Hanabi is a cooperative game. Your goal, as a team, is to build a fabulous fireworks show. You have to put together 5 fireworks ( 1 white, 1 red, 1 blue, 1 yellow, 1 green), by making series rising in number (1, 2, 3, 4, 5) with the same coloured cards.
However, you cannot see your own cards; you can only see the cards of your team mates.
Also, communication (and non communication) between the players is essential to Hanabi.
You can only communication with your teammates when you give them a clue (see below).
No other communication is allowed.
Your final score is the total of the top cards on the five piles:
|0-5||horrible, booed by the crowd...|
|6-10||mediocre, just a spattering of applause.|
|11-15||honourable, but will not be remembered for very long...|
|16-20||excellent, crowd pleasing.|
|21-24||amazing, will be remembered for a very long time!|
|25||legendary, everyone left speechless, stars in their eyes|
- Each card has a colour and a number.
- There are three variants that add a 6th colour: the multicolour.
- There are three difficulties:
- These cards get marked by a separate multicolour clue.
- These cards get marked by a separate multicolour clue.
- These cards get marked by any colour clue.
- 8 clue tokens.
|# Cards (official rules)||5||5||4||4|
|# Cards (unofficial rules)||6||5||4||3|
Players take turns to either:
• Give a clue
• Play a card, or
• Discard a card
- Possible when there is at least one clue token available.
- Clues indicate all cards of a certain colour, or all cards of a certain number.
you may indicate no cards of a color or number on BGA (you have no 5s, or you have no reds) you may not indicate only a single card of the matching criteria if there is more than one
To do this:
1. The active player selects a card in another player's hand.
2. The active player selects a clue option.
3. One clue token is moved from 'available' to 'used'.
1. The active player selects a card in their hand.
2. The active player selects the 'Play selected card' button.
- • If the played card fits in any of the sequences, it is placed in the appropriate colour stack on the table.
- • If the played card does not fit, it is placed in the discard pile and the team gets a misfire token.
3. A replacement card is drawn.
This is only possible when there are fewer than 8 clue tokens available.
1. The active player selects a card in their hand.
2. The active player selects the 'Discard selected card' button.
3. The discarded card is placed in the discard pile.
4. One clue token is moved from 'used' to 'available'.
5. A replacement card is drawn.
- The game can end in several ways:
- 1. All cards have been drawn from the deck. After this happens, everybody gets 1 more turn. The game ends with the score at the last turn.
- 2. Your team played all possible cards. In this case, the game ends immediately and you receive the maximum possible score.
- 3. Your team made three mistakes that caused misfires. In this case, you lose the game and get negative ten (-10) points.
- 4. Your team 'abandons' the game. Although discouraged, you receive zero (0) points.
- Clues cannot be given for the colour black.
- Black cards must be played in reverse order, from 5 to 1.
|Black Powder (k)|
After playing the last card of a colour, randomly select one of the following bonuses for immediate use:
- Gain a clue token.
- Gain a clue token and recover a life.
- Give a clue regarding a colour.
- Give a clue regarding a number.
- Shuffle a discarded card into the deck.
- Play a discarded card, if possible.
When you play with others, certain words are used to describe certain objects or situations. Here is a short list.
- A card that is highlighted by a clue.
- A card that is not clued.
- The oldest, unmarked card. The card that you have no information on and will chop from your hand.
- The newest, unmarked card. The card that you have no information on and was freshly drawn.
- To bomb
- A play that causes the third misfire-token. Sometimes called explosion
- A card that has no copy (left) in the deck.
- Double discard
- When consecutive players have the same card on chop. A special problem, since, to each of the consecutive players, a number to chop does not look like a unique save.
- Twin chop
- 2 consecutive players have the same card on chop. Same as Double discard
- Double save
- When a player has two consecutive unique cards, requiring consecutive save clues.
- A set of rules / guidelines / conventions on how to interpret actions by players (including spending of clue-tokens).
- If you are playing with ELO rating on, your ELO may be changed at the end of the game.
- Here is how it works:
- 1. Every player on the team is temporarily considered as having the average ELO rating of the team.
- 2. The system will generate a bot associated with the score your team has achieved (let’s call it Hanabot).
- • Hanabot’s ELO rating depends on the variant you are playing (50 cards, 60 cards, 60 cards multicolour), the number of players on the team and, most importantly, your team's score.
- • All Hanabot’s ELO ratings have been set by an experienced player, they are not random or simply proportional to the score/number of players.
- 3. Your team (actually your team's average ELO) will now compete against Hanabot.
- • The system will calculate your team’s ELO gain/loss as though your team had tied with Hanabot.
- • If your score is below 18 (50-card game) or below 21 (60-card game), Hanabot's ELO is always 1000.
- You can find all bots' ELO ratings here
- The 55-card variant (50 cards + 1 of each value in the sixth color) cannot be played with ELO rating on. (This is because this variant is highly dependent on draw and a team’s score may not always reflect the players’ skills.)
- If you achieve the perfect score and this should cost you ELO points, you will be considered as having beaten the bot associated with the score, so that you lose no ELO points.
- To be confirmed: I think the ELO-system has had an overhaul.
- Players can cheat at Hanabi by:
- 1. Sharing information via the chat user interface.
- 2. Deliberately abandoning games.
- This is why ranking has been disabled for this game.
- If you want to play "full information" style, meaning that you wait until you know both the colour and number of a card before you play it, you will likely be short on clue tokens in >99% of the decks.
- The logical conclusion is that you cannot play a full information style game and the team must agree on some sort of meta-information. The meta-information is coded in a convention.
- Many players have different kinds of conventions.
- Note: not all conventions are added to the BGA list. Please respect the convention the table creator wants to play.
- When different players in the same team assign different meta-information to the same clue, the game becomes unplayable.
- Ideally, all players adhere to the convention, to avoid miscommunication.
- Below, the meta-information of several conventions is explained.
- Players assume the following meta-information:
- Every marked card will be playable at some point.
- Cards with a number-clue on them must be saved to be played later on in the game
- Saved cards can be safely played when all "number-1" cards are played.
- Saved cards are played from oldest to newest.
- The newest, previously unmarked card, is the next playable card in the sequence.
- The oldest, unmarked card is safe to discard.
- If a player has a card that they are sure they will never be able to play, e.g. cards marked by a blue clue when the blue stack is already complete, these card must be discarded in priority.
- A clue that marks 0 new cards.
- A clue that tells a lie (i.e. does not conform to the meta-information above).
- A clue that marks unplayable or duplicate cards, unless it saves a card at the chop position that does not have another copy of the card in the deck or other players' hands.
- e.g. Marking 4s in another hand if your hand could have a copy of that 4.
- Sometimes, a "bad" clue cannot be avoided.
- e.g. Save a unique 3 using a number clue which also marks unplayable cards.
- e.g. A colour clue from a flamboyant must be used and it is used to mark 0 new cards.
- When a lie has been told, it needs to be corrected.
- A correction clue can never mean "play", it can only mean "discard".
- Finesse builds on the standard convention (making it a bit more complex).
- In the finesse convention, the timing of the clue gives you extra information.
- You can assume the following meta-information:
- Every card that gets marked, will be playable at some point.
- If the oldest, unmarked card gets marked with a number-clue, then the clue meant "save this/these card(s)". (It is custom to play saved cards from oldest to newest when appropriate.)
- A clue that doesn't mark the oldest card, means that the newest, previously unmarked card, is the next playable card in the sequence.
- The oldest, unmarked card is safe to discard unless that player is busy by having a "known" play or discard.
- Obviously, if a player has a card that they are sure they will never be able to play, for example a card on which is marked as 1 when all the 1s have already been played, should be discarded in priority.
- Clues are given by the last possible player.
- When the timing of a clue doesn't match your expectations, you can draw certain conclusions (see next section: special interpretations)
- When receiving a clue for the "next playable card in the sequence", the linking cards are not limited to already marked cards (like in standard convention).
- The newest, unmarked cards in every player's hand should be considered as well.
- If a marked card gets discarded it means:
- The copy of that card is already marked in someone else's hand, and the discarding player is the first to realize this; or
- The copy of that card MUST be marked in a future clue (because a save-clue will be given that will mark the copy card as collateral).
Clues you should avoid giving (bad clues)
- As a special remark: don't mark useless, or duplicate cards. So if you have a saved 4, you are forbidden from marking 4s, unless you are sure you are not holding a copy of that
Meaning of skipping a player / stealing clues in Finesse convention
- There are different approaches among finesse-convention-players about who should clue whom. So this section will be devided into sub-sections. The order does not say anything about which approach is more common on BGA. There are a lot of players following each idea.
- If you allow for skipping (without assigning information to it), a lot of potential information is destroyed. Note that some players are religious in their beliefs about their interpretation on convention, so this section gets re-edited over and over.
To summarize what we already know about finesse convention: 1. Ever marked card will be playable at some point. 2. A number-clue that marks the CHOP card is a "save this for later"-clue. 3. Every other clue means: the newest, previously unmarked card, is the next playable card. 4. If you have no known playable card (or known safe-to-discard card), your CHOP card is safe to discard. 5. A clue is given by the last possible player.
- The text within the sections are written by people who follow the idea.
a) Skipping / stealing should be avoided
- If more than 1 person are able to give the same clue, then the last possible person should give that clue.
- In finesse convention skipping (Alice clues Cat, instead of Bob; then Bob is skipped) means either:
- 1. The Bob is forbidden from giving that clue; or
- 2. The Bob has something to play.
- Note that skipping a player to give a save-clue (number-clue on the CHOP card) is NOT a finesse, and can (should) be used
- to save "nice-to-have" cards;
- to save one of the twin-CHOP cards;
- to speed up the discard of useless cards.
b) Skipping / stealing is for discard-management
The main problem with this interpretation, is that "discard management" destroys the assumption that your CHOP is safe to discard, making the convention contradict itself. There is no reason for Alice to avoid discarding her safe-to-discard CHOP card, just because she sees a safe-to-discard card in Bob's hand.
When applying this style of finesse convention, you assume that players use a bad strategy (see section Strategy) of not-saving-valuable cards. Your CHOP card might be not-unique, but nice-to-have (not playable, but almost playable). So to avoid discarding a nice-to-have card, you force people into discarding by skipping them, and leaving them without anything useful to do. Skipping/stealing is a term that is often used for a situation, where giving a clue is not left for the last possible player to do. e.g. Alice saves a card on chop of Cat, potentially leaving Bob without a known play or a useful/necessary clue to give. Skipping players is a useful move to give other players the time to play / discard / clue more cards.
When getting skipped without having a play or useful option to clue, it most likely means that your chop card is not very useful. These are valid reasons for skipping you:
- Your chop-card is trash (a copy of the card is already played/marked)
- Your chop-card is redundant: Another copy of it is on another player's (or your own) hand as well, so discarding it is no loss (or it has low priority like a high number when it's very early in the game).
- You might still have something nice on chop, but the player sitting before already received information that their chop might be something very good (from earlier discard-behaviour like NOT being skipped over by the player sitting before them on earlier turns) and they are reluctant to discard for that reason.
- The clue given is a play-clue for a card that might be in your hand so you couldn't know whether you want the other card or not. The player skipping you knows that you want it and therefore takes away the decision.
This approach sacrifices some saving-clue-tokens option (that might be achieved on a stricter dogma about who should clue whom) in order to be able to keep nice cards in the game for longer and discard useless cards instead and by that often get those cards played before they would be discarded.
- This convention is better for players, who play efficient enough to rarely struggle with clue-shortness and can therefore afford being more careful about not losing first copies.
- If a group often struggles with clue-shortness then skipping-dogma (as proposed by other conventions) might as well improve their overall results. If the problem for not achieving 30 points usually isn't clue-shortness but getting bottom-decked (or heavily delayed from losing earlier a card that could be played now), then results can get better with this discard-management convention.
Planning ahead is a vital aspect of the game. By carefully deciding on who to let clue and whom to skip, it is often possible to
- avoid twin-chops before they become an issue
- avoid not having enough clue tokens to save all necessary cards
- get the discard you were waiting for (when a player has 2 copies of the same playable card and you want them to discard before clueing it)
- For others conventions, see the previous link: this forum post
- In addition to the convention(s), there is also some strategy to the game.
- You should not confuse strategy with convention! Strategy is team vs deck; convention is communication within the team.
- A simple example of strategy is saving 2s in 2-player games.
- Using number-2-clue to communicate that those cards need to be saved, is a convention.
- More advanced players will try to avoid losing as many "first" (non-unique) cards as possible.
- It's a good strategy to keep good cards in the game as long as possible - even if they are not yet playable.
- The other copy of that card might be far down the draw pile.
- Trying to mark at least 12 cards with the first 8 clue-tokens.
- This gives 99% chance of always having enough tokens in the game.
- Not using the last two clue-tokens for "single plays".
- This avoids running out of clue-tokens in times when you need to save critical cards.