If anything in this doc is unclear, please just ask on the Hanabi Central discord or edit it yourself!
Reactor is a conventional framework oriented around giving, receiving, and taking safe actions. Players expect to receive safe actions from clues if they are the next player without one. When players receive information, the primary type of information they expect to receive is actionable information. While every system has to provide actions to a degree, Reactor implementations take that idea to a limit. Reactor is built with the belief that in most gamestates, immediately actionable information is significantly better than inactionable information.
Reactor began as an attempt to tackle the Winter 2022 Holiday Mix Hanabi competition. Referential Sieve originally motivated Reactor: one perspective is that Reactor 1.0 transforms Referential Sieve to further appeal to players who love giving playful positionals.
We begin by discussing basic and advanced conventions in 3-player games.
We then discuss conventions that change in 4-, 5-, and 6-player games and additional conventions that apply in certain Hanabi variants.
Finally, we include convention proposals and conventions not used by everyone.
A card is said to be 100 percent playable if its empathy only contains playable cards.
A card is obviously playable if it was either signaled to be a play by a stable or reactive clue, or 100 percent playable, or gotten by gentleman/sarcastic discard.
A card is globally known (globally known trash/ globally known play) if its identity (its trashiness/ its playability) can be determined with the help of targeting priority by all the players.
A good touched card is a card whose empathy only contains playable cards and trash cards of the standard suits (the standard suits are red, yellow, green, blue, purple, and teal).
The lock card is the rightmost unclued card in someone's hand.
Alice is the player whose turn it is. The players who play after her, in order, are then Bob, Cathy, Donald, and Emily. (Some players also use Zelda to refer to the player right before Alice.)
A reverse reactive clue refers to a clue from Alice to Bob when Bob is loaded. (Since Bob cannot be the reactor, Alice's clue to Bob must be reactive.) A scenario that might happen is Bob playing the card he's loaded with and Cathy reacting to Alice's clue.
Most of the time, Bob will be the reactor so we will use Bob/reactor and Cathy/receiver interchangeably in the sections below unless specified otherwise.
Basic conventions from Referential Sieve are required to play Reactor 1.0.
Stable clues are either lock clues or one of the following types of Referential Sieve style clues:
In reactor, a playable rank clue plays only promises that the leftmost card is obviously playable.
Unlike in referential sieve, any unloaded rank clue which touches the reactor's lock card is a lock clue. The reactor should not discard unless they receive another instruction while locked.
(This is desirable because if we really wanted Bob to discard something, we would often have ways of signaling that with a reactive clue instead.)
A rank clue to Bob's lock card from a locked player or a player at 8 clues is still a lock clue.
Reactive clues are among the hardest parts of Reactor. We break them down into three pieces and cover targeting priority, important details (concerning duplicated cards and chop resets), and bad stable clues.
In a three-player game, each reactive clue provides one safe action to both the reactor and the receiver.
To give a reactive clue, add together the slots of the two actions you're signaling, and give a clue to the receiver focusing that slot number (modulo 5).
The focus is defined in the same way as referential play clues, except that the whole hand is treated as unclued: the focus is the leftmost (not necessarily newly) touched card, except that slot 1 has the lowest precedence. This means slot 1 can only be the focus if it is the only card touched.
Bob: g4 r1 g1 g4 g5
Cathy: y2 p3 b1 y1 y4
To get Bob to play red 1 from slot 2 and Cathy to play blue 1 from slot 3, Alice would clue rank 4 to Cathy which focuses Cathy's slot 5, since the sum of the targeted slots (2 and 3) is 5.
Alice can also get Bob to play green 1 from slot 3 and Cathy to play blue 1 from slot 3. The sum of the slots (3 and 3) is 6, which itself is not a valid slot but wraps back around to slot 1. Thus Alice needs to find a clue which focus Cathy's slot 1 - a rank 2 clue to Cathy would work!
Alice can also get Bob to discard his duped green 4 on slot 1 while getting Cathy to play her blue 1 from slot 3. The sum of these slots is 4, so now Alice needs a color clue that focuses slot 4. Yellow to Cathy would do the trick, because even though yellow touches the yellow 2 on slot 1, that slot has the lowest precedence so the focus of the clue would be the next leftmost card touched, which is slot 4.
Unfortunately, it is not possible for Alice to get any two cards in Bob and Cathy's hands - like green 1 from slot 3 and yellow 1 from slot 4 for example. Alice and Bob must agree about what action in Cathy's hand a reactive clue is getting.
This is accomplished using the concept of targeting priority. It consists of 4 priority levels (from highest priority to lowest priority).
(Cathy's playable)
(Bob-Cathy finesse)
This priority level generally doesn't apply to color clues.
(Cathy's trash)
Bob targets in order:
Bob: g4 g5 g1 g2 g3
Cathy: b1 y1 y2 y3 y4
Alice gives a 1 clue to Cathy, touching both the blue 1 and the yellow 1 on slots 1 and 2 and thus focusing slot 2. From Bob's perspective, the leftmost non-obvious play in Cathy's hand is the yellow 2 on slot 3. For this reactive clue to work, Bob must play the slot which when added to slot 3 produces the focused slot of 2, or 2 - 3 = 7 - 3 = 4 (mod 5).
Bob blind plays the green 2 from slot 4. Cathy now works out which slot she must play, which is 2 - 4 = 7 - 4 = 3 (mod 5). Cathy blind plays yellow 2 from slot 3.
However, Cathy also knows from the targeting priority that the yellow 2 was the leftmost non-obviously playable card in her hand! This means that none of the clued 1s could have been playable, as otherwise Bob would have reacted by playing a different slot.
Even though Cathy has two clued 1s which look like they would be playable from empathy, everyone should mark known trash on the two clued 1s because of targeting priority.
Bob: b1 r2 g3 y3 y4
Cathy: p5 r4 b3 y5 g4
Cathy is currently locked and the team is in a difficult position while being low on clues, so we need to unlock Cathy by telling her to discard something. To determine the card to sacrifice, first determining how far away from playable each non-critical card is.
Since blue 3 and green 4 are both 2 away from playable, tiebreak rules picks the higher ranking card (green 4) to sacrifice. Therefore, Alice can clue Cathy purple (focusing slot 1) to get b1 to play from slot 1 and g4 to discard from slot 5.
The status of cards ("clued"/"100 percent playable") is determined before the reactive clue touches the receiver's cards.
For example, if the leftmost unclued trash card is revealed through the clue, or if the leftmost non-obviously playable card is revealed after the clue is given, the target does not change.
If Cathy's hand has a clued but non-obviously playable card, and another unclued copy of the same playable card to the left of the clued copy, then the unclued copy is ignored when computing the target.
The ignored card has lowest priority among playable cards above finesses.
Cathy's hand has no playable cards, but has two copies of a finessable card (i.e. a one-away card in 3p):
The ignored card has the lowest priority among finessable cards (above dupe of globally known cards).
If Cathy's hand has a non-playable and non-finessable dupe card, and another unclued copy of the same playable card to the left of the clued copy, then the clued copy is ignored when computing the target.
If Alice or Bob has a globally known playable card, additional copies of that playable card in Cathy's hand are treated as follows:
In general for more than 3 players, the status (trash or playable) of the playable duped card is chosen to maximize the number of plays. For example, for 4 players, we consider the duped card to be playable if a color clue is given.
Discard signals obtained from a stable referential discard clue are revoked when the player is given another safe action.
Discard signals obtained from a reactive clue are almost never revoked.
The only way to revoke such signals is to give a lock signal.
A stable discard clue to someone with a reactively signaled discard is simply a stall.
If Alice gives Cathy a clue when Bob is loaded and Cathy is not (so she is the reactor), Cathy should expect to interpret it as a Referential Sieve stable clue. However, if Bob sees that the clue is a bad referential sieve clue, it is always a reactive clue where Bob is the reactor - regardless of how many pending plays each player has!
Bob should react to the clue before Cathy has the chance to respond to the RS interpretation.
Specifically, the following apply:
These conventions apply to players who are either locked (every card they hold has been clued) or have 8 clue tokens on their turn.
This includes Turn 1, but not "double discard avoidance" (when the previous player discarded a now-unique card).
Sometimes a locked player has no good stable referential play or discard clues available to give. In that case, the following types of clues are permitted as stalls:
If Alice is locked, clue that Alice gives to the receiver are respected as a true Referential Sieve stable clue, unless the reactor sees any of the following:
In particular, a slot 1 discard is not a reaction in response to a clue given to Cathy by a locked player if Bob has no other safe actions. Thus locked Alice must remember that any reactive clue which depends on a slot 1 discard from Bob to work can fail and be interpreted as a simple stable clue.
Alice is allowed to lock Bob by cluing rank to his lock card even if Alice herself is locked, if the team expects one of the players to be unlocked in the near future.
The endgame is when:
Clues in the endgame follow the same conventions as in the Locked and 8 Clue Players section.
Bombs reset all information.
All players therefore are unloaded from the clue that caused the bomb (if it is isolatable). Players can re-get these cards, but can also "hope" the cards play on their own. But either way, they are never globally known after the reset.
If a player plays into a previous reaction that was suspect, they are implicitly giving permission for the entire clue to be back on.
Knowing only Basic Conventions is sufficient to play your first games using Reactor, so go play some games!
It can be useful to the team to know that there are circumstances when a player will not discard, so that e.g. instead of saving Bob's chop, Alice can play her card.
On occasion, players will defer blind-plays. This is often done by the last player in a reactive chain when they need to save their Bob's chop.
It is also possible for one of the players in the middle of a reactive chain to defer their blind-play to give a clue of their own, but then they should consider whether doing so would be
You should respond to the most recently given clue if there are multiple reactive clues given before your turn.
Reactors with a pending reaction that is a play are considered loaded.
Note that this convention does not apply to 4 or more player (See Hesitation Nullify Signals).
If a reactor sees that reactive clue does not work (induces a bomb), they must choose the next working interpretation in the targeting priority table. For example, if a playable card cannot be targeted, then the next leftmost playable card should be.
This type of clue is an illegal reactive clue as it breaks targeting priority inference. Although it is called illegal, it can still be given and Cathy must take this possibility into account clue when inferring information about her hand.
If a legal reactive clue, that does not target a different card than the prioritized one, is available and gets the same amount of plays as an illegal one, the legal one should be given first allowing Cathy to make more inferences about her hand.
When a stable rank clue is given with a rank all of whose members are known trash, it is a Referential Play Clue on the card to the left of the leftmost clued trash card (wrapping around if necessary). It does not make sense as a Referential Discard Clue, because cluing number on the card to the left as a Referential Discard Clue is more efficient.
In pinkish variants, one can clue a trash rank (usually 1 when all 1s are played) as a trash push, this additionally promises that the focused card is actual trash (i.e. not a pinkish card).
The exception is for Brown, Muddy Rainbow, and Null variants, where trash clues with rank must be respected as Direct Discard Clues unless one of the trash cards is on chop or all cards of the special suit has been accounted for, or Bob has already a known safe discard. It simply indicates to discard the touched cards.
Alice can give a reactive clue to trick Bob into playing the connecting card into Cathy's hand when he is really not, as long as Cathy knows that the card that she is supposed to play into Bob's reaction is definitely unplayable after the clue is given.
Bob: p1 p2 g1 y1 b1
Cathy: g3 b3 r3 p3 y3
Alice clues 3 to Cathy touching all five 3s, which focuses slot 2. Since the only 1 away 3 is red 3, Bob plays yellow 1 from slot 4 (4 + 3 = 7 = 2 mod 5). Cathy knows that all of her cards are unplayable after the 3 clue, however she can mark the 3 on slot 3 as exactly red 3 because it needs to have been one-away from playable in order for the bluff to have worked.
Hard elimination occurs in one of the following scenarios:
In hard variants, we do not have hard elimination for valuable non-playable 2s. Hard variants are defined as variants having an efficiency of greater than 1.42. Hard eliminations may play on their own.
Any other valuable card that is either explicitly told to discard or implicitly discarded through PTD may have soft elimination notes. Most notably, non-playable 3s and 4s can only have soft elimination notes. In practice, most "soft elimination" notes are hard in the sense that the player who has them will almost certainly play them on its own when it becomes playable.
Discards of known cards follow the same conventions as in Referential Sieve.
Note that these discards only get the card that was actually discarded (not all possible identities).
Reactor also uses the Zero Clue State from Referential Sieve, with one change: if Alice has 1 or more clues at the start of her turn, only Bob erases his ZCS note.
When pace is +1 or lower, all reactive clues which would normally be one discard + one play become double play reactive clues instead.
However, if pace is exactly +1 and there exists a clear double play reactive clue which does not look like a stable clue, then the one discard + one play interpretation still takes precedence.
4-player games can be much harder than 3-player games.
In a 4 player game, there are up to two reactors. Thus, there are three types of clues:
Since there are 4 cards in each player's hand, the sum of the slots is computed modulo 4:
The most common sources of lost games in 4-player games are not knowing who is loaded, thereby confusing a double reactive clue with a simple reactive clue, and giving ambitious finesses involving three players which do not work.
As a result, the following are cornerstone conventions for 4-player games. Please read the following conventions carefully and fully understand the examples.
1. Loaded Cathy Skips
2. Cathy's Connecting Principle
3. Ambiguous Finesses
Whenever one of the reactors defers playing or discarding into a reactive, that card is no longer gotten for reactive targeting purposes and loading purposes.
As usual, the first reactor is the first person going clockwise from Bob who is not loaded with a play. Reactive clues to the person immediately after the first reactor in clockwise order are simple reactive clues.
If Bob is not loaded and unlocked Alice clues Donald while Cathy is loaded with a play, Cathy's reaction is always skipped over and the clue becomes a simple reactive clue on Bob + Donald.
If Bob is loaded, we shift each of Bob, Cathy, Donald cyclically and apply the same principle. Thus, a clue from Alice to Bob when Bob and Donald are both loaded is a simple reactive clue on Cathy + Bob.
If only Cathy is loaded, a clue from Alice to Cathy is a simple reactive on Bob + Cathy.
If Bob and Cathy are both loaded, a clue from Alice to Cathy is a simple reactive clue on Donald + Cathy.
If Alice gives a bad stable clue, it becomes a reactive clue which ignores the loadedness of all players except for Cathy, in accordance with the Loaded Cathy Skip principle.
Rarely, Bob must also react if Bob is loaded and Alice gives a bad simple reactive clue to Cathy and Donald.
Nothing has been played on the stacks. Bob has a red 1 on slot 1 which was previously called to play.
Bob: [r1] y3 p1 y4
Cathy: p5 b1 p2 p3
Donald: y1 r5 g5 y5
Alice gives a red clue to Donald focusing slot 2. If Bob plays slot 1 as red 1, then Cathy will either discard purple 5 from slot 1 (if Bob plays something other than y1), or play purple 5 from slot 1 (if Bob plays y1) since she will interpret the slot 2 focused clue as a simple reactive between Cathy and Donald. Since either scenario is bad, Bob knows this that this is instead a double reactive clue and thus plays purple 1 from slot 3 (since 3 + 2 + 1 = 6 = 2 (mod 4)). Cathy and Donald know that since Bob did something unexpected (something other than playing the red 1) that the clue by Alice was a double reactive clue.
Note that this can only occur if Bob knows definitively that the simple reactive clue will lead to a bomb or the discard of a critical card. Otherwise, Bob has to respect a simple reactive clue between Cathy and Donald.
A double reactive clue occurs in one of the two following scenarios:
Suppose, without loss of generality, that Bob is the first reactor, Cathy is the second reactor and Donald is the receiver.
The parity of the number of plays called for by rank and color clues do not change compared to 3 players. When Alice clues Donald:
The focused slot will be the sum of the slots of three actions by Bob, Cathy, and Donald (modulo 4). When Alice gives a reactive clue to Donald, Bob will have to determine which actionable slots to "anchor to" and subtract from the total sum of slots to determine his action. This process is referred to as reactive reduction. The reduction priority table is as follows - Bob finds the first reduction in the reduction table that maximimize the amount of plays. A reduction that would lead to a higher number of plays has priority over a reduction that would lead to a lower number even it appears after in the reduction table.
One tricky situation is whenever Bob plays a card which connects to a card Donald's hand, Cathy's Connecting Principle (described in a below section) applies and may change Cathy's action to be different from what Bob originally intended for Cathy to do. It is Alice's job to make sure that both interpretations work before giving the clue.
Bob: b1 r1 p1 g1
Cathy: g2 y1 g2 g3
Donald: y1 g4 g1 b5
Donald has a playable on slot 1, so that slot is "locked in". Since y1 is being called to play, Bob will always target the leftmost g2 trash in Cathy's hand, so that card is also locked in. Alice has a choice of cluing 1, 4, or 5 to get b1 (slot 1), g1 (slot 4), or r1 (slot 2) respectively.
Bob: b1 b2 b3 b4
Cathy: g2 y1 g2 g3
Donald: y2 g4 g5 b5
Alice sees a finesse from Cathy to Donald is possible if Cathy + Donald's targeted slots sum to 3. A blue clue to Donald (focusing slot 4) would get three plays - b1 from Bob, y1 from Cathy, and y2 from Donald.
Bob: b1 b2 b3 b4
Cathy: g2 y1 g2 g3
Donald: y4 y4 g4 r3
No finesse is possible from Cathy to Donald, but Donald does have two duped y4s. The target for Donald would be the y4 on slot 1, so a good clue here would be 3 to Donald (focusing slot 4), getting b1 + y1 played and the left y4 discarded.
Bob: b1 r1 p1 g1
Cathy: g2 y1 g2 g3
Donald: b2 y4 g4 r3
Donald has no trash and no playables and a finesse from Cathy to Donald isn't available. However, Cathy has a playable y1 on slot 2 so that will be the target. Alice can then clue red to Donald focusing slot 4 which will get 3 plays - Bob subtracts the slot 2 target in Cathy's hand to get a slot sum of 2 for himself and Donald, and playing b1 from slot 1 is a possible finesse into the slot 1 b2 in Donald's hand.
After all legal interpretations that maximize the number plays have been exhausted, Bob targets the next playable/finessable card in Donald's hand and assumes that he (Bob) is playing a copy of the original target in Donald's hand. If that is not possible, Bob assumes a lower amount of plays.
Suppose Alice gives Donald a double reactive clue where Bob is signaling for Cathy to discard. If Cathy has a card that is somewhere in Donald's hand, that card takes precedence over the usual sacrifice discard criteria. So the priority for discard targeting (in order of decreasing "trashiness") in games with 4 or more players is:
This section applies to when Donald has no immediately playable cards.
Suppose Alice gives a triple play clue to Donald. If Bob's play connected with a card in Donald's hand (making it now playable), Cathy must choose that target rather than finding the first of her finesse positions which would have originally made one of Donald's cards playable.
Nothing has been played on the stacks.
Bob: b1 p1 b5 p5
Cathy: g5 g1 r1 r5
Donald: g3 b2 r3 p2
Alice gives a Green clue to Donald focusing the g3 on slot 1. Bob sees that there does not exist a finesse from Cathy to Donald, and thus must play into Donald himself. He subtracts the g1 (slot 2) target from Cathy's hand to get a sum of slots of 3. A slot 1 play of b1 could play into Donald's b2 in slot 2, so Bob blind plays slot 1.
Now Cathy has to decide if the finesse was between her and Donald or between Bob and Donald. By Cathy's Connecting Principle she sees that b1 connected with b2, so that must have been the intended reaction. Cathy thus plays g1 from slot 2 = -2 = 1 - 1 - 2 (mod 4).
Note that it up to Alice to make sure that Cathy's Connecting Principle is respected before giving a clue that could look like a Bob-Donald Finesse! As a general rule of thumb, triple play clues involving finesses in 4p are complicated and if a simple 3p reactive clue is effective then that is often preferred for clarity reasons.
Bob: b1 p1 r1 g1
Cathy: y1 g5 b5 r5
Donald: b2 y2 r3 p5
Alice clues Purple to Donald focusing slot 4. Bob sees that Cathy's y1 (slot 1) plays into Donald's y2 (slot 2), so Bob plays b1 from slot 1. However, Cathy applies Cathy's Connecting Principle to deduce that the finesse was between Bob and Donald (slot 1 + slot 1), so bombs g5 herself from slot 2.
In general, to avoid violating Cathy's Connecting Principle, Alice must ensure that one of the following holds:
Bob: b1 p1 r1 g1
Cathy: y1 r1 b5 r5
Donald: b2 y2 r3 p5
Alice clues Purple to Donald. This is the same setup as before except the green 5 has been replaced by a red 1. Bob plays blue 1 expecting Cathy to play y1 into y2, but Cathy plays r1 instead because Cathy's Connecting Principle tells her that the original finesse is between Bob and Donald.
The conflict in the previous example is that the Bob assumes Cathy-Donald finesses have priority according to the targeting priority table while Cathy follows the Cathy's Connecting Principle and the two point to different slots in Cathy's hand. However, if both of the slots turn out to be playable, then Alice can proceed with the clue as usual.
Suppose Alice gives a reactive clue and Bob sees multiple possible finesses between Cathy and Donald. How should Bob react? We call this situation an ambiguous finesse and the correct reaction from Bob is to target the finesse between Cathy and Donald that would result in Cathy playing the first possible slot in the finesse ordering 1->4->3->2.
Bob: y1 p1 g1 p5
Cathy: b5 g5 r1 b1
Donald: b2 y5 r2 r5
Alice clues Yellow to Donald focusing slot 2. Bob sees that there are two possible finesses (r1->r2 or b1->b2). He chooses the one that results in Cathy playing the first possible finesse position, so in this case slot 4 which is b1 into b2. The sum of these slots is 1 and thus Bob plays yellow 1 from slot 1.
Note that it is up to Alice to make sure that the Cathy + Donald finesse does not fail because Cathy ends up playing a card even earlier in the finesse ordering.
Bob: y1 p1 g1 p5
Cathy: b5 g5 r1 b1
Donald: b2 y5 r5 r2
6 player games follow the same principles as for 4 and 5 player games.
In a 5 player game, there could be up to three reactors.
The reduction table for 5-player is as follows:
In general, the following 4-player principles should apply, for future conventional development:
The meanings of reactive clues change significantly in "Rainbowy", "Pinkish", and "Omni" variants.
Our reactive clue values change to be based on the value of the clue given rather than the slot focused or what cards it touches.
Pinkish variants impact the reactive rank clues, while rainbowy variants impact the reactive color clues.
Pinkish and rainbowy variant rules apply to any variant with special ones or special fives with a pink or rainbow property, respectively. Both apply if an omni suit, omni special ones, or omni special 5s are present.
Additionally, the following are considered to be pinkish variants:
and the following are considered to be rainbowy variants:
In some rainbowy variants, it may be beneficial to have color clues call for two plays or two discards, and rank clues call for one play + one discard. This is up to the team's agreement before starting the game.
Examples of common variants where this may be helpful:
When swapping, turn 1 color clues to Cathy are not stable.
If there are 6 different color clues available:
In a pinkish special-ones/fives variant where a rank 1/5 clue is not possible, the 6th color has the same meaning as a rank value of 1/5 (respectively).
Otherwise, the 6th color has the same meaning as the 1st color (with a value of 1).
If there are 1 to 3 different color clues available, reactive color clues are assigned starting at the value 3 rather than 1. For example:
If there are 4 different color clues available, reactive color clues are assigned the values 1,5,3,4 for the first, second, third, fourth colors respectively.
When Bob receives a 1 clue, Bob's rightmost (newly clued) 1 is called to play if all 1s are from the starting hand otherwise Bob's leftmost (newly clued) 1 is called to play. All the other "1"s are not expected to play.
A lock clue (rank clue to Bob's rightmost unclued card) which does not immediately reveal a safe action to Bob promises the rank of the rightmost unclued card.
In a pinkish-ones variant, rank to rightmost unclued may not promise the true rank as it is permissible to lock a player with a 1 on their rightmost unclued card.
In a pinkish-fives variant, a 2 clue must be given to lock a player who has a 5 on their rightmost unclued card.
When possible, a referential discard stable clue should promise the true rank of the clued card next to the card being told to discard.
Bob: i5 i2 g3 g3 r3
However, it is permissible to disobey pink promise in the following example as rank 3 won't get g3 to discard. Alice should clue 4 or 5 to i3 depending on the situation.
Bob: i3 g3 g3 r3 r2
If the rank being clued is a trash rank then the focused card is promised to be trash inducing a trash push (or a trash cm in variants like brown).
If the rank being clued is a playable rank, then the leftmost card is promised to be playable.
One exception is when the clue touches the lock card then it is the lock card** that is promised to be playable.
Outside these two cases, the information gained by a discard pink promise may not be used to determine if a card is obviously playable.
All lock clues are turned off in Clue Starved variants.
In other words, rank clues that touch Bob's rightmost unclued card are to be treated as a normal referential discard clue rather than a lock.
For example, a rank clue touching slots 2 and 5 in Clue Starved calls for slot 3 to discard.
In brown variants, a brown clue touching the rightmost unclued card is a play clue.
Zero clue safety promise is especially important in these variants - all players should write zcsp on everyone's leftmost unclued card that was not called to play and not cm'ed by context whenever a clue brings the team down to 0 or 0.5 clues.
Most trash pushes are turned off.
If Bob has no safe actions, a brown clue that touches Bob's rightmost unclued card is a lock clue.
Most trash pushes are turned off.
If Bob has no safe actions, a red clue to Bob's rightmost unclued card is permitted as a lock clue, provided that the rightmost unclued card will not be known to be a non-muddy-rainbow card to Bob.
Red locks can be fake - if the rightmost unclued card was later revealed to be red, Bob should not retroactively interpret the lock as a stable color push.
Most trash pushes are turned off.
Null playable cards have targeting priority over other playable cards.
Nothing is played on the stacks. Alice sees the following hands:
Bob: b1 r1 p1 y5 g5
Cathy: g1 y1 u1 p4 b5
Alice gives a rank 4 clue to Cathy. By null playable priority targeting, Bob targets the null 1 over the other playable cards to the left, and thus plays his blue 1 (slot 4 - 3 = 1).
This convention does not apply to null finesses.
Null trash, also called waste, has the highest targeting priority over other trash of the same quality (i.e. cluedness).
The priority is thus: clued trash > null trash > unclued trash > known trash.
The reasoning is that null cards inhibit our ability to give effective clues the most, and thus we want them out of players' hands as soon as possible.
Trash null 1s are also considered to be null trash.
Any stable clue to Bob which does not touch any new cards or fill in any positive information is considered to be a lock on Bob. In addition, Bob marks their rightmost unclued card as a null card and treats it as permanently clued, so the next rank clue to Bob touching the card one to the left also becomes a lock.
If Alice intentionally bombs a known trash card or blind bombs chop, it is a lock on Bob's entire hand. Bob also marks their rightmost unclued card as null and permanently clued, as before.
Rank - color swaps apply to this variant with the deficient color clue convention applied (i.e. "Odd" has a discard value of 3 and Even" has a discard value of 4).
(Experimental) Instead of normal referential rank discard clues, a stable "Odd" clue means to cm one card and a stable "Even" clue means to cm two cards. All cards touched by rank stable clues are unclued (should be marked with an x), except those which have either been cm'd or touched with a color clue. In addition, permanent cms are on, which means that if a card was previously cm'd, the next cm clue skips over that cm'd card.
The START card has a higher targeting priority than all other playable cards in the same hand.
Bob: g4 g5 g1 gS g2
Cathy: y1 b5 bS b4 b3
Alice gives a 3 clue to Cathy as a reactive play clue focusing slot 5. Since bS has a higher targeting priority than y1 and b5, Bob should blind play slot 5 - 3 = 2 into this clue.
If a sacrifice discard needs to be made, the 2 and 4 of an unplayed stack are both considered to be 1 away, and the 3 is considered to be 2 away from playable.
(Experimental) The same as in Up or Down variant we want to keep more flexibility for each suit as soon as possible. That's why we propose the following idea:
Play targeting prio:
Stack: r4 x x x x
Cathy: y1 r5 x x x
In that case when reacting Bob should target r5 as a play over y1 since we want to leave more flexibility for y suit starting card depending on a draw
Stack: x x x x x
Cathy: r1 y2 r2 r5 g3
In that case when reacting Bob should target r5 over anything else since both Alice and Bob can agree that playing red suit in order r5 r1 r2... very valuable since they both can see those cards on board.
Stack: x x x x x
Cathy: r5 g4 g3 g2 r1
In that case when reacting Bob should target g2 as long as it "starts" the longest possible sequence (g2 g3 g4) in Cathy's hand.
Once all suits' starts were determined normal prio targeting rules apply.
Cards touched by 5 considered unclued.
Cards touched by 1 considered unclued.
Blue cards are considered unclued.
Red cards are considered unclued.
Example: If Bob is the receiver, the clue is stable. If Bob appears after the receiver then the clue is reverse reactive.
These conventions are turned off by default but can be turned on before any game through agreement. They might be useful for difficult variants such as double dark.
Optional Conventions lie in the middle of the advanced conventions to convention proposals/discussions spectrum.
If a card in a normal suit can only be playable or trash by empathy and targeting priority inferences, it is considered obviously playable for loading and targeting purposes.
If this convention is turned on, one can queue playable cards on top of good touched cards. Therefore, good touch cards must be play before signaled cards if they could lead into a signaled card, otherwise signaled cards must be played before good touched cards.
When a card is singled out with a hard elimination note, that card is considered obviously playable for loading and targeting purposes.
When Bob has hard elimination notes for a playable card and Alice clues a finesse to Cathy, Bob should prioritize the finesse involving the card with hard elimination notes over other finesses (in the typical 15432 order). If there are multiple such cards (very rare), play the first possible slot in the 15432 finesse ordering which corresponds to a valid elimination finesse.
This convention generally does not apply in the endgame.
When a card is called to play in Alice's hand but the identity of that card is unknown, then Alice is expected to play all of the cards (known or unknown) that were previously called to play which could lead into the new unknown playable card before playing that card.
However, if Alice has multiple obviously playable cards, the expected order of play of these cards follows the below priority table:
If Alice has multiple globally known playables and does not respect this priority table from the perspective of another player, it triggers one or more priority plays on the first such player, who is said to have been "prioritized".
The prioritized player is promised that they hold the card which Alice's card leads into, and are also promised that they may repeatedly play the rightmost possible card, starting from clued cards and continuing into the unclued cards, until they find the promised card.
Cathy knows that red 2 should play before green 3 since it has a lower rank, so Cathy knows she has been prioritized for green 4. Cathy will first play the yellow 4 from slot 5 since that card could match, and then blind play the remaining cards starting from the right until she finds green 4.
Fix clues can be given to a player who will bomb a card from a priority play at some point. This is referred to as a load clue.
The exception is if the prioritized player is Bob and Bob has no clued cards that could match the card being prioritized. In this case, Bob is expected to blind play his rightmost unclued card and nothing else. This is referred to as a priority bluff.
Cards that are called to play from priority do not load for the purposes of considering reactive targets.
In endgame, when Cathy has a non-obviously playable card, a positional discard is a considered reactive. For example, Alice discarding slot 4 does not mean play slot 4 for Bob but 2+2 or 1+3.
When Bob is loaded (and possibly Alice), both Alice and Bob can give a clue to save Cathy's chop.
The player that is the least likely to dupe has responsibility for saving that chop.
If the card is not playable, then the clue is either a self-prompt (as in rs) or a finesse/bluff on leftmost unclued card in the player's hand.
In particular, it is not a regular prompt, finesse or bluff (as in H or RS).
When a direct play clue is given with number touching only chop, it promises that the unclued card to the right is trash. Otherwise, a color clue must be given.
(White/Null): The card to the right is known to be either White/Null or trash.
Feel free to post or discuss proposals on the Hanabi Central discord or here!
If Cathy has a card that was called to play but is subsequently duped and about to bomb, a reactive color clue to Cathy will target the card that is about to bomb with priority over any playables in Cathy's hand.
The reasoning is that good cards are often sieved in from slot 1 into slot 2. After some time these cards become playable, so it makes sense that slot 2 should have higher finesse priority. This strikes a good balance between getting important finesse or just-sieved cards vs cards towards the back of the hand which also tend to be good and are sometimes stuck there for long periods of time due to targeting priority rules.
The finesse algorithm changes to the following: