Life In 19x19 http://lifein19x19.com/ |
|
This 'n' that http://lifein19x19.com/viewtopic.php?f=12&t=12327 |
Page 9 of 53 |
Author: | bayu [ Wed Jan 13, 2016 9:13 am ] |
Post subject: | Re: This 'n' that |
Bill, you write faster than I can read. I really appreciate your effort. Very enjoyable quality read. It was great to use the oldest known game record for a joseki question. It shows that there must have been a sensible amount of training and knowledge around back then. Do you happen to know, why that game (or half of it) got recorded and survived till today? I got lost with the tinies and minies. Is there a good reason for the difference? As I see it, a tinie for one player is a minie for the other and vice versa. I assume that they also appear in some sort of capture go. |
Author: | Bill Spight [ Wed Jan 13, 2016 11:21 am ] |
Post subject: | Re: This 'n' that |
bayu wrote: Bill, you write faster than I can read. I really appreciate your effort. Very enjoyable quality read. Thank you, bayu. ![]() Quote: It was great to use the oldest known game record for a joseki question. It shows that there must have been a sensible amount of training and knowledge around back then. Do you happen to know, why that game (or half of it) got recorded and survived till today? A question better addressed to John Fairbairn or Peter Shotwell. ![]() Quote: I got lost with the tinies and minies. Is there a good reason for the difference? As I see it, a tinie for one player is a minie for the other and vice versa. Tinies and minies are infinitesimals. A tiny is a White sente, and is a plus for Black; a miny is a Black sente, and is a plus for White. Here is a miny in straight No Pass Go. Each player has 1 point of territory, for a net count of 0. But there is a Black sente, or miny. Minies are less than 0, which means that White wins, regardless of who plays first. ![]() ![]() ![]() ![]() ![]() Quote: I assume that they also appear in some sort of capture go. There are no known tinies and minies in regular go, except in chilled go, as explained on SL. I suppose that that holds for the Capture Game and Capture-N, as well. ![]() |
Author: | bayu [ Wed Jan 13, 2016 12:13 pm ] |
Post subject: | Re: This 'n' that |
Quote: Here is a miny in straight No Pass Go. Each player has 1 point of territory, for a net count of 0. I always feel embarrassed finding out that I can't even count to 3. ![]() How do you count white as 1? I naively count it as 1.5: if white plays first, it's 2 points; if black plays first it's 1 point; take the average. Maybe got something to do with Sente gains nothing. Black has the privilege to play the sente, so white's count is 1. But then: Quote: Minies are less than 0 Sente gains nothing. I also thought that sente loses nothing either. Is this not true for straight No Pass Go? ![]() I hope these questions make sense. I'm not entirely sure;) Quote: There are no known tinies and minies in regular go, except in chilled go, as explained on SL. Thanks for explaining this. Maybe, Senseis could stand some improvement. The examples are given for chilled go, but regular go is not mentioned under tinies. It might be obvious when coming from a CGT perspective. |
Author: | Bill Spight [ Wed Jan 13, 2016 2:46 pm ] |
Post subject: | Re: This 'n' that |
bayu wrote: Quote: There are no known tinies and minies in regular go, except in chilled go, as explained on SL. Thanks for explaining this. Maybe, Senseis could stand some improvement. The examples are given for chilled go, but regular go is not mentioned under tinies. It might be obvious when coming from a CGT perspective. Thanks for pointing that out. ![]() Quote: Quote: Minies are less than 0 Sente gains nothing. I also thought that sente loses nothing either. Is this not true for straight No Pass Go? ![]() I hope these questions make sense. I'm not entirely sure;) Thank you for your questions. They are definitely sensible. ![]() The saying that sente gains nothing refers to points. OC, playing a sente takes away the opponent's option of playing the reverse sente. That is worth something, but it is not something that is measured in points. The analogy that taking sente is like cashing a check is a good one. And sente gains nothing in No Pass Go as well. ![]() A miny (a Black sente) is less than zero because White gets the last play in it, whether by playing the reverse sente or by answering the Black sente play. The difference between a miny and zero cannot be measured in points. This idea is new and strange to most people, who have no experience with infinitesimals. ![]() Quote: Quote: Here is a miny in straight No Pass Go. Each player has 1 point of territory, for a net count of 0. How do you count white as 1? I naively count it as 1.5: if white plays first, it's 2 points; if black plays first it's 1 point; take the average. Maybe got something to do with Sente gains nothing. Black has the privilege to play the sente, so white's count is 1. There are a couple of things going on here. First, as you say, sente gains nothing. So let's look at the result after the Black sente. This position we can count. ![]() The second thing that is going on is that territory does not mean the same in straight No Pass Go as it does in regular go. In regular go Black has three points of territory and White has four points of territory. But in no pass go Black has only one point of territory because of the "group tax". Black has to keep two eyes to stay alive. So Black can only fill one eye, and has only one point. As for White's territory, if White did not need the eye with the ![]() ![]() ![]() And here is why a miny is an infinitesimal. Let's look at the position after White plays the reverse sente. In this case White has only one point, as well. Not only does the sente gain nothing, the reverse sente gains nothing, too! ![]() |
Author: | Bill Spight [ Thu Jan 14, 2016 8:18 pm ] |
Post subject: | Re: This 'n' that |
Here is an easy problem. Like all my problems. ![]() Enjoy! ![]() |
Author: | Bill Spight [ Fri Jan 15, 2016 11:45 pm ] |
Post subject: | Re: This 'n' that |
OK. Easy problem. No particular point in hiding this. ![]() ![]() ![]() ![]() Usually a difference of 1/32 point will not matter, but here it does. If White takes the 1 point play, Black will get a net local score of 1 point. More on this little problem in the morning. ![]() |
Author: | Bill Spight [ Sat Jan 16, 2016 6:49 pm ] |
Post subject: | Re: This 'n' that |
Difference games I learned about difference games in the 1990s, from Berlekamp and Wolfe's Mathematical Go. Although the idea is mathematical, applying it does not require much in the way of math. The basic idea is that independent regions of the go board may be added and subtracted. Go tends to break up into independent regions in the endgame, and even before then regions may be only loosely coupled. To subtract one region from another one, change the colors of the stones in the first one and add the two together. For instance, for this region: (The marked points are the points in play.) this region is its negative. OC, a region minus itself is zero. If you play the difference game out, the second player just plays mirror go. A major use of difference games is to compare two plays. Let's take the little problem and subtract it from itself. The bottom region is the negative of the top. In between is no man's land. This is a zero game, OC. ![]() In the problem White has two plausible moves, which we wish to compare. We let White make one of those moves in top region, and let Black make the other move in the bottom region. If White's move is better than Black's then the combined position should be better for White than for Black (and vice versa). That means two things. First, White to play should win in the difference game. Second, Black to play should not win. Let's check that out. White to play wins by one point. You can let Black try different moves, but White always wins. (Note that if White starts at 2 instead of 1, Black can play at 1 and tie by mirror go.) Again, if Black starts at 2 White can play at 1 and play mirror go. ![]() ![]() Now, let's play the difference game for the case when Black has the choice of moves. ![]() Black wins by one point, as expected. Again, if White starts at 3, Black at 4 is jigo. We can stop after ![]() What does that mean? It means that we cannot say that one play is better than the other for Black. In some situations, one may be better, in different situations the other one may be better. Yes, the ![]() Note that what allows White to win the difference game, going first, is the fact that ![]() ![]() ![]() One advantage of comparing plays by difference games is that you do not have to calculate the size of the plays. ![]() The caveat is ko. Why? Because we assumed that the regions were independent, and kos may break the independence of different regions. For instance, the number of possible ko threats may matter. And in this case we treated "a" and "b" in the original problem as miai, with one player getting one and the other player getting the other one. In a game with a ko, however, one player might win the ko while the other player gets both "a" and "b". Each case must be decided on its merits. Difference games can help us to make close calls. ![]() |
Author: | Bill Spight [ Sun Jan 17, 2016 4:32 pm ] |
Post subject: | Re: This 'n' that |
Difference games (ii) I was quite surprised when David Wolfe told me that the White descent in the next diagram was not as good as the hane-and-connect. The reason being, he explained, was that it allowed Black to play ![]() This surprised me for two reasons. One, in estimating territory here, the exchange of ![]() ![]() ![]() The difference game makes David's point clear. ![]() For the difference game we set up the negative of the original position, then in the top we let White play the descent and in the bottom we let Black play the hane-and-connect. Black to play makes one point of territory (marked). White to play only gets jigo. The hane-and-connect is correct. Takagawa was wrong. ![]() ---- When I was a beginner one thing puzzled me -- among others, OC. ![]() How should Black save the ![]() Unlike a lot of beginners, I was not attracted to the descent because it gave White ![]() Can difference games can shed some light on this question? You bet. ![]() First, let's compare the descent with the solid connection. ![]() White to play makes an extra point of territory (marked) and wins. Somewhat surprisingly, Black to play also wins. We cannot say that the solid connection is better. Now let's try the descent vs. the hanging connection. White to play wins, as with the solid connection. White gets jigo. But suppose that Black does not submit with ![]() White simply ignores ![]() ![]() ![]() ![]() So my beginner's intuition was correct, and the descent is normally inferior. ![]() Finally, let's compare the solid connection and the hanging connection. White makes one more point of territory (marked) and wins. No surprise. ![]() White gets jigo. So the difference game says that the hanging connection is best. ![]() Wait! Are the pros wrong to play the solid connection, unless there is a tactical reason? Remember the caveat about kos. The hanging connection leaves possible ko threats which the solid connection does not. Does that matter? Well, every case must be judged on its merits. ![]() |
Author: | Bill Spight [ Wed Jan 20, 2016 3:46 pm ] |
Post subject: | Re: This 'n' that |
Difference games (iii) More on the second line hane. ![]() The second line hane occurs frequently, as does the descent on the second line. Often the difference between playing one or the other depends upon tactics. But what about when there is no particular tactical reason to choose one or the other? Let's explore that question with difference games. ![]() Which is better for the player to the right, the hane or the descent? On the top side we let White play the descent, on the bottom side we let Black play the hane. The answer is not obvious, nor is the play in the difference game, which could get complicated. Fortunately, in difference games we do not have to find the best play to win or tie; good enough play will do. ![]() With Black to play ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() What if Black plays the hanging connection instead. That does not look promising, but let's see. White wins by 2. The first player wins the difference game, which means that whether to play the hane or the descent depends upon what's on the rest of the board. ---- Some time ago I noted that the hane-and-connect on the second line was fairly usual play, but that the descent was normally sente. That being the case, if you foresee that your opponent can get the hane-and-connect, why not forestall that by playing the descent with sente? Furthermore, by symmetry, that is so for both players, so normally shouldn't one player or the other play the descent with sente? In that case, we should see the pros make a lot of double descents on the second line, but we don't. ![]() However, if you play the hane you do not necessarily have to connect, even when tactics don't matter. What do difference games tell us about playing the hane with sente versus doing nothing? ![]() With Black to play, the descent is good enough to win. If White responds on the top side with a descent, Black connects on the bottom side. OTOH, if White captures the Black stone on the bottom side with gote, Black plays the monkey jump on the top side. With White to play, the descent is also good enough to win. If Black responds with a descent on the top side, White captures the Black stone on the bottom side. OTOH, if Black saves the Black stone, White plays the monkey jump. The upshot of it is this. Sometimes you play the descent to prevent your opponent from playing the hane, sometimes you play the hane with sente to prevent your opponent from playing the descent, and sometimes you wait. ![]() |
Author: | Bill Spight [ Sat Jan 23, 2016 12:24 pm ] |
Post subject: | Re: This 'n' that |
Difference games (iv) Comparing plays with difference games has some practical advantages by comparison with calculating the size of plays. First, you do not have to calculate the size of plays. In my little problem White did not have to calculate the size of the either the play that gains 1 point or the play that gains 1 1/32 point. True, neither is that difficult to calculate, but to calculate the value of a play may mean reading out a somewhat sizable game tree, and calculating the values of other plays in the tree. Reading out a difference game may also involve a sizable tree, but the effort involved is not necessarily more. Besides, once you reach a mirror go position you can stop reading that variation. ![]() Second, unlike with calculating values, you do not need to find perfect play. Good enough play will do. We saw the practical advantage of that in the second line hane vs. descent example. ![]() Third, difference games can distinguish between plays of the same size. There are even plays of the same size where the difference game prefers one to the other, regardless of who plays first. Commenting on such a play, one pro suggested adding half a point to the value of that play. ![]() Fourth, difference games can tell us when a smaller play is a viable option. In my little problem, the difference game tells Black that the play that gains 1 1/32 point may not be as good as the play that gains 1 point. ---- Difference games in straight no pass go Back in the 90s the late John Rickard and I explored play in some large eyes in no pass go. One problem, OC, is that no pass go values are not familiar to regular go players, so that it was not always easy to eliminate certain inferior lines of play. Difference games could have helped, but I did not use them, and I don't think that John did, either. (We did not get together in person, but corresponded via email.) Here is a simple example. ![]() ![]() Should White play at "a" or "b"? Our regular go sense tells us "b", because it threatens to take away a potential eye. But life and death is not the issue. OC, a one point eye does not have a fractional value in no pass go, so that may matter. We might work out the values of the resulting positions. (In fact, didn't I do that last week? But what were those values? ![]() ![]() Let's do the difference game. Now, let's say that all I am interested in is whether ![]() ![]() ![]() That means that all I have to do is to look at the game where Black plays first. If I get "jigo" (a zero game) as White, then ![]() ![]() ![]() If ![]() ![]() ![]() Well, if ![]() ![]() Now if ![]() ![]() ![]() ![]() ![]() ![]() ![]() After ![]() After ![]() So ![]() ![]() BTW, in case you are curious, In this position White is 0.125 point better on average than Black. ![]() ---- Now, in straight no pass go we know that it is generally a good idea to play inside your opponent's eye towards the end of the game. But what about playing inside your own eye? Which plays are good? Well, since, group tax aside, a one point eye is worth a full point, while larger eyes are worth a fraction of the number of board points they surround, making a one point eye seems like a good idea. OC, that may not always be so. ![]() And, in fact, it is not so when making the first play inside your own seven-point linear eye. As this difference game illustrates. Is ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Black has 2 + 2 + 1 = 5 points. (The eyes with the single dead White stones are worth 2 points, one to capture the eye and one to fill it. ![]() White has 1 + 1 + 2 = 4 points. (The eye with the two Black stones is worth 2 points, as you can see after White captures the two stones and then Black plays inside the eye.) Black wins. So we cannot say that ![]() ![]() But perhaps ![]() ![]() ![]() ![]() ![]() ![]() So ![]() ![]() ![]() The Black "territory" is worth 4.5 points. The White "territory" is worth 4*. Black is half a point ahead, give or take the equivalent of a dame. ![]() |
Author: | Bill Spight [ Mon Jan 25, 2016 5:16 pm ] |
Post subject: | Re: This 'n' that |
Yet Another Difference Game I had not intended to post another difference game now, but this seems kind of interesting. Straight no pass go, Black to play. Ignore the group tax. ![]() Should Black play at "a", "b", or "c"? Or does it depend upon the rest of the board? ("c" and "a" are equivalent, OC, so forget "c". ![]() At first I thought that "a" was best, because the count after "a" is 4.5, while the count after "b" is 4. But if Black "a", White "c", the local score is 4, while after Black "b" the local value is infinitesimally greater than 4, by 2 UPs. So maybe there are times when "b" is right. OC, we set up a difference game. ![]() Enjoy! |
Author: | Bill Spight [ Wed Jan 27, 2016 8:42 am ] |
Post subject: | Re: This 'n' that |
OK. Here is the solution. Which is better depends on the rest of the board. ![]() I.e., whoever plays first wins. ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ---- Many, if not all of these No Pass Go positions have analogs in regular go. For instance, here is an analog for the White position after ![]() ![]() ![]() ![]() See http://senseis.xmp.net/?DOWN |
Author: | emeraldemon [ Wed Jan 27, 2016 10:15 am ] |
Post subject: | Re: This 'n' that |
Hi Bill, I was watching a Haylee game on youtube and this position came up in the endgame: Haylee was white. After ![]() ![]() |
Author: | Bill Spight [ Wed Jan 27, 2016 3:03 pm ] |
Post subject: | Re: This 'n' that |
emeraldemon wrote: Hi Bill, I was watching a Haylee game on youtube and this position came up in the endgame: Haylee was white. After ![]() ![]() My first thought on seeing this was that ![]() ![]() ![]() ![]() This is a difficult ko for White to play, as White puts more at stake than Black. The practical thing is usually to play ![]() ![]() As ![]() ![]() ![]() ![]() Let's compare that with the result after ![]() ![]() ![]() ![]() ![]() ![]() ![]() Since our current theme is difference games, what does the difference game tell us about these two positions? Difference games do not work well with kos, but let's assume that Black avoids the ko in the bottom right corner. ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() It seems that Black to play wins the difference game by taking or threatening to take the ![]() That means that under some circumstances, depending on the rest of the board, the second line atari is right. ![]() |
Author: | Bill Spight [ Thu Jan 28, 2016 7:30 am ] |
Post subject: | Re: This 'n' that |
Correction Actually, we can remove the exception about needing the corner eye if we go back to the original comparison. This is the original game to compare ![]() ![]() Normally, to play ![]() ![]() ![]() ![]() OC, White to play wins. ![]() ![]() OC, we are assuming that Black avoids the ko on the right. ![]() |
Author: | Bill Spight [ Sun Feb 07, 2016 3:32 pm ] |
Post subject: | Re: This 'n' that |
Error in Tesuji and Anti-Suji of Go? This example comes from Sensei's Library ( http://senseis.xmp.net/?canceledit=Tesu ... o%2FErrata ), where Dieter (Knotwilg here) suggests a different play from the one Sakata gives. Problem: How to connect the three isolated black stones to the corner? Sakata's solution starts with ![]() Dieter suggests this way. Is Dieter right? Edit: That is, is Dieter's play better? Enjoy! ![]() |
Author: | Pio2001 [ Mon Feb 08, 2016 10:34 am ] |
Post subject: | Re: This 'n' that |
Author: | Jhyn [ Mon Feb 08, 2016 11:28 am ] |
Post subject: | Re: This 'n' that |
Pio2001 wrote: |
Author: | Shaddy [ Wed Feb 10, 2016 1:39 pm ] |
Post subject: | Re: This 'n' that |
Author: | mitsun [ Wed Feb 10, 2016 7:22 pm ] |
Post subject: | Re: This 'n' that |
Same principle:
|
Page 9 of 53 | All times are UTC - 8 hours [ DST ] |
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |