Taking Your Chances

[Kirby]

It's the middle of the game - let's say mid to late middle game.

You have been trying to count the score. It seems that the score is somewhat close. Maybe you're a bit behind. Hard to say, since there are several big points left.

Then, your opponent doesn't defend his group. You think to yourself, "Can I kill it?" You read ahead some sequences - looks... feasible. Maybe. Not quite sure. If you had to guess, let's say you feel there's a X% chance you'd actually kill the group and win the game. But if you go for the kill, it will drive him into your territory - you'll lose the game for sure. On the other hand, you could decide not to try to kill the group. But the game seems really close. If you play normally, and just try to move into the endgame, and finish the game, maybe you estimate the chance of winning is Y%.

How do you make this decision?



Is this simply a matter of trying to kill if X > Y? Followup: How do you assess these probabilities?

[Uberdude]

There's another choice in some games: look for a way to get some outside moves in sente so that the likelihood of kill is increased. And read more.

[EdLee]

Hi Kirby,

I think this can happen at any level,
all the way up to top pro levels.
I forget who was Fujisawa sensei's opponent
in the famous game where Fujisawa read for ~3 hours on 1 move.
( Was his opponent Sakata or Kato ?
Maybe someone can post that game here. )
Well, Fujisawa sensei killed the huge group.
Afterwards, he said he read as far as he could --
he thought he could kill it -- but he said he still was not sure.

That's top level.

My anecdotal experience is if I had a situation like that,
and I got a high-level review atterwards, invariably it turned out
there were a myriad of things -- Z,W,V,U,A,B,C,... --
that I had missed, either long before that "tricky" position,
during the tricky position, or after it; or, all of the above.
So I had blind spots Z,W,V,U,A,B,C,... and they were
all "bigger" than X or Y. (I know I'm comparing apples and oranges,
but you know what I mean ? )

For example, I missed some sequence or tesuji much earlier;
I missed a kill earlier; I had a wrong direction earlier;
I had a bad joseki choice earlier; I missed some tesuji
during the tricky situation; I misjudged the situation --
it was not close at all; I missed some endgame sequence
or tesuji, etc. etc.

So what I learned was that yes, eventually this kind of difficult position
happens all the time, even at top levels.
But, I have a lot of problems that are much bigger
and more fundamental, that I have to fix or improve,
way before I'll reach those very difficult positions.
I dunno if this makes any sense to you?

[Shenoute]

I forget who was Fujisawa sensei's opponent
in the famous game where Fujisawa read for ~3 hours on 1 move.
( Was his opponent Sakata or Kato ?
Maybe someone can post that game here. )
Well, Fujisawa sensei killed the huge group.
Afterwards, he said he read as far as he could --
he thought he could kill it -- but he said he still was not sure.

That's top level.

Here is the game. Fujisawa spent 177 minutes on move 93.

As for the original question, I guess it depends on how important is the game to me. In a tournament game, I'd go for the kill only if I am pretty sure it works. If it is a friendly game at the go club, I would probably try to kill even if not 100% sure it is feasible.

[Bill Spight]

It's the middle of the game - let's say mid to late middle game.

You have been trying to count the score. It seems that the score is somewhat close. Maybe you're a bit behind. Hard to say, since there are several big points left.

Then, your opponent doesn't defend his group. You think to yourself, "Can I kill it?" You read ahead some sequences - looks... feasible. Maybe. Not quite sure. If you had to guess, let's say you feel there's a X% chance you'd actually kill the group and win the game. But if you go for the kill, it will drive him into your territory - you'll lose the game for sure. On the other hand, you could decide not to try to kill the group. But the game seems really close. If you play normally, and just try to move into the endgame, and finish the game, maybe you estimate the chance of winning is Y%.

How do you make this decision?

Sounds like you are playing against me.

[PeterN]

In theory should just go with whatever of the two you think has the highest %, though good luck estimating that. In reality if I think there's some reasonable chance of a kill I will almost certainly go for it when I get caught up in the game. This does not always work well....

PeterN

[mlund]

If there's a likelihood of a kill that you can't read and you don't think your opponent can read it either, then the primary question left is not "can I kill my opponent's group?" - since the answer is already: "maybe." The question is: "What can I get if I force my opponent to live?"

If the game is too close to count, then the difference in score made up my pushing against your opponent over and over again without losing sente should be enough to swing it in your favor, yes? Instead of chasing him out into your territory, reducing him while sealing your territory. He's got to respond, right? If you just seal him and he plays away then you attack his vital point and try to kill him - with a much higher change of success. What the worst that can happen at that point? He lives in Gote and you get to play your response to his original move? Not a bad deal if you ask me.

I've been taught never to try to kill unless it is a last resort. "Do or do not - there is no try," as Yoda says. When a group's life is actually in question, I make profit forcing it to live. Diving in on a gamble kill is something I'm only supposed to do if I'm otherwise so far behind I can't salvage the game. If I win because I killed something that I couldn't read the kill on, I've already failed my strategic objectives for this game and the important take-aways in the review will have nothing to do with this attempted kill.

Marty Lund

[Solomon]

There are a number of confounding variables that X and Y are dependent on, including:
* whether my opponent is a friend or a stranger
* the number of times my fiance asked me to quit the game so that I help her prepare dinner in the kitchen
* how badly I need to use the restroom if the byoyomi period is too short for a restroom trip
* how many glasses of wine I had

[mhlepore]

Not trying to evade the question, but would perhaps a better exercise be to identify how to avoid such situations?

For example, maybe not looking for the best move, but just "good enough" moves, during the games can culminate in a tense situation toward the end that you describe?

[Kirby]

Not trying to evade the question, but would perhaps a better exercise be to identify how to avoid such situations?

For example, maybe not looking for the best move, but just "good enough" moves, during the games can culminate in a tense situation toward the end that you describe?

Yes, of course. I'm still interested in this topic, though.

[Bki]

I guess it would depends on many factor. If I know my opponent well enough to be aware of his endgame skills, I think that would be the deciding factor : If I'm better, then assuming the game is close after I make all the forcing move I can against the weak group, I will likely get the advantage. If he is better at the endgame, then I would be more likely to lose.

More generally, it would depends on how certain I would be about the kill. Whether the group is completely eyeless, how solid is the position I will run toward it..

If I can't estimate which option would be more likely to lead to victory, then I guess I would chose to go for the kill. It's impossible to tell which choice is better, so I might as well chose the more fun and complicated way. Whether I fail or succeed, I would study quite a bit the game afterwards to see whether I should have failed or succeeded. That way my fighting skills will improve and the next time I find myself in such a situation I will be just a little bit better at judging the best solution.

[Uberdude]

I find this "never try to kill" philosophy a load faux-zen kumbaya bollocks. If your opponent makes a mistake and gives you a good chance to end the game decisively now then take it. If you make them resign then you can't screw up and lose later. Guo Juan often makes this point in reviews. Another point my friend Andrew Kay likes to make is it doesn't just win you this game, it helps you win your next game against them by crushing their spirit.

This is not to say go for those ridiculous kills where you chase them through your former territory with little chance of success, but killing is part of Go and to reject it is to be an incomplete player.

As an example, here's a game I played in last weekend's British Championship Candidates' tournament. At move 85, due to his earlier mistakes making my outside stronger, I had a chance to cut off his group and go for a kill. Now this isn't so much like Kirby's situation in that there is fairly minimal risk in going for the cut and kill: he is contained and I only lose a few points and a little bad aji in the boundary of my centre moyo. But given that situation, and the clock situation (I had used over 1 hour of 90 minutes main time, he had used less than 1 hour) and my history of being stupid in byo-yomi, I went for the kill, and succeeded in ending the game promptly.

http://files.gokgs.com/games/2015/5/4/Uberdude-4.sgf

[moyoaji]

I am currently facing this challenge in many of my games. As of late, I've been trying to hold back on the killing to see if I can win with endgame. However, sometimes I look at the kill as a learning experience and go for it.

But, if I'm really truthful, it depends more on my mood at the time...

Sometimes, I just feel like killing a group. Other times, I feel like playing calmly. That will usually be the deciding factor in this type of a game. This is not scientific, just me being honest.

[skydyr]

I find this "never try to kill" philosophy a load faux-zen kumbaya bollocks. If your opponent makes a mistake and gives you a good chance to end the game decisively now then take it. If you make them resign then you can't screw up and lose later. Guo Juan often makes this point in reviews. Another point my friend Andrew Kay likes to make is it doesn't just win you this game, it helps you win your next game against them by crushing their spirit.

This is not to say go for those ridiculous kills where you chase them through your former territory with little chance of success, but killing is part of Go and to reject it is to be an incomplete player.

I wouldn't say that one should never kill by any means, but there is a point in a player's development, often in the SDK range, where they have to learn to profit by attacking without expecting to kill. In addition, killing can be a very do or do not kind of thing if you go all out to do it, so trying to kill speculatively and creating weaknesses in the process can be quite counterproductive. If you can be relatively certain of a kill, then by all means go for it, but it takes experience to tell when one can be relatively certain of a good result without being able to read to the end of the sequences. A lot of the players who try to think this way, though, do it because they have the opposite tendency and need to dial it back to balance their game and continue to improve.

[Bill Spight]

I find this "never try to kill" philosophy a load faux-zen kumbaya bollocks. If your opponent makes a mistake and gives you a good chance to end the game decisively now then take it. If you make them resign then you can't screw up and lose later. Guo Juan often makes this point in reviews. Another point my friend Andrew Kay likes to make is it doesn't just win you this game, it helps you win your next game against them by crushing their spirit.

This is not to say go for those ridiculous kills where you chase them through your former territory with little chance of success, but killing is part of Go and to reject it is to be an incomplete player.

As an example, here's a game I played in last weekend's British Championship Candidates' tournament. At move 85, due to his earlier mistakes making my outside stronger, I had a chance to cut off his group and go for a kill. Now this isn't so much like Kirby's situation in that there is fairly minimal risk in going for the cut and kill: he is contained and I only lose a few points and a little bad aji in the boundary of my centre moyo. But given that situation, and the clock situation (I had used over 1 hour of 90 minutes main time, he had used less than 1 hour) and my history of being stupid in byo-yomi, I went for the kill, and succeeded in ending the game promptly.

An instructive game, but painful to watch.