KoThreat/Discussion

Sub-page of KoThreat

Table of contents

Discussion whether the follow-up of a ko threat is a "Sente" move

tderz: Sente is relative and – or, hence – a ko threat implies sente simply because the opponent loses (at least) one move to resolve the ko.
It does not mean, that – if the very same move – was played without any ko involved, one had to classify this move as a sente move.

Bill: No, it could be an ambiguous move or part of a reversal, but this is a page for beginners. If we use the term “ko threat” for any play, then it loses its meaning. A ko threat at least has to maintain the ambient temperature, i. e., to produce a play that is the largest play on the board (aside from the ko). If there were no ko, such a play would be sente, in the vernacular. It might not fit all of the different senses of the term, however.

I thought that the qualification, aside from the ko, was clear, but, just in case, I have amended my statement on the main page.

Anyway, the reason I had for stressing the sente aspect of ko threats is that the typical introduction to ko threats stresses winning the ko. The natural question there is whether to answer the threat or not, and you compare the threat with the ko (naively -- there's much more to it). Well, that's all very good, but ignores the idea of the ko exchange, which is the typical end of ko fights. You don't get something for nothing, as a rule. If your ko threat is ignored, you want to get something for making the threat, at least to stay even. So the question becomes, not whether the threat is answered, but whether it makes a profit, if ignored, and you compare the threat with other plays on the board.

So the right way to think about ko threats is the latter. Not, does it let me win the ko, but does it let me gain something if my opponent ignores it. That is not a difficult concept, and it is best for beginners to get it from the start.

It also makes kos less daunting for beginners. You don't have to worry so much whether you can win the ko, as whether you can get something in exchange for losing it. That's less to worry about.

tderz: I agree with your argumentaion in your last 3 paragraphs.
I only estimate your 2 premisses "... produce a play that is the largest play on the board (aside from the ko)." and
"If there were no ko, such a play would be sente, ..." as wrong, because too general (i.e. not always true).
Both are neither necessary (to win a game) - i.e. if you even win with your smallest ko-threat move - or ally your ko-threat continuations are gote moves,
nor sufficient (you might always play those threats with the biggest continuation or which are sente, but because of that you lose the game).

I have to check those terms: ambiguous, reversal, ambient temperature, I do not yet know/understand their concepts.

JuhoP: How can you say that "... produce a play that is the largest play on the board (aside from the ko)" is wrong? Think of a situation where the follow-up is not the biggest play on the board. You make your threat, your opponent ignores it and fills the ko. Now you cannot play the follow-up to your 'ko threat' because it is not the biggest play on the board. You have to play somewhere else (where you could have played anyway). Surely this means that the 'threat' was no threat at all. Or am I missing something?

Bill: Upon reflection, I am eliminating the reference to sente. While I think that the idea is pedagogically helpful, I think it is inappropriate for a brief comment.

tderz: Bill, then add it as pedagogical hint like "... and we know that sente plays are usually quite big (threats)".
For most cases you were right anyway :-)

Bill: TDerz, one problem is that the term “sente” is ambiguous. In endgame discussions I usually use the technical sense which refers to a classification of plays. Here I intended it in the common usage as a more or less forcing move. In addition, I am talking about the rest of the board besides the ko. Too confusing for this sort of page, which should be brief.

JuhoP, in my comment I forgot what I stress so often: The largest play is not always the best play. That's why I have added my usual disclaimer, "as a rule". Surely there are cases, even with an ordinary, regular ko, where a ko threat does not leave the largest play on the board.

tderz: The whole confusion derives from the word 'threat'.
Actually one is making that 'ko threat' move only because we are deprived by rules to re-take that ko immediately (and we do not want a jigo or no-result').
The two - perhaps not sufficient or complete - considerations for that move are:

  • might it allow me to re-take the ko? Independent of YES or NO as an answer, the next question is :
  • what are the consequences for the rest of the play?

That's it.
Of course the answer to these related questions can be complete mini-games (involving the concepts of sente and big(est) moves).

Bill: Well, there is another question. If the threat is not answered, what advantage is gained in the current position from having made the threat?

It seems to me that the question of the consequences of making the threat subsumes the questions of what happens if the threat is answered and what happens if it is not.


Discussion whether the follow-up of a ko threat is the largest play on the board

tderz: Bill stated in KoThreat#100: "ko threat is a move that creates a new play as an alternative to the ko. The new play is the largest play on the board, aside from the ko."
I rather suggest amending Bill's wording to "The new play is usually/most often the largest play on the board, aside from the ko."

Disclaimer1: LocalKoThreat displays examples which could be interpreted that the value of the move itself which follows the ko threat could be as low as 1 point or even zero (i.e. a connection), but in the ko/game context they still might hamper the opponent from winning the ko or the game. (my argument is somewhat flawed, because also the "zero-point connection" has a value in the range of things at stake in the ko).

Disclaimer2: A ko threat does not have to be the largest play on the board. (I understand that Bill said the following next play should be the largest on the board) There are many alternatives, depending on he type, size and importance of the ko, whether one is behind or winning, the stage of the game and its particularity.
I'll give some examples:

  • loss-making ko threats?: one is playing moves which are losing 1, 2 or many more points (hence move value itself is minus). The ko might be an all-dominating ko and each player has to win it. You do not care about the individual (negative) size of the ko threat, as long as the sum of all such plays is not overshadowing the value of the ko itself.

Bill is right here that the very ko threat, i.e. the created next play should be big, hopefully bigger than the (game-dominating) ko value.

    • Example:

Ko value = 10 (i.e. a direct, not game dominating ko?)
stage: late yose
situation1: you seem to be 3 points in front if nothing rare happens and you win the ko ([200] i.e. you do not gain 10 points by winning it, you're just preventing losing 10)
ko material (your own): 3 threats KT1, KT2, KT3

      • KT1 loses 3 points if answered, but has a threat of 50 points (big, more than needed)
      • KT2 is loss-neutral if answered and has a threat of 12 points
      • KT3 is loss-neutral if answered and has a threat of 10 points

ko material (opponent): only 2: KToppo1, KToppo2

      • KToppo1 loss-neutral if answered, threat of 4 points
      • KToppo2 loss-neutral if answered, threat of 3 points

ko material difference: you have one threat more.

Variant 1: It's the opponent's turn (= burden of finding the 1st threat):
KTo1, reply, ko, KT2, reply, ko, KTo2, reply, ko ... you win the ko (and the game by 3 points) because the opponent did run out of ko threats (s/he had to start first).
Please notice that you did not start with the loss-making ko threat (which was the biggest threat) only to obtain jigo (q.e.d.).

Variant 2: It's your turn:
KT2, reply, ko, KTo1, reply, ko, KT3, reply, ko, KTo2, reply, ko, now you have to play KT1 and only can get jigo.

Bill: It appears that it is a mistake to answer the opponent's 3 point threat. Play continues:

win ko, KTo2 completion, elsewhere.

tderz: ... resulting in jigo, due to premise [200] above (jigo is not as good as winning by 3 points)

By comparison, you have reply, ko, KT1, reply, ko, elsewhere, win ko, elsewhere.

Let's suppose that things are normal, and that the correct alternative to the ko is a gote, which we shall call G, and that playing G gains g points for each player.

Then the result of making the losing ko threat to win the ko is 10 (for the ko) - 3 (for the losing threat) - g (for the opponent's playing G) - s (for the value of what's left) = 7 - g - s = 0.

The result of ignoring the opponent's 3 point threat is 10 (for the ko) - 3 (for the opponent's threat) + g (for playing G) - s' (for the value of what's left) = 7 + g - s' = 2g + s - s'. Now, the value of what's left may differ in the two cases, but the difference (s'- s) will not exceed g. So the result is >= g, and g > 0. Ignoring the 3 point threat wins.

Lesson: it is not (only) about the size of the threat but about the sum of won/lost values during the whole ko sequence.


tderz: Another realistic example could be construed, where you are in the middle of the game, have a (non-game dominating) ko of moderate size (10-20 points), do not know whether you are ahead or behind, have several, non loss-making, extreme monster ko-threats, and many (you do not know how many you will have or get during play) ko threats and you do not know whether your opponent will have/get more ko threats.
Why to throw nuclear ko threats first if you might need them later if there starts something really important during the game?
I would suggest to play those threats with an estimated value a little bit bigger than the ko and wait and see (preserving the nukes for later).


Other material moved from main page

1. Kris was concerned that the description be easy for beginners to follow, which I think has been accomplished now, but here is his comment:

Kris Rhodes: I want to insert a comment here for any relative beginners to the game before you get to the slightly complicated comments which follow on this page.

I think the simplest way to explain a ko threat is to say it is a place on the board where, if you could get two moves in a row, you would win a lot of points. So, if a ko situation happens, and you want to win that ko, then you can play the first of the above-mentioned two moves (which need have no direct relationship to the actual area of the ko fight situation itself). You hope that this will demand a response by your opponent in that area of the board, and the ko will remain available for you to retake.

Bill: [100] A ko threat is a move that creates a new play as an alternative to the ko. If the opponent can win the ko on the next play, the new play is the largest play on the board, aside from the ko, as a rule.



Note to tderz: Please note the level of this page: it's marked "Beginner", for a reason. Please move your comments either to the /Discussion page, if it is something you want to discuss or another page, and restore this page to its previous state. This page is intended to be a readable, concise, simple explanation of Ko threats for beginners.

Bill: This is a reference page. (Our beginners ko threat page is Finding ko threats.) However, I agree that it should only say what a ko threat is -- not an easy task, BTW. Strategy, etc., belongs elsewhere. Much of the recent discussion here should be moved and maybe an earlier version revived.

tderz: Thanks Bill. Indeed I did not look at all at the page type. I will move it to discussion.

Bill: Thank you. But I do feel that the discussion of strategy (which is still on the main page) deserves its own page, with a link. Keep reference pages short and sweet. :-)


Static Evaluation

  • Before a ko fight one should consider several numbers:
    • Value of ko "V" (roughly put here in one number, which often is not possible)
    • number of own ko threats exceeding value V ("n(owk>V)")
    • same for opponent ("n(opk>V)")
      • number & size of own ko threats of same or lower value than V
      • same for opponent

We could describe several outcomes and attribute the names komaster, koloser to the possible constellations (e.g. (n(owk>V) >> n(opk>V) => excess of number of threats = komonster).

In any case one should try to have picture concerning these figures. This is a workable method if we were assuming a quite static game (i.e. in the very late engame it is like this, some "1/2-point" kos, you simply have a kind of NIM?-game, if you cannot find a special move for breaking out).

In the middle game, the situation is very different:
ko threats dis-continuously alter the game in a dynamic way and
it is difficult to foresee the changes of the game by the threats made.

Dynamic Consideration This leads me to the point that one has to consider carefully how the (middle-)game would change by all the threats made and replied to,
even if the ko is eventually won/lost.

  • If you have one or more ko threats "big enough" (threatening enough damage), so that your opponent should answer them, play the smallest one.

If you will win the ko at your next opportunity, and a new ko begins or threatens to come about while the ko threat situation remains the same, you will be glad you only played your smallest effective threat.

  • Otherwise, play the biggest one you have.
    • tderz: This "otherwise" refers to "If you have (enough) ko threats (bigger than the ko)".

Hence, this describes the situation where you have less ko threats being worth more than V than your opponent. The author (>>) writes:

If you will lose the ko, you want to get the most you can in exchange for it.

tderz]: ... this meaning seems [30] to be that both players exchange all their ko threats until those of value > V are used up one wins the ko and the other takes the next big threat + its follow up.

I wonder, if in such a case it would not be better for the expected koloser not to use up all his threats - knowing that he will lose the ko any way - but just to take the appropriate threat, say the biggest one of those smaller than V.
Thereby the expected koloser keeps all his ko threats (as the opponent, of course), but you never know what could happen during a game.
For deciding on this consideration perhaps the most important comes into play:
do you think that you are behind (and seeking complications) or the opposite?

Winning Ko responses

I'm curious on Ko threat responses. Let's say White plays a sente move as a Ko threat, and Black must respond. Now what? Black stares, examines the Ko, examines the threat. Black could cover the peep... but he finds another way out. Black plays there instead, effectively guarding the threat; also, black has just played a dual-purpose move that resolves the local threat and also threatens a nearby White position in Sente.

Now we have three choices here.

If White now re-takes the Ko, Black has just saved the otherwise threatened group and now makes good on his threat. White now has no reason not to settle the Ko, and settles in Gote after Black has carried out his threat, leaving Black to chose the direction of play. This is as opposed to settling in Sente, where Black must carry out his threat, and White chooses where to play.

White can re-take the Ko and Black can chose not to settle the local threat, instead playing another Ko threat. If White chooses to settle here, Black takes his pick of threats and still ends in Sente, or plays a strengthening move to improve his position and allows White to chose what to save in exchange.

Strengthening is better, for example, if White cannot attack your position successfully, or cannot attack your strengthened position; and if the new position gives Black strong influence to attack something else. Strengthening could also be useful if White's settled position leaves Black taking his other threat, but then White may attack the previously threatened stone to reduce Black's influence. In any case, Black cannot take both positions and will take one, just whether he chooses or White chooses is in consideration. Black also gives White Sente in either case, barring moves that continue to threaten.

If White instead chooses to respond to the threat, Black settles the Ko without a loss elsewhere in exchange. Black has now completely won the Ko without exhausting White's Ko threats-- indeed, Black may have fewer Ko threats than White and, by a clever response, win the Ko. Of course, if the Ko is to Black's advantage, he may chose to protect his threatening stones and strengthen the position there rather than concede locally to White, as he can win the Ko.

Still, it seems to me that finding a response that both threatens your opponent and protects against the Ko threat is a brilliant and game-changing maneuver worth many, many points in Sente. I wonder about single magic-bullet plays; perhaps a larger, continuous threat-threat response would be more realistic, though, where you respond to a Ko threat by playing threats in return rather than directly defending, challenging your opponent to either retake the Ko, make good on his threat, or defend himself against your threat. A few Sente peeps does a good position make, and played correctly may give you an effective defense against a Ko threat while the Ko is still yours on the board, and then you settle, and then you use your position to both nullify the follow-up and, perhaps, extend out to destroy or claim territory on top of winning the Ko.

Ko threats can be globally strategic, ne? -- JohnMoser

Ko threat classification

moved from independent page, but probably belongs here (short+old discussion about ko threats as it is)

What is a ko threat? Depending on your definition, every tenuki or pass could be called a ko threat. More generally, one speaks of restriction threats. If passes are involved or threatened to be involved, then one calls the particular restriction fights pass fights. --RobertJasiek


BillSpight: What is the threat of a ko threat?

Not the threat to retake the ko. Any board move does that (and sometimes passes, depending on the rules). The threat is to make another local play if the opponent wins the ko (or, in complicated kos, makes another play in the ko).

Yes, some people define ko threats so that every play or virtually every play is a ko threat. But that definition is so broad as to rob the term of its meaning, isn't it?

So we see that "ko" and "ko threat" definitions depend on the purposes of stating them. From a view of move legality about each move is a ko threat; from a view of perfect play fewer moves might be ko threats. --RobertJasiek

Sometimes ordinary plays (and even passes) have one of the functions of a ko threat. I used to call these phantom ko threats, but now I call them virtual ko threats.

I classify ko threats according to function on ko threat functions.


[30] the original text was "if you lose the ko" which I interpreted as "after you have have lost the ko"

Make the initial example simpler to read for beginners

For a beginner, Ko threats come in three main flavors. The easiest to find is attacking a loosely connected wall, like a bamboo joint or knight's move pair, and it's immediately clear what the cost of ignoring the threat is. Second is an atari, or making the stronger move of a pair which kills a group. Harder to spot, and more importantly, harder for a weak opponent to gauge. Finally a shape threat, which is the hardest to master, and the hardest to judge the value of.

And that doesn't even begin to cover the idea of using a similarly valued ko threat rather than using the strongest threat you have all the time, and considering possible threats as "koins" which you're using to play silent auctions with.

But I digress. I'd really like to see some simple ko threat boards to better illustrate the idea with threats strong enough for beginners to grasp the idea, because this is an important aspect that is difficult to learn, but rarely appears on a smaller board. ~~~~

unkx80: I agree that the board in the current version looks quite complicated for a beginner. So I decided to reintroduce the example presented in an earlier version of this page.


KoThreat/Discussion last edited by PJTraill on February 10, 2019 - 20:02
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
RecentChanges
StartingPoints
About
RandomPage
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Goproblems.com
Login / Prefs
Tools
Sensei's Library