Layered vs. reverse bluff

[Travis Hall]

Actually I missed a very important information…
My Y card was completly kwown. I already knew the number, and with the Y clue also the color.
So me and every other players know I can’t answer that Y clue.

So that’s a case covered by my caveat. You know that no card in your hand can be the next card in a sequence of plays leading up to 2y. It’s relatively rare, but it does happen.

(But notice how the clue is marking colour on a card already marked with number, presumably by a prior bluff in that position. One of the reasons this is so rare is because it gives relatively little benefit. 2y will usually be safe already, due to already being marked. That means you’re usually better off simply allowing Alice to bluff, thus saving another card. Even if Donald can’t see a good bluff for Alice, Donald may be holding one. And if there is no bluff Alice can give usually at worst Alice just clues the card directly.)

[Travis Hall]

Bob doesn’t see the missing card in Cathy’s hand so from his point of view he has it and should play it.
This is the same reasoning as a classic bluff.

Yes, Blacktango has it right here. The basic principle of bluff is that the very next person who can respond by playing the indicated card will do so. (Otherwise everyone is just guessing about who should respond, and the unpredictability often makes it impossible to clue safely.)

In Blacktango’s updated example, Alice can’t respond because it is impossible for her to be holding 1y. She knows she can’t respond to the clue, and looks around to see who can. She sees no 1y, but can see that Bob has another playable card, so lets him do it.

Unlike Alice, Bob does not know he doesn’t have 1y, and he can’t see 1y (in finesse) anywhere else, so he assumes he has 1y and responds. There difference between Alice and Bob is that Alice knows this is a bluff before any plays occur, while Bob does not.

Something I often say is: If you are asking who the (possible) bluff is for, it’s for you.

There’s another variation on this. Suppose Bob did not have a playable card in finesse, but Alice receives the same clue. Now, she knows that she can’t play 1y, but if she does nothing to reveal the bluff, Bob has to respond. How can she reveal the bluff? Only through a blind play. So Alice plays blind despite knowing that the card will not be 1y. (But if players don’t expect the first partner who can respond to do so, then this could be an attempt to bluff Cathy, and then everything falls apart. And that’s why we don’t bluff Cathy like that.)

[Romain672]

Ok I finally get the first example. Please make Alice someone who clue something, and Bob the first answer of the main move.
And use () on specific cards to make it clearer what is going on.

So for your situation, please confirm me it's that:
Played: nothing
Donald clue yellow to Alice.
Alice: x-x-x-y(2) - y2 was previously saved with 2
Bob: g1-x-x-x

And so you got two interpretations possible:
- self color bluff (hgroup's choice: https://hanabi.github.io/docs/extras/special-bluffs/ ): so if Alice don't answer, it's layered (note this is in advanced section, so above level 23)
- out of position bluff: Alice should not play, and Bob will play

Let's assume Bob got g4 in finesse position instead what happen?
- self color bluff: nothing change, Alice play her finesse position
- out of position bluff: here the clue became a self color bluff, and Alice must answer. That tell Bob his finesse position isn't playable

Now let's assume instead Cathy got a bluff target, a b2 not clued:
- self color bluff: here Donald decided to not clue b2 which is absolutely weird, he declined a free bluff. That should tend to lead for Alice to not play. The thing is, it depend of the value of the color clue. Like if there is another y2 saved, immediately that clue gain a lot of value, especially if the bluff target is instead a 4, where another copy is visible in Bob's hand.
- out of position bluff: here the clue make total sense, Alice just discard a trash card, and Bob answer it. You lost the value of the bluff. But except if it's early, you can find situations for it. You can find situations where this doesn't make sense too


I think in most cases bga assume the 'minimum' and assuming that out of position bluff look like the minimum.
But in summary, take the situation of this post. And tell me if Alice should play or not, you will get the answer of your initial question.

[Blacktango]

Since it’s the first time I read something about the self color bluff, and I never saw it any time on BGA, I wouldn’t considerate it in my situation.

I’m not sure what difference it makes between a G1 or a G4 on Bob’s finesse position (since they are playable).

We talked a lot with Donald after the game to explain each other what happened on that Y clue.
In their opinion, it should have started layered plays from Bob, which Cathy should have canceled, playing their left most card (which was a playable Bk3): a kind of reverse double bluff.
Since I thought we were not in such layered situation, I clued the Bk3 myself. So I canceled the « double bluff ».

[Blacktango]

My example can be rewritten with Alice as clue dealer :

Alice: XX XX XX XX
Bob: Y2 XX Bk2 X5 (Alice clue Y to Bob, which already had two marked 2)
Cathy: G5 XX XX XX (G5 is playable but not marked at all)
Donald: Bk3 XX XX XX (Bk3 is playable but not marked at all, but it should not matter imo)

Bob and everyone know that Bob can’t answer the clue themselves.
Since Bob can’t see anyone with the missing card, they know this is a bluff (which the player right after them will answer).
Cathy answers thinking they have the missing Y1.

[Romain672]

Yeah, you seem to be totally on that situation of self color bluff.
If you assume Bob should play, then if he doesn't play, it's layered on Cathy. (see my next post if you want to assume 'reverse double bluff' )
If you assume Bob should assume Cathy should play, then Bob only play if Cathy's finesse position isn't playable.

[Romain672]

(hard post: feel free to skip it)
I wanted to add something else about the 'reverse double bluff' (which I would name 'out of position double bluff': 'reverse' is for me for stuff after the person clued), and I think I finally find what I wanted.

Bob: ?-x-x-x
Cathy: r1-5-x-g2
Donald: b1-x-x-x
Emily: x-y2-g2-x
Alice clue yellow to Emily.

So here, the basic though is to saw a y2, and just said: oh I have y1, let's play it.

Now, if you think more about the clue, you can see that 2 is available to Emily. But Alice decided to clue yellow anyway. Why?
Maybe Alice don't want g2 but it's unlikely, and here Bob know that Alice and himself saw the other copy of g2, so 2 clue is far better. I insist since it's the main weakness of my example, but maybe Bob is holding y1-w1-x-x, and so yellow have the upside to play y1+y2 immediately. On the other hand, clueing 2 would make it ambiguous from Emily's pov and force her to delay... Even if it doesn't seem a strong enough reason.

So now, we can said that Bob is bluffed. It's a known bluff: he can't hold y1.
In hgroup, my book, and Blacktango's book, it doesn't change anything, Bob should play his slot 1.
But with this 'reverse double bluff', you could argue you are in the previous case. And so, if Bob decide to not play, it will cause Cathy to blind play, and she will think it's a layered finesse since Bob didn't answer, and will continue to play.
So now, we are in Donald's pov, if Bob's finesse position isn't playable, that's easy: Donald play to cancel the layered finesse, it make sense to me. But then Alice removed some options to Bob. Alice could have instead just clue Donald's b1, and let Bob bluff Cathy's r1 with yellow to Emily. Worst in effeciency, but immensely clearer.

If Bob's finesse position is playable (like w1), then Donald should assume a pass bluff from Bob's pov, which will play r1 during his next turn. And I believe Pass bluff should have priority on most clues of this type. And so Donald would play only if Bob's finesse position isn't playable.

The only difference with the example of the first post of Blacktango, is that the negative bluff part is done by 'strong' context, and not by negative clue. Zamiel in hgroup dislike using information not known by all players around the table since it cause desynchronisation if anything goes wrong (and can lead to 2/3 bombs/errors), which I would agree here.


But I personnally don't like much this interpretation, self color bluff look so much simpler to me.
If I want to create a new convention set, I would really like this kind of move, but that look out of place from hgroup and bga.

[Blacktango]

Interesting example.

I didn’t understand why 2 clue on Emily would be ambiguous ? If Bob has the Y1 in finesse position and plays it, why should Emily wait before playing her left most 2 ? (If Bob doesn’t have another playable 1.)

I don’t like basing bluffs on the context either. It’s not easy to read, and very hard/impossible to fix when they fail.
If you play with a lot of different players, it will be hard to read that kind of clue, especially if you don’t know them personally (are they tired, in a distracting environment, …).

I never tried the self bluff color, and it looks great because it’s more often usable than the reverse bluff.
However, I think it’s not based on any logic (it’s purely a convention) and I don’t like this.

[Romain672]

I didn’t understand why 2 clue on Emily would be ambiguous ? If Bob has the Y1 in finesse position and plays it, why should Emily wait before playing her left most 2 ? (If Bob doesn’t have another playable 1.)

2 clue is not ambiguous in my example. It's ambiguous if Bob is holding y1-w1-x-x. It could be an hidden finesse: Bob answer with y1, then Emily can't play. If Emily discard, Bob can play w1 from hesitation and then Emily will play her 2 as y2/w2. ( https://hanabi.github.io/docs/level-22/ ... blind-play )
If Emily delay, Bob can suspect w1 but not play it. Emily will play her 2 as exactly y2 next turn.

I never tried the self bluff color, and it looks great because it’s more often usable than the reverse bluff.
However, I think it’s not based on any logic (it’s purely a convention) and I don’t like this.

The basic situation is logic:

Bob: g1-x-x-y(2) - y2 was previously saved with 2
Cathy: 5-x-x-x
Alice clue yellow to Bob.

If Bob don't answer, Cathy will bomb. So to prevent the bomb, Bob has to play.
I don't know the position of bga on it, but on hgroup, it was not done for a long time, then added, and it work just fine.
If you start to add others things, it's where you have differents interpretations. But it's like most moves.

[Blacktango]

Oh, fair enough. I didn’t think about it like this.

I guess the "bluff" part of this technique’s name doesn't suit me, and didn’t help me to understand it.
Maybe I would call it "self canceling reverse" or "anti reverse" instead, or something like that, because no one is bluffed here.