here's an interesting tidbit from the article. Does anyone know anything more about this study?

Levinovitz in Wired wrote:
According to University of Sydney cognitive scientist and complex systems theorist Michael Harré, professional Go players behave in ways that are incredibly hard to predict. In a recent study, Harré analyzed Go players of various strengths, focusing on the predictability of their moves given a specific local configuration of stones. “The result was totally unexpected,” he says. “Moves became steadily more predictable until players reached near-professional level. But at that point, moves started getting less predictable, and we don’t know why. Our best guess is that information from the rest of the board started influencing decision-making in a unique way.”


I'm checking his Google Scholar and can't really find it :/

Here but I don't know if it is accessible to all.
Author:  RBerenguel [ Thu May 15, 2014 7:52 am ]
Post subject:  Re: Man vs Machine article in wired
Shenoute wrote:
Here but I don't know if it is accessible to all.


Not even my university has access to it (first time this has happened!) It's available for free, though, through one of the writers' page.
Author:  Bill Spight [ Thu May 15, 2014 8:13 am ]
Post subject:  Re: Man vs Machine article in wired
Levinovitz in Wired wrote:
According to University of Sydney cognitive scientist and complex systems theorist Michael Harré, professional Go players behave in ways that are incredibly hard to predict. In a recent study, Harré analyzed Go players of various strengths, focusing on the predictability of their moves given a specific local configuration of stones. “The result was totally unexpected,” he says. “Moves became steadily more predictable until players reached near-professional level. But at that point, moves started getting less predictable, and we don’t know why. Our best guess is that information from the rest of the board started influencing decision-making in a unique way.”


RBerenguel wrote:
Not even my university has access to it (first time this has happened!) It's available for free, though, through one of the writers' page.


Thanks for the ref. :)

I took a quick look and my guess was right. The local configurations of stones come from joseki. :) (That was my guess.) So amateur dan players are better at playing joseki than kyu players and pros are better at not playing joseki than amateur dans. ;)

From the paper:

Michael Harré, et al. wrote:
In this work we focus on the 7×7 corner regions of the board where well studied patterns of moves, called Joseki, are played. Studying the move trees in this area provides an insight into how these well understood sequences of moves change with skill.
Author:  RBerenguel [ Thu May 15, 2014 8:22 am ]
Post subject:  Re: Man vs Machine article in wired
Bill Spight wrote:
Levinovitz in Wired wrote:
According to University of Sydney cognitive scientist and complex systems theorist Michael Harré, professional Go players behave in ways that are incredibly hard to predict. In a recent study, Harré analyzed Go players of various strengths, focusing on the predictability of their moves given a specific local configuration of stones. “The result was totally unexpected,” he says. “Moves became steadily more predictable until players reached near-professional level. But at that point, moves started getting less predictable, and we don’t know why. Our best guess is that information from the rest of the board started influencing decision-making in a unique way.”


RBerenguel wrote:
Not even my university has access to it (first time this has happened!) It's available for free, though, through one of the writers' page.


Thanks for the ref. :)

I took a quick look and my guess was right. The local configurations of stones come from joseki. :) (That was my guess.) So amateur dan players are better at playing joseki than kyu players and pros are better at not playing joseki than amateur dans. ;)

From the paper:

Michael Harré, et al. wrote:
In this work we focus on the 7×7 corner regions of the board where well studied patterns of moves, called Joseki, are played. Studying the move trees in this area provides an insight into how these well understood sequences of moves change with skill.


Meh! Not "so" interesting :/
Author:  gowan [ Thu May 15, 2014 12:52 pm ]
Post subject:  Re: Man vs Machine article in wired
I, too, am disappointed that they just examined how well players know joseki. In considering predictability of moves in general, I would expect very weak players to be difficult to predict and predictability to increase as the strength of the players increases, but after a certain point (amateur 1d?) I would expect the moves to be less predictable. The reason, I think, would be that moves are more and more predictable as players learn standard patterns (whole board, not just corner) but stronger players play forcing moves and interrupt patterns more often. Of course if we work with whole board situations the very idea of predictability becomes unclear (what is the predicted move?).

As an aside, there has always been a feeling that speed at playing through pro game records is an indicator of playing strength. It is also thought that this is the case because stronger players are better at predicting where the next move will be and so limit the search in the diagram.
Page 1 of 1 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/