Anyone a fan of Hanabi (card/tile game)?

It’s a cooperative game in which you all work together to try to fill in sequences of cards or tiles in the middle of the table. The trick is that each player has a hand of cards, but you can’t see your own cards, you have to get information based on clues that your teammates provide. Fascinating and deep game.

Anyhow, I’m interested in chatting with other people about what conventions they use and have developed… BUT, I specifically do NOT want to just google it, because it’s fun to test and try things out, and if some group of very smart redditors using computer simulation have developed what can be proven to be optimal conventions, I don’t want to know that… at least, not yet.
Anyhow, anyone else a fan?

[Moderating]
Moved from CS to the Game Room.

I haven’t heard of it, but I will be looking into it now. Thank you for the recommendation.

Right now it’s one of my favorite games to the point where I always carry it with me just in case an opportunity arises to play. It’s easily one of the best cooperative games and the game-play is smart and unique. One fun (for me) aspect is trying out different strategies with your partner(s) and evolving together.

I’ve played it.

It can be frustrating because of the enforced limits on communication. It makes it really difficult for first timers. I was playing a co-op games this evening (The Big Book of Madness). The two people I was playing with had never played the game before. But it wasn’t a problem because they could ask me questions as we were playing.

Me too. Sorry not to provide a more substantive response - maybe in a few months!

I’m a fan of the game but I don’t really have any conventions for giving hints. I do, however, find that I adjust my strategy if I perceive that the player has a solid understanding of logic.

I, on the other hand, played this game once with somebody who had no grasp of logical deduction. He couldn’t even grasp why it was a bad idea to repeatedly shuffle his hand.

Here’s a quick overview of the conventions that my friends and I use. There are two primary ideas on which they are build (note, by the way, that we always play for total victory… so if we discard a 5 early in the game, we just lose and start over, as opposed to playing to get as many tiles out of 25 as possible):

(1) Everyone’s hand has an order, newest to oldest. (When you start, your tiles/cards are in a random order, of course, but you just pretend they are sorted newest to oldest). Everyone keeps their hand in this order, and when you discard a tile, you always discard the oldest tile that you have not been given clues about. The benefit of this is that everyone knows what tile the other players might possibly discard. So, for instance, if you have a 5 in your hand, which is a crucial tile, but it’s not currently your oldest tile, then I am 100% confident you are not going to discard it, so I don’t yet need to spend time cluing you about it. When you draw a new tile, you add it to the other side of your hand, the newest side, which is referred to as the “play side”, for reasons I will explain below.

(2) Generally speaking, clues mean “play all these tiles right now”. So what you do not do with these conventions is tell someone “hey, these three tiles are all fours”, just because that’s useful information to have, and they can remember it and gather it up along with other information and eventually make useful plays down the road. So when you indicate multiple tiles in someone’s hand, and say “these are all red” or “these are all twos”, you are instructing that player to play them all, newest to oldest, one by one. Now, the obvious objection to this is, what if the tiles are in the wrong order in the person’s hand? Or what if there are gaps, like they have 124 of reds, but the rules require that you clue all the reds? And the answer is that once the player has played at least one tile out of their “play queue”, any other clue given to that player means “also, stop playing from your current play queue”. So, if everything was working out smoothly, and Bob had 124 of Reds, Alice would say “these are reds”, pointing at all three of them. He would play the 1. Then on his next turn he would play the 2. Then before his next turn Alice would point to some other tile in his hand and say “this is green”, and he would then play that green tile, and remember that he had a red tile, but know that it was not the red 3. You can also address tiles being in the wrong order by two players both giving clues. So if Charlie’s hand contains 321 of reds, but in the backwards order, then Alice can point to them and say “these are all reds”, and then Bob points to the middle one and says “this is a 2”. Charlie then realizes that the tile closest to his play side can’t be a 1, because if it was, Bob wouldn’t have given the second clue. He knows the middle is a 2, and now he deduces that the third one must be the 1, so he plays all 3, and you got 3 plays out of 2 clues.
Putting those two principles together, you get to what is the most fun and awesome component of these conventions, which is chain reaction clues (chain reactions generally only apply in 3- or more player games).

So suppose I look at the next two player’s hands, and they are (newest to oldest) (no tiles yet played):

Player #2 R1 G3 W4 W5 W4
Player #3 G4 G2 R2 B3 B3

So the Red 1 and 2 are both there ready to be played. But it will take two clues to get them both played, won’t it? As it turns out, it will not. What I do is point to the red 2 in the third player’s hand and say “this is red”. The second player sees me do that. Well, that seems like an insane play. I’m telling the third player to play their red 2 when there are no red tiles played yet. Madness! What could explain that? What I’m doing is giving a chain reaction clue. So player #2 pulls the tile off of his play side and plays it, without ever having been given a clue at all. And, presto, it’s the red 1. A chain reaction!

Now, that might seem very unlikely to ever come up. But remember, as we draw new tiles, we always add them on the play side. So imagine a situation partway through the game, where we’re all waiting for a Green 2 to be drawn. When the green 2 is drawn, it will always be added to someone play side. At that point, as long as anyone as a Green 3 anywhere in their hand, a chain reaction will be available.
Some more examples of chain reactions:
(1) The same situation as before, but the order of the players is reversed
Player #2 G4 G2 R2 B3 B3
Player #3 R1 G3 W4 W5 W4

I can still give precisely the same clue, and if player #2 is alert, the chain reaction can still work. This is a reverse chain reaction. It’s more dangerous, because what makes the normal chain reaction work is that player 3 doesn’t even get a chance to (incorrectly) play their R2 before player 1 has played the R1. But the key is that everyone should always be keeping an eye on the play side of everyone else’s hand. If you are clued a red tile in the middle of your hand, and someone else has the R1 literally on their play side, then you probably should NOT play, because it’s almost certainly a reverse chain reaction.

(2) Player #2 R1 R2 W4 W5 W4
Player #3 G4 G2 R3 B3 B3

A multi-tile chain reaction. I tell player #3 their middle tile is red. Player #2 sees that, sees that it seems insane, and deduces that his first and second play tiles must be R1 and R2 in that order. Player 3 sees player 2 play his R1 despite not being clued, realizes a chain reaction is afoot, sees that player 2 also has R2 coming, so deduces he must have R3. Three plays with one clue!

(3) Player #2 R1 R2 W4 W5 W4
Player #3 G4 G2 R3 R4 B3

Same as previous example, but my clue to player 3 is both is red tiles, which become a play queue, but which he doesn’t even begin to play until R1 and R2 have both been played

(4)
Player #2 R1 B3 W4 W5
Player #3 R2 Y2 Y2 Y4
Player #4 G4 G2 R3 G5

A double chain reaction! I clue player #4 his red 3, and player #2 and player #3 both blind play off their play sides.

(5) Here’s where it starts to get REALLY awesome
Player #2 G1 G2 R1 B5 Y1
Player #3 Y3 Y3 R2 B4 B4

I clue player #3 his red 2, “this is red”. Player #3 sees me give that clue, and thinks to himself “ahh, yes, it’s a chain reaction, I must have the R1 here on my play side”. So he pulls the tile off his play side and plops it down. And it is NOT R1… Oh no! But, wait, it’s G1, and G1 can ALSO be played. Then it gets to player #3s turn. Remember, he was clued a tile, which normally means he’s expected to play it. But, hold up, he just saw player #2 do something really weird, just pull a tile off his play side and play it despite being given no clues whatsoever. So he realizes that the red tile he was clued must NOT be the R1, because if it was R1, there would have been no chain reaction. So he just hangs onto his red tile. Then the next time we get back to player #2, he’s thinking “well, I must have the R1 in here somewhere”, so he plays the next tile on his play side (not the newly drawn one, of course), and it’s the G2. Then on his NEXT turn he plays the R1, and then Player #3 plays the R2. Presto! Four plays with one clue!

This is called a “lying chain reaction”, and it is SO MUCH FUN.

More later.

I just played this for the first time this weekend. I found it fun but very very tense. Especially when only two players as I played it. At one point I discarded a 5 and felt miserable because I let my partner down. It also took longer than you would expect to not look at cards as you draw them. Still, I enjoyed it and would play it again.

I think it gets VASTLY better with 3 or more players.

And remember, if they didn’t warn you it was a 5, and you discarded it, it’s THEIR fault, not yours :slight_smile:

(That said, you should absolutely play with a hand-ordering convention, as mentioned in my point (1) in the very length post above… that makes it much easier, although not automatic, to avoid fatal discards.)

I’d never heard of the game but I noticed and skimmed this thread just now. I think your strategy is very clever, MaxTheVool ! As for computer simulations, I think it would be a significant effort just to implement your strategy in software. Software that would invent alternate strategies would be very difficult.

Per your request, I won’t mention what I saw when I Googled “Hanabi strategy” … except that the top Google hit takes advantage of “A player … can … reorganize their hand” to signal. The Googled “Winning Hanabi Strategy” does not use such card shuffling to telegraph everything the player knows, but, since your strategy depends on first and last cards, one can imagine a player wanting to rearrange cards based on other knowledge … and players making inferences when he does. Is this legal? If so, how do you draw a line to prevent massive signalling?

I hope no-one minds if I bump this thread rather than starting a new one. I just recently got round to acquiring this game and have played a couple of times with my wife, so 2-player only so far.

My initial thoughts are that there is a fine balance between playing the game with zero communication other than what is explicitly mentioned in the rules (which makes it really difficult), and at the other extreme developing such detailed conventions that it becomes too easy. But we’re certainly much closer to the first than the second at the moment, so it seems like it wouldn’t hurt for me to start introducing some of the ideas in this thread.

We’re pretty much at the stage that the only thing we have agreed on so far is putting known cards at one end of our hand, so as to remember what they are. I know, super-basic.

This seems like a very powerful move - without it, you’re often stuck deciding between which card to inform your partner about, knowing that they might discard the one you don’t mention (as they have no information to say otherwise). I’m going to give this one a try immediately.

Love this but think it might be a little advanced for us after only 4 games. We are developing an understanding of a much more basic version of this, whereby a clue generally means “you can play this immediately”. But that only works well if you get lucky with the cards. Without getting more sophisticated, like your version above, we’ll often be stuck with only cluing one card, which is not a good use of resources.

The rest of your post is indeed awesome but too advanced to attempt yet - I hope to return to it in future. And also, of course, play with 3/4 players.

septimus, I think even with detailed signalling conventions, the game will never be easy. Don’t forget that most people are not top Bridge players. No doubt a world-class Bridge partnership could learn how to regularly beat Hanabi within a few games, because they are used to signalling, conventions, memory etc. Most of us are not that gifted, and as in Bridge, the more conventions you try to learn, the more you will get them wrong!

I’ve actually never played with only two players. Seems like most of the really intricate fun would be missing.

I did learn an INSANELY hard variant, which my friends at work and I have been attempting… when you give the clue, you only point to the tiles, you don’t actually say what the linking characteristic is. So you just say “these two tiles” or “this tile”.

I’ve only played it once and didn’t love it. It feels too much like a solitaire game to play it with others. I’d rather just play a real solitaire game, on my own. If I’m doing something with others, I’d prefer a more real cooperative game, and I’d play something like Apocalypse Chaos, The Lost Expedition, Mechs vs. Minions, Unicornus Knights, Spirit Island, etc.

The Lost Expedition, in particular, would probably be the one that I would go for if I just wanted a small and easy-to-teach coop to carry around.