Sometimes, players will figure out asymmetric information about the identities of cards in their hand.
For example, in a 3-player game:
It is the Early Game and there are 3 clues available.
Alice discards, initiating the Mid-Game.
Bob knows that there must be nothing left to clue, or Alice would have been obligated to clue it.
Bob has three 2's in his hand that were saved with a 2 Save clue earlier on.
Bob sees that Cathy has a red 2 on chop. Since Alice did not clue it, Bob must have the other copy in his hand.
Bob has not discarded since the 2 Save was given to him earlier on. Thus, he knows that one of his three clued 2's must be the other copy of the red 2, so he marks all three cards as "possible red 2".
However, these notes are asymmetric - Alice and Bob write the notes, but Cathy does not. (Cathy has no idea that this is happening because Cathy cannot see her own red 2 on chop.)
Alice and Bob will have to keep this information desync in mind going forward.
By default, players should never assume that other players have asymmetric information. This is because having asymmetric information is the exception rather than the norm.
In the previous example, Bob might perform an action later on in the game that seems to demonstrate that he knows he has the red 2. But Cathy should not assume that he does, unless there is no other explanation for the clue.
Usually, if a 2 needs to be saved, it is not the responsibility of a player with already-clued 2's in their hand, because they could potentially violate Good Touch Principle by cluing it. So, they should defer and let someone else on the team do it. (The same general concept also applies to cluing playable cards on chop, for example.)
However, if all players have a clued 2 in their hand, then someone has to take a risk of violating Good Touch Principle in order to satisfy Save Principle. Who should do it?
In this situation, the player who has the least number of matching clued cards should take responsibility. If 2+ players are tied, then the player with the less specific type of clue on their card(s) should take responsibility.
If 2+ players have the exact same number of clued cards and type of clued cards, then either player should save the card.
Example 1:
In a 3 player game, Alice has a blue 2 on chop that needs to be saved.
Bob has 2 unknown 2's.
Cathy has 1 unknown 2.
Bob should let Cathy perform the save, since 1 matching card is less than 2 matching cards.
Example 2:
In a 3 player game, Alice has a blue 2 on chop that needs to be saved.
Bob has 2 unknown 2's.
Cathy has 2 unknown 2's.
The responsibility is shared, so Bob should perform the clue if Cathy has a good discard, or he can discard and let Cathy do it.
Example 3:
In a 3 player game, Alice has a blue 2 on chop that needs to be saved.
Bob has 2 unknown blue cards.
Cathy has 1 unknown 2's.
Bob should let Cathy perform the save, since 1 matching card is less than 2 matching cards.
Example 4:
In a 3 player game, Alice has a blue 2 on chop that needs to be saved.
Bob has 2 unknown blue cards. No blue cards have been played yet, so from Bob's perspective, each blue card has a 1/9 chance of being the blue 2.
Cathy has 2 unknown 2's. No 2's have been played yet, so from Cathy's perspective, each 2 has a 1/9 chance of being the blue 2.
The responsibility is shared, so Bob should perform the clue if Cathy has a good discard, or he can discard and let Cathy do it.
Example 5:
In a 3 player game, Alice has a blue 2 on chop that needs to be saved.
Bob has 2 unknown blue cards. No blue cards have been played yet, so from Bob's perspective, each blue card has a 1/9 chance of being the blue 2.
Cathy has 2 unknown 2's. One 2 has been played already, so from Cathy's perspective, each 2 has a 1/8 chance of being the blue 2.
Thus, it is Bob's responsibility to save the blue 2.
gives a Play Clue to the next connecting card (e.g. red 3)
In almost all cases, there is no reason to give a Selfish Clue. Players can simply play their playable card and let someone else on the team "get" the next card. Being Selfish is typically associated with having bad teamwork.
However, in rare cases, giving a Selfish Clue can be good.
For example, in a 3-player game:
All the 1's are played on the stacks.
Alice has a globally known playable red 2 in her hand.
Bob hand is as follows, from newest to oldest: blue 1, red 3, red 4, red 5, blue 1
Both Bob and Cathy has a safe discard.
Alice knows that on this turn, she is expected to play the red 2, unless she has a really good clue to give.
Alice sees that if she plays the red 2, then Bob will probably have nothing to do, and will go on to discard the blue 1.
At that point, the red 5 will be on chop, and it will be difficult for the team to get all of the red cards. Specifically, someone will probably have to give a number 3 clue to Bob as a 1-for-1 on the red 3.
Alice sees that if she gives a red clue to Bob right now (e.g. before she plays the red 2), then she will be able to "lock-in" a nice 3-for-1 clue. Even though the team will lose Tempo, this will be inconsequential because Bob has a safe discard and Cathy has a safe discard.
Alice clues red to Bob as a Selfish Play Clue.
The Selfish Finesse (A Finesse Through Your Own Hand)
One excellent reason to give a Selfish Clue is to "lock in" a Finesse before the cards move out of Finesse Position. This is called a Selfish Finesse.
Like any other Finesse, players in this situation should just mark the cards for later and wait patiently for the cluer to play their card.
For example, in a 3-player game:
It is the first turn and nothing is played on the stacks.
Alice clues red to Bob, touching a red 1 on slot 1 as a Play Clue. (This is now a globally known red 1, since Cathy does not have any red 1's in her hand.)
Bob clues red to Cathy, touching a red 3 on slot 1 as a Play Clue.
Cathy sees that Alice has as red 2 on her Finesse Position, so this must be a Selfish Finesse. Bob did not want to play the red 1 because then the Finesse might go away and he wanted to lock it in. Cathy marks her red card as potentially red 3 and discards.
Alice sees that Bob gave a Play Clue to a red 3. Bob knows for sure that he has the red 1, so Alice must have the red 2 on her Finesse Position. Alice marks her Finesse Position card as red 2 and discards.
When you have a playable card, it is usually a mistake to defer playing that card in order to "steal" a clue from another player to force them to discard.
When this occurs, it is called an Ego Clue.
For example, in a 3-player game:
Red 1 is played on the stacks.
Alice holds a globally-known blue 1.
Bob has a red 2 on chop, followed by another red 2.
Alice wants Bob to discard so that she can cleanly clue the red 2 to get it played. Thus, Alice decides to give a Play Clue to Cathy on a yellow 1, and hope that Bob has nothing else to do.
Bob discards.
Here, Alice made a mistake, because the line with better teamwork was for Alice to play the blue 1 and Bob to clue the yellow 1. Accelerating the red 2 is not as important as allowing for Bob to potentially clue Alice's hand.
Ego Clues are usually not good because the cluer is forgetting that they cannot see their own hand. In some cases, it really is important to have Bob discard as soon as possible, but it is the exception rather than the norm. Good Hanabi players know that the best lines empower their teammates, instead of forcing them to do actions that are only good from one specific perspective.
Ego Clues only apply when the cluer has a playable card. They do not apply when the cluer has a known trash card.
Ego Clues are different from Selfish Clues because Ego Clues are defined as cluing cards of a separate suit (instead of the same suit).
We have many conventions that specify what a clue means. But combined with this, we are always expected to look at the context of the game.
For example, sometimes players will receive a clue that just feels strange. If this good clue could have been given a bunch of turns ago, why is it only being given now? Were the other players not paying close enough attention, and are only getting around to cluing this card now?
In these types of situations, from High Value Principle, you can assume that the team is not making any mistakes, which means that the meaning of the clue should be altered from the "textbook" definition. We call these types of clues Context Clues because they rely on contextual reading of the game state.
The most common Context Clue is given to "stale" 1's towards the beginning of the game. For example, in a 3-player game:
After a few ordinary clues happen, Alice discards, which ends the Early Game and initiates the Mid-Game. By doing this, she implies that there is nothing to do (because players are mandated to "extinguish" all of the available Play Clues and Save Clues in the Early Game).
Bob then immediately clues Cathy about two 1's.
Normally, from Good Touch Principle, Cathy would assume that both of the 1's were "good" and play both.
However, Cathy also knows that if both of these 1's were good, then Alice was required to clue them before discarding.
Thus, Cathy can reason that one of the 1's are bad. Furthermore, by convention, she knows that the oldest (rightmost) 1 is bad, because that would be the one that she would ordinarily play first.
So, Cathy should skip over the oldest (rightmost) 1, and play the other 1. Cathy should also treat the other 1 as known-trash.
The Stale 1's Clue is specific example of a more general concept called Focus Inversion, which is covered below.
Sometimes, a Play Clue is given that both touches the chop card and one or more other card. The normal interpretation of this is to treat it as a chop-focus Play Clue, and play the chop. However, sometimes a player can know that the chop is not actually playable. This can be determined through:
explicit information - negative clues present on the card
implicit information - historical / contextual information about the card derived from moves that teammates performed earlier on in the game
When this occurs, the clue is meant to be a Play Clue on the leftmost card instead of the chop card.
Just like a "normal" Play Clue that touches multiple new cards, the chop card (and other new cards introduced) are not necessarily playable right now.
For example, in a 3-player game:
Alice clues 1's to Bob, which introduces one card on slot 1.
Bob plays the 1 and it is red 1.
Cathy clues blue to Bob, which introduces two cards on slot 1 and slot 5.
Normally, Bob would think that this is a Chop-Focus clue, and play the blue 1 from slot 5.
However, his slot 5 card has a negative 1 clue on it (from Alice's number 1 clue), so it explicitly cannot be the blue 1.
Thus, this must be Focus Inversion and blue 1 must be on slot 1. The blue card on slot 5 can be either blue 2, blue 3, blue 4, or blue 5.
Note that Self-Finesses with rank always take precedence over Focus Inversion, which is a specific case where Occam's Razor does not apply. (Self-Finesses with rank are quite common, but Self Finesses with color are nonsensical.)
If you are reading this section, you are presumably a more-experienced Hanabi player who has the fundamentals down. So, now is a good time to clear up a potential misconception that you may have about our conventions.
Up until now, you may have the impression that the group has a lot of conventions, and that if you just memorize all of the conventions, you will become a really good Hanabi player. Or, you may have the impression that the conventions are like laws and that you must always follow them. Neither of these things are true.
Context is defined as all of the meta-information about the game state. In intermediate and expert games, when determining what a clue means, players must always take into account the full context of the game. This means that figuring out what a clue means is never truly as easy as following one of the flowcharts listed on the website.
When writing the strategy documentation, it would be distracting to constantly write the word "probably" and to constantly write "this may not apply in all situations" all over the place. For advanced players, this is obvious - Hanabi is complicated enough such that every convention is meant to be broken if the particular situation demands it.
Thus, players need to read the reference documentation with a grain of salt, especially when it says the word "always". Even though this website outlines the "correct" thing to do in a lot of situations, these are not strict rules. Rather, they are intended to show the "default" or "common" case. Memorizing the default cases is really helpful, for intermediate and advanced players alike. But players have to always remember to keep their brain fully engaged on the particular situation.
Typically, players will only have one card to play at a time. In this case, unless there is some really good clue to give, there are no decisions to be made - they just play their one playable card.
What if a player has two or more playable cards to play? Which card should they play first?
If one of the cards leads into a teammate's hand, then they should play that one, because that would be good teamwork. Otherwise, they can play any card that they want.
However, if a player does not play their cards in a certain order, nothing special happens. In other words:
When you have multiple cards to play, try to think about which card would be best using common sense. And then play that one.
When other people have multiple cards to play, try to predict which one they will play. But if they play an unexpected card, do not read into it too much. They are either making a small mistake or they have a plan that is not yet visible to you.
So, players have to balance Efficiency, Tempo, and "locking in" a touch on one-away-from-playable cards.
In the Early Game, it is very common for players to give a number 1 clue when they see a hand that has two or more 1's in it. And this type of clue is normally pretty good. But because of the previous considerations, sometimes it would be better to get the 1's with several color clues instead.
Players should not just consider how good their individual clue is - they should also consider how their teammates can build on top of their clue.
For example, in a 3-player game:
It is the first turn and nothing is played on the stacks.
Bob's hand is as follows, from newest to oldest: green 4, red 1, red 3, blue 2, blue 1
Alice would like to get the red 1 and the blue 1 played from Bob's hand. (No-one else has a red 1 or a blue 1.) Should Alice clue number 1, red, or blue?
If Alice clues number 1, it would be a 2-for-1 clue. But then the team will probably spend a clue getting the blue 2. This will result in an efficiency of 3-for-2. And the red 3 will probably end up being discarded. Bad.
If Alice clues blue, it would be a 2-for-1 clue. But then the team will no longer be able to give a red clue, because it would be Chop-Focused on the red 3. So they would have to clue number 1, which would result in an efficiency of 3-for-2. Bad.
Thus, Alice should clue red first, and then clue blue later, resulting in an efficiency of 4-for-2.
As explained in the beginner's guide, it is almost always better to give a Play Clue instead of a Save Clue if a player has a playable card in their hand.
However, sometimes it is better to give a Save Clue first and then a Play Clue later.
For example, in a 3-player game:
It is the first turn and nothing is played on the stacks.
Cathy's hand is as follows, from newest to oldest: blue 4, blue 4, red 3, red 1, blue 5
Alice would like to get the red 1 played. (No-one else has a red 1 in their hand.) Alice also knows that the team needs to eventually save the blue 5.
If Alice clues number 1 first to get the red 1, then the team will need to give a 5 Save to save the blue 5, resulting in an efficiency of 2-for-2.
Instead, Alice clues number 5 as a 5 Save, which allows Bob to follow up with a red Chop-Focus Play Clue. This touches the red 3 as a 2-for-1, allowing for an overall efficiency of a 3-for-2.