Any specific black powder strategies?

[Travis Hall]

b) 1bk 1y 5r 1g - same as before; two ones play and then a reclue and 5 save. It can be discussed further if an early 5 save (if no 5bk played yet) could work as both reclue 1bk & 5save.

I’m going to dismiss the idea if the early 5 save serving as a secondary clue after discussion, because in BGA’s environment, the discussion just won’t happen.

For this, I’d start with green, and then 1s. That should serve for the coloured 1s to play, and the black 1 to save. Why? That requires walking the logic a bit.

If the hand was three playable 1s, including chop, you can just clue 1s, and let the off-chop 1s play. Then you can probably finesse out the 1g when you need it, or clue it if you need it played quickly. But we’ve gone to the trouble of cluing it differently, so it’s not that. That means the black 1 is in the hand somewhere.

If the hand was 1 1p x 1g, we would wait for the 1p to be on chop before cluing the save. But we don’t wait, so it isn’t that.

So the only place for the 1p to be is on the far left (at the start; second from left if clued after the play of 1g). That makes the card second-from-left safe to play.

And that’s two clues for two plays and a save.

The 5r is fairly irrelevant here. It probably just needs to be saved, whatever happens. If it wasn’t a 5, we’d look at letting it be discarded.

[Travis Hall]

C: 1g 4r 3g 1bk - 1s clued twice immediately, and we know that the second one was playable, unless stated otherwise.

Whereas my approach would simply be to clue 1s and left the off-chop play. One clue for a play and a save.

If at the same time we have a player with a hand like D: 1b 1m 1g 1y,

I’m confused by Ankeszu’s description of play for this. Given those two hands in combination, though, it seems pretty obvious that we need to clue 1 to both hands, then warn off C with green. Three clues, five plays, save 1p and 3g, and mark a safe discard.

[Ankeszu]

But I will say that I really prefer that the signals I use be deducible.

I could say exactlu the same. I like for people to be able to deduce Bk powder strategy not because they read about it on forum but from within their experience with game without a variant - thus, the variant strategy should lie on the same basic principles. In my case it is playing 1s from right and recluing 1s if we need to stop a 1 from playing. I would say your solution - playing 1s from left and assuming a chop-clued 1 is always a specific 1/70th on the deck - is the harder to figure on your own, actually.

I suspect that Ankeszu’s approach may have some benefits that may only materialise when you can be sure that your partners are using exactly the same techniques;

...all conventions need people playing accordingly, your bk powder strategy included. It would be hard to you to save 1bk or other useful cards if you have no idea what another player is going to think about your clues, or he would discard random cards.

The attitude you described in response to a) I rather use when the troubling cards are of the same number (2m 2b x 2g , and i want to make 2b 2g plays - especially as a finesses[not bluffs]).
Imagine the hand 1m 1b x 2g, and one player clues green (to make 1m play with cosave 2g, and let 1b be finessed) but the next player has nothing useful to finesse 1b with (but wants to clue 1b in this turn), and he clues blue normally. The clue receiver sees exactly the same clues as you described: green ten blue. Or blue then green (if 1m 2b x 1g). And he should choose xm 1b x 1g explanation for the clues - because? Even in your play-1s-from-left vision, for these two 1s to be played, for clued player it looks like they could be clued by 1s for the first play to be safe, regardless of what xm is. Unless there are three clues before any play at this player: g,b,1. But then it feels hardly efficient.

b) 1bk 1y 5r 1g - same as before; two ones play and then a reclue and 5 save. It can be discussed further if an early 5 save (if no 5bk played yet) could work as both reclue 1bk & 5save.

I’m going to dismiss the idea if the early 5 save serving as a secondary clue after discussion, because in BGA’s environment, the discussion just won’t happen.

For this, I’d start with green, and then 1s. That should serve for the coloured 1s to play, and the black 1 to save. Why? That requires walking the logic a bit.

If the hand was three playable 1s, including chop, you can just clue 1s, and let the off-chop 1s play. Then you can probably finesse out the 1g when you need it Ank: you assume that if you clue 2g after few turns, the player with marked 1 suspected as black (and at least two unkwnown draws) would play the marked 1, not the newest card? I mean, it can happen, but I would say it is a bit too situation-specific to be sure that the player would assume 1g as marked 1 always., or clue it if you need it played quickly. But we’ve gone to the trouble of cluing it differently, so it’s not that. That means the black 1 is in the hand somewhere.

If the hand was 1 1p x 1g, we would wait for the 1p to be on chop before cluing the save. But we don’t wait, so it isn’t that.

So the only place for the 1p to be is on the far left (at the start; second from left if clued after the play of 1g). That makes the card second-from-left safe to play.

And that’s two clues for two plays and a save. Same as within my example. What differs is the order or plays and clues.

I like how you applied logic to the situation.
However, along with your answer to my example c, it misses one quite important point: 1bk is far less common than any other 1 in the deck, thus saving it will be needed in the starting hands less often than a typical play 1s clues. You suggest to use more clues to make plays (and potential cosaves, IMO rather unclear, and that also makes not understanding additional color clues as a playclues, which also slows the plays). I suggest to use more clues to save 1bk, which is statistically more rarely needed than the play clues for playable 1s, ergo statistically this attitude would require less clues.

C: 1g 4r 3g 1bk - 1s clued twice immediately, and we know that the second one was playable, unless stated otherwise.

Whereas my approach would simply be to clue 1s and left the off-chop play. One clue for a play and a save.
Ank: Yup, this is the case where your attitude is better - if 1bk is really at chop. For the fairer comparison, what if it isn't (1/70 vs 18/70)? How it would approach 1bk 4r 3g 1g hand? Let the 1g be discarded? Started with green (which may be beneficial if saving 3g, but how about 4g/4m, especially if we see the other 4g/4m would need to be saved on another hand due to saves?) And if green, that makes any color clue touching already marked 1, not marking a play but marking a 1 as not black - thus slowing to clue the plays, because 2g would need additional clue to be known?

If at the same time we have a player with a hand like D: 1b 1m 1g 1y,

I’m confused by Ankeszu’s description of play for this. Given those two hands in combination, though, it seems pretty obvious that we need to clue 1 to both hands, then warn off C with green. Three clues, five plays, save 1p and 3g, and mark a safe discard.

I am confused what is confusing in my description three clues (1 to C, green to C [which can be given as non-finesse only because all players can see 1g is nicer to be played at D than C; green clue followed by 1g discard], and ]after three plays of D*] 1 to D for 1y marking, unless it was played/marked somewhere else in between) for a save 1bk and 3g, one gentlemans discard 1g at C, four plays at D. I am not sure what is fifth play in your version...
*Notice that if C was B (the second player in the game), after first clue was 1 to B, B would himself clue 1s to D though.

[Travis Hall]

But I will say that I really prefer that the signals I use be deducible.

I could say exactlu the same. I like for people to be able to deduce Bk powder strategy not because they read about it on forum but from within their experience with game without a variant - thus, the variant strategy should lie on the same basic principles. In my case it is playing 1s from right and recluing 1s if we need to stop a 1 from playing. I would say your solution - playing 1s from left and assuming a chop-clued 1 is always a specific 1/70th on the deck - is the harder to figure on your own, actually.

Ah. Well, there’s a fairly fundamental difference of opinion.

Because I don’t consider playing 1s from right to be a fundamental principle. Far from it. There are many expert-ranked players on BGA who don’t do it, and (in response to questions) I have briefly explained the logic on playing 1s from the right several times, but on some of those occasions the player who asked explicitly decided not to pick up that technique.

I regard “a clue to chop where the card may be unique should be regarded as a save” to be far more fundamental than “play 1s from right”. It’s a technique developed very early on, both in communities’ and individuals’ development of Hanabi techniques. Almost nobody reaches expert rank without knowing it, because you just need a save mechanism to consistently get high scores.

So when I’ve explained Black Powder rules to other players (at expert rank or higher, because I generally haven’t been playing Black Powder with less-experienced players) the understanding that a clue of 1s that includes chop may indicate a save has been immediate. It is the new expectation right away.

And maybe then the question is asked, “how are we going to signal play-on for 1s?”, but often we just get on with the game. And I’ve had players then clue a colour to prompt play of my chop-clued 1. Or I’ve had players question why I’ve clued colour to their chop-clued 1, and I’ve just said (after the fact), “I needed to know you would play it, and I couldn’t be sure you wouldn’t read it as a save for the 1p,” and they have understood and accepted that.

You could even drop the “from the right” part, and say that the principle is to play clued all 1s by default. Even then, though, that’s not the fundamental principle here. The principles involved are Good Touch and the fact that all 1s start playable. Once all the 1s are played, those same principles tell us to play out all cards marked 2 (unless indicated otherwise - but that, too, applies to 1s).

The problem, though, is that not all 1s are playable any more. The principles have changed.

Really, a good chunk of what I’ve suggested is just a description of what I’ve been seeing in play. And then you’re telling me that something else is better, and even if it may save a token in some percentage of games, I’m just not convinced it’s enough to overcome conceptual inertia.

[Travis Hall]

However, along with your answer to my example c, it misses one quite important point: 1bk is far less common than any other 1 in the deck deck, thus saving it will be needed in the starting hands less often than a typical play 1s clues.

I recognise that the case of 1p on chop is less common than the case of any other 1 on chop. However, what we’re really talking about here is risk management. What are the chances of a particular situation occurring, what actions will our clues prompt from our partner (again, with what probability), and what are the consequences of particular actions?

If our aim is to score 30 and flamboyants are not in use, the consequence of discard of the 1p is immediate failure. (If using flamboyants, you can get the card back, but it’s still a severe blow to your chances of achieving the objective.)

Whereas a more cautious approach, considering the chop-clued 1 to be 1p until told otherwise, generally costs a clue. I think losing for wont of one token is pretty rare. There’s enough slack in the game to make it up.

Combining probabilities with assessment of the possible consequences, I’m still inclined towards my approach.

[Travis Hall]

Ankescu, I also suspect you may overrate the significance of 1/70 vs 18/70. Sure, those are the probabilities of an on-chop coloured 1 and an on-chop black 1, but they aren’t the probabilities of having to decide how to deal with each situation.

Given that a 1 is on chop, we are far more likely to want to clue it if it is black.

If a coloured 1 is on chop, there may be a duplicate visible elsewhere. If the duplicate is in the same hand, we don’t want to clue the chop; we will generally want to let it be discarded. If the duplicate is not in the same hand, it is likely to be easier to clue. Neither of these situations can arise for the 1p.

If the coloured 1 is not blocked by a card of the same colour or multi, and there is no other 1 in the hand, it is generally better to clue it by colour. That is not possible with the 1p.

Even if the 1p is colour-blocked, it can still be worth cluing by colour - x x 2g 1g is better clued with green and 2, assuming the other cards don’t interfere. Again, not so the 1p.

Basically, all these scenarios and more eat away at that 18-to-1 bent towards the coloured cards. It’s very hard to actually calculate the probability of desiring to clue 1 to an on-chop 1, but the weighting is nowhere near as heavy as 18:1.

[Travis Hall]

Ank: you assume that if you clue 2g after few turns, the player with marked 1 suspected as black (and at least two unkwnown draws) would play the marked 1, not the newest card? I mean, it can happen, but I would say it is a bit too situation-specific to be sure that the player would assume 1g as marked 1 always., or clue it if you need it played quickly.

Yes, I do think we should generally be able to count on that. That’s how most players would play it if we called for a finesse on a 2, 3, or 4 that had been number-marked by a previous play clue (and I would avoid finessing a newly-drawn match to the marked card as a result). It’s just applying the same technique to 1s.

But if you don’t think you can count on it in a specific situation or with particular players, you can just clue the colour when convenient, and you’re not that much worse off.

Also, there’s a trick I’ve been known to use in such situations where I’m not sure how the player following me will read the clue. Sometimes, I can hold on the finesse until the next player draws a playable card. Then I clue the finesse, and if they misread it, it becomes a bluff instead.

[Travis Hall]

Ank: you assume that if you clue 2g after few turns, the player with marked 1 suspected as black (and at least two unkwnown draws) would play the marked 1, not the newest card? I mean, it can happen, but I would say it is a bit too situation-specific to be sure that the player would assume 1g as marked 1 always., or clue it if you need it played quickly.

Yes, I do think we should generally be able to count on that. That’s how most players would play it if we called for a finesse on a 2, 3, or 4 that had been number-marked by a previous play clue (and I would avoid finessing a newly-drawn match to the marked card as a result). It’s just applying the same technique to 1s.

I just realised, with Black Powder, we can finesse 5s too. And I’d expect the marked 5 to be played by default (although context can tell me otherwise).

[Travis Hall]

1bk 4r 3g 1g

Clue 1 and green, in that order, in succession, without letting the player holding this play in between, if we judge that the 1g is important enough to get played. (If there’s too much else to do with early clues, it may be better to let the 1g go.)

The green clue issued before the play of the off-chop 1 tells the holder that the off-chop 1 is not safe to play (because otherwise we would let it play before cluing the green). The 3g does not play because if it was 2g, we would clue green and 2 instead.

Two clues for one play and two saves (better than two clues for one play and one save).


1bk 4r 4g 1g
1bk 4r 4m 1g

Exactly the same. The extra save provides less benefit the further from playable the card is, but it’s still not generally bad to have incidental saves.

I’m not too worried about duplicated saves in general (I find that incidental saves prevent more problems than they cause in general). But if there is a sign that 4g or 4m is going to be incidentally-saved elsewhere, that’s something to factor into the decision of whether the 1g is worth digging out.

In fact, 1bk x x 1g all play out the same, unless one of the x cards is 2g or a 1. All that changes is the number and quality of the incidental saves.

“ And if green, that makes any color clue touching already marked 1, not marking a play but marking a 1 as not black”

No, that is unnecessary. The off-chop 1s would be considered playable by the holder when clued, so the other players should allow that player to play them out. If the sequence of plays is interrupted (barring a time-sensitive reason to call for a different play - finesse for a new draw would not be considered an interruption) it indicates that the holder has reached a non-playable 1. So, cluing the chopped 1 by colour is the only colour clue required.

That is, you don’t need to mark every 1 as not-black. Just the 1 clued on chop.

[hakha3]

I agree with Ank way of playing around 1 black.

Playing 1s from right should already be a thing, which is actually pretty much just logic not really convention. There are situations where playing from right is better and there arent for playing from left, so for sake of simplicity just play 1s from right everytime, so you dont miss situations where playing ones from right matters and that it helps other players coordinate with what you will play, which is especially important when we have 1 black and we want to control so that it isnt played. Logic is pretty similar to why we discard cards from right always, especially on start, where it shouldnt matter.

How intuitive it is that reclue 1 means its black, depends on if you know that move that reclue 1 means to discard (so in cases like 1r 1r 1g x you clue 1, they play from right and then when 1r is left you reclue 1 to signal, something is wrong with this 1, without black only explanation is that its trash) and then when you know that this move is available you should be playing without worry 1s from right until you get this signal.

This method is foolproof, as we dont put 1 black at any risk at any point and it looks more effiecient and doesnt require sometimes to lose playable 1s, which seems to me like big thing