[Home]Clive/SimpleHanabi

ec2-3-21-248-47.us-east-2.compute.amazonaws.com | ToothyWiki | Clive | RecentChanges | Login | Webcomic

Overview



This is a simple set of guidelines for playing Hanabi, intended to be suitable for using with relatively new players.

I am creating it because I'm about to write an... other page describing altogether scarier stuff. I don't want to give the impression that's the only way I ever play.

Hanabi is an endlessly fascinating game, which can be played in lots of ways. It needn't be daunting. However, suppose for a moment that you want to score as close to 25 points as possible in a five-colour three-person game. This means that:


(Similar calculations say 29 clues in a 2-player game, 25 clues in 4-player, 22 clues in 5-player.)

If you play completely naïvely, to play a card you need to know its rank and colour. That takes two clues. Sure, a single clue often gives information about more than one card, but not nearly often enough. And sometimes you have to spend a clue warning against a dangerous discard, leaving even fewer for identifying plays.

That is why, to score well in Hanabi, it is useful to have a few understandings between the players.

Card age



(For simplicity, I'm talking about a 4/5-player game. In 2/3-player games, you also have a fifth card, E, but these principles continue to apply in the obvious way.)

Label your cards A,B,C,D. When you play or discard and draw a new card, put that in as the new card A and shunt everything else along. Now D is your oldest card and A is your newest.

If you do this, a pair of principles can apply:


In themselves, those principles are little help. What helps is being able to act in the knowledge that other players will be following them.

For example, if someone has a 5 in their hand, but you know which card they're going to discard, you don't need to warn them until it's their discard candidate. This gives you valuable time and clues for other things.

Clues to play



So when you're given a clue about several cards, which seems to be a clue to play, you play the newest. Then what?

There are two options:


In one sense, it doesn't matter which option you pick. However, it's hugely important that players have agreed an option, otherwise they don't know when it's necessary to give a supplementary clue and when that's counterproductive.

Which option to pick isn't immediately obvious, and it's fairly well balanced, but if you play Hanabi many times, experience teaches that the second option works out better.

So: when you are clued to play several cards, play the newest. Next turn, unless anybody's told you otherwise, play the next newest and so on until you've played them all.

There is always the choice to tell someone about a card's rank or its colour when you want them to play it. Choose shrewdly, considering which is likely to leave the game in a better situation the following round.

Clues against discarding



Sometimes (all too often) you identify the risk that someone will discard a card they should not or must not. In these circumstances, you need to warn them.

Cards a player must not discard are those which reduce the maximum achievable score. At the start of the game, that means preventing 5s from being discarded. Later, when 2s, 3s and 4s have been discarded, it means preventing the second of a pair of identical cards being discarded. (If things have gone wrong though and you've lost, say, both G3s, G4 and G5 are no longer undiscardable, since they're now useless.)

From time to time you might decide a player should not discard something, perhaps because it's the 2 corresponding to your missing 1, say, or Y4 when you can see someone else is about to be clued to play Y3.

Clues against discarding present two main challenges:


Keep in mind that you know each players's discard candidate. You only clue against discarding if that card is undiscardable. However, when considering a potential clue against discarding, always think about what they'll then discard instead. Sometimes, more than one of their cards is undiscardable; be alert for this.

I suggest two simple, hard-and-fast principles:


Try to pay attention to the logic of the situation: what cards are dangerous? What cards are playable? What else have you been told previously about your hand? What cards can you see in other players' hands? Who gave the clue, and what could they see? Such considerations help in distinguishing the meaning of a clue.

If, at your discretion, you decide to warn against discarding something that's not absolutely undiscardable, be careful. Remember that warnings aren't free. And discretionary clues are easier for the recipient to misconstrue as clues to play.

Clues against discarding the discard candidate



This is the simplest kind of warning: you tell someone about their discard candidate, and they refrain from discarding it.

When a clue against discarding identifies multiple cards, treat them all as undiscardable. Your new discard candidate is the rightmost remaining card.

For example, if you hold cards w,x,y,z and are told x,z are 5s, y is now your discard candidate. After you've discarded y and drawn v (leaving v,w,x(5),z(5) in your hand), w is your next discard candidate.

If you already know your card D is a 5 and now you're told that C is a 3, look at whether there are any 3s it would be dangerous to discard (because the other one in that colour has already been discarded). If so, treat C as undiscardable and B as your new discard candidate.

Clues against discarding other cards



Suppose you know nothing about your hand and someone says "your card C is a 5". Are any 5s currently playable? If so, it's a clue to play, but what if no 5s are playable?

Why did they give that clue when you weren't going to discard C anyway? The natural conclusion is that D isn't discardable either.

So, more generally: when given a clue against discarding, it is a clue against discarding any of the cards identified by the clue, and any cards older than the oldest card you were clued against discarding.

Clues to discard



Rarely, a clue identifies cards as definitely discardable. In normal circumstances, this would be wasteful: you assume cards can be discarded anyway, unless told otherwise. Therefore, it follows that if you're told something is discardable that's because something else isn't.

When clued to discards, treat all cards older than the oldest of them as undiscardable. For example, suppose all the 1s have been played and you're told "card B is a 1". This means C and D are both undiscardable. (And notice that "card B is a 1" informs you about both of them in a single clue, where otherwise you might have needed two.)

Clues against playing



So you've clued someone to play several cards and now you want them to stop? Depending on the circumstances, there are likely several options:


Try to avoid clues which might cause confusion. Suppose I hold w,x,y,z and am told w,x are green. I play w and it's the G1. Good! I draw v to replace it and now have v,x(G),y,z. Now I'm told x,y are 3s. The W1 and W2 are already out; do I play card y or not?

Context, consistency, empathy, tempo, priorities



Here, too, some general principles apply:


Don't be afraid to discard. People will tell you if you shouldn't. Many new players feel they should "play safe" by always giving clues, but remember that on average there has to be at least one discard for every two clues given. And people expect to act on clues they're given. "Informational" clues are not merely wasteful, they're actually dangerous; people will attempt to construe them, and therefore misconstrue.

If a player clues past the person on their left ("the skipped player") to the player two to their left, this generally means one of five things:


If you give a skip clue, be aware people will see it as unusual and be alert for such circumstances. If you don't want them looking for such circumstances where none exist, don't give a skip clue. (-8

Context



Actions in Hanabi are never taken in a vacuum. When reading the situation, try to consider:


That last point is subtle but important. If the game is nearly over and the B1 is out but one B2 has been discarded and you urgently need the other, when the player to your right discards instead of cluing you that's a pretty strong indication you don't have it in your hand. Similarly, if you're told card A is a 1 and you play the G1, the fact you were clued "1" not "green" makes it unlikely card B is a G2. (The more mundane principles that you keep playing unless told not to and discard your oldest card unless warned against it are special cases of this idea.)

Consistency



Try to be consistent in how you play, especially within one game.

There will be misunderstandings. Suppose Alice takes an ambiguous action and Bob misconstrues. Now what? Two options:


Unless Alice and Bob know which option to take, they'll be like two pedestrians trying to work out which side of one another to pass. This is completely arbitrary, but my recommendation is that, at least for the rest of the game, each person continues to act consistently and people interpret their actions accordingly.

Trying to iron out the misunderstanding during the game is cheating. After the game, you can have a post-mortem and try to clarify the shared understanding between the players.

Empathy



Each player is different. When you play Hanabi with someone, you get to know them better. Regardless of any conventions, Hanabi works best when players are on the same wavelength.

If you know how bold or risk-averse another player is, how precise or vague they are, how consistently they behave, how easily flustered or riled they are, how much experience at Hanabi they have, what conventions they do or don't know, even how awake and alert they are right now, things will run more smoothly.

Also, remember each person around the table knows something different. When acting, consider what the people trying to understand your action do and don't know; when construing someone's action, consider what they do and don't know.

Tempo



Try to see not only what should be done, but when it should be done and by whom.

If somebody discards and draws, what will their hand look like afterwards? Will it be easier or more difficult to clue them usefully? (e.g. if someone starts with R1 W3 Y1 R1, perhaps you let them discard one of the R1s before cluing them.)

When two players each hold the same card, which is it better to have play it?

When that dangerous card becomes the discard candidate, will there be enough cluestones left to do something about it?

Which of their W2 and R4 is it better for them to play first?

Remember people may be waiting for you to play something, or even for you to discard something, before they can take some important action. (And remember that, if so, they have no way to tell you!)

Priorities



The highest priority is staving off disaster. If, on the basis of existing clues, the person on your left is going to mis-play or discard something they shouldn't, fix that.

Similarly if the person two to your left is in a pickle it'll take two actions to fix, take the first of them.

Beyond that, the approximate descending priority is:


Whether to clue or play is a matter of judgement, depending on how useful the clue would be and how many cluestones are left.

Dirty tricks



If in doubt, stick to what you know.

After a while, you may be tempted to try something cunning. This is only natural and part of the fun of Hanabi. However, before you do something strange, consider:


The palette of strange actions available includes:


In advanced play, there are a variety of extra conventions and techniques which codify meanings for a lot of this stuff. Initially, however, I'll divide them into two broad categories:


When a player lies, they take an action that has an obvious and clear meaning which turns out to be false. In this case, play will continue for a short while before the lie becomes apparent. At that point, players can move on to bafflement. (-8

When a player does something baffling, the first question is: did they make a mistake? This is firmly a judgement call based on empathy but it's often safer to give them the benefit of the doubt. If, however, you're pretty sure a person's deliberately done something odd:

ec2-3-21-248-47.us-east-2.compute.amazonaws.com | ToothyWiki | Clive | RecentChanges | Login | Webcomic
This page is read-only | View other revisions | Recently used referrers
Last edited May 25, 2015 5:15 pm (viewing revision 2, which is the newest) (diff)
Search: