Skip to content
 
If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Lifehacker)   Math based strategy to the classic game of 'Guess Who'. Stats enthusiasts have already guessed it. Claims that you have a 96% chance of winning, as long as your opponent doesn't do the same thing. So don't show your 8yo niece the video   (lifehacker.com) divider line
    More: Interesting, Game, Big Nose Kate, Play, Prince, character's name, YouTube, whole video documents, Mark Rober's YouTube channel  
•       •       •

1027 clicks; posted to Geek » on 13 Aug 2020 at 2:42 PM (6 weeks ago)   |   Favorite    |   share:  Share on Twitter share via Email Share on Facebook



33 Comments     (+0 »)
 
View Voting Results: Smartest and Funniest
 
2020-08-13 2:31:19 PM  
Then, you finagle your way around the rules of the game by turning three questions into the one question you get per turn: "Does your character have red hair, or glasses, or a big nose?"

So, basically the winning method is to cheat.

Great plan, Ray. Real scientific.
 
2020-08-13 2:44:44 PM  
Anyone else wacky parse that as "Meth"
 
2020-08-13 2:49:20 PM  
Was the Vintage tag busy getting its QQ plates?
 
2020-08-13 2:50:27 PM  

bloobeary: Then, you finagle your way around the rules of the game by turning three questions into the one question you get per turn: "Does your character have red hair, or glasses, or a big nose?"

So, basically the winning method is to cheat.

Great plan, Ray. Real scientific.


Technically, it is just a yes or no question, the opponent doesn't answer all three
 
2020-08-13 3:00:03 PM  
"How to Beat Children"
 
2020-08-13 3:00:19 PM  
Amazing what math can do.

Remember "Deal or No Deal"?  I figured out that the average you can expect to walk away with if you play to the very end was somewhere in the neighborhood of $38,000.   Meaning, from the very start, if you total up all the amounts in play and divide by the number of cases, that's the average expected payback.

The way to win was to play until the Banker offered more than that average expected value, and then take it.
 
2020-08-13 3:01:55 PM  

dittybopper: Amazing what math can do.

Remember "Deal or No Deal"?  I figured out that the average you can expect to walk away with if you play to the very end was somewhere in the neighborhood of $38,000.   Meaning, from the very start, if you total up all the amounts in play and divide by the number of cases, that's the average expected payback.

The way to win was to play until the Banker offered more than that average expected value, and then take it.


But but but....what if I have the million?!?!?
 
2020-08-13 3:06:51 PM  

Relatively Obscure: "How to Beat Children"


cdn.shopify.comView Full Size
 
2020-08-13 3:08:38 PM  

Sasquach: dittybopper: Amazing what math can do.

Remember "Deal or No Deal"?  I figured out that the average you can expect to walk away with if you play to the very end was somewhere in the neighborhood of $38,000.   Meaning, from the very start, if you total up all the amounts in play and divide by the number of cases, that's the average expected payback.

The way to win was to play until the Banker offered more than that average expected value, and then take it.

But but but....what if I have the million?!?!?


That's the stupid way to play.

The right way is to continuously adjust the expected value with each case of new information that is opened.  If you open a few of low value cases first, the expected value of the case goes up.  When faced with a choice between a million dollars and one dollar, take the $500,000 offered, but when face between a choice of five cases, four of which are worth more than $100,000, and one that's worth $100; and you're offered $40,000, take the chance.  You'll be able to say you took the best option available with the information you had and shouldn't have any regrets.
 
2020-08-13 3:09:33 PM  

phalamir: Relatively Obscure: "How to Beat Children"

[cdn.shopify.com image 720x793]


Fark user imageView Full Size
 
2020-08-13 3:10:17 PM  

dittybopper: Amazing what math can do.

Remember "Deal or No Deal"?  I figured out that the average you can expect to walk away with if you play to the very end was somewhere in the neighborhood of $38,000.   Meaning, from the very start, if you total up all the amounts in play and divide by the number of cases, that's the average expected payback.

The way to win was to play until the Banker offered more than that average expected value, and then take it.


I disagree with that logic. None of those techniques mean anything if you only get to play once. Same goes for almost any game of odds. For example, there's really no point calculating the long-term odds if you're only going to play one hand of poker. You don't go all in on a marginal play because 1/100 times you'll score HUGE.
 
2020-08-13 3:12:35 PM  
"Is your character an American Woman?"

"Nope."

"Does your character have These Eyes?"

"No."

"Well, I'm fresh out of ideas."
 
2020-08-13 3:19:49 PM  

Russ1642: I disagree with that logic. None of those techniques mean anything if you only get to play once. Same goes for almost any game of odds. For example, there's really no point calculating the long-term odds if you're only going to play one hand of poker.


Disagree all you want. He's got the simple version but he's right - that method gives you a significant advantage versus playing completely naively.
 
2020-08-13 3:22:56 PM  
Take that shiatty 1980s board game.
 
2020-08-13 3:24:13 PM  
It's more interesting to ask what the optimal strategy is when complex questions are disallowed.

Hair color partitions the set of faces, so you can knock off at least five per turn until you're down to nine.  Then you can do a binary search as usual.  In the worst case, it takes seven turns to make a positive ID and eight to win: 24, 19, 14, 9, 5, 3, 2, 1.
 
2020-08-13 3:25:06 PM  

MusicMakeMyHeadPound: Russ1642: I disagree with that logic. None of those techniques mean anything if you only get to play once. Same goes for almost any game of odds. For example, there's really no point calculating the long-term odds if you're only going to play one hand of poker.

Disagree all you want. He's got the simple version but he's right - that method gives you a significant advantage versus playing completely naively.


But it ignores the probability that you're more likely to choose a "stupid" case than one that has any real value. 3/4 of the cases have a trivial amount of money. If you remove 8 cases, including 6 bad cases and 2 marginal cases, your expected value is now over a hundred thousand.

There is an additional factor over deciding whether to keep playing versus not to keep playing: is a random case elimination more likely to increase or decrease the expected value of the remaining cases?
 
2020-08-13 3:36:20 PM  

Sim Tree: There is an additional factor over deciding whether to keep playing versus not to keep playing: is a random case elimination more likely to increase or decrease the expected value of the remaining cases?


I'm going to confess that I don't know that specific game very well and you may be right.

The broader point is that it's not useless to calculate expected value when playing a game one time, even if it's "napkin math" that may be imprecise but able to be performed mentally under stress. 

Of course it's even more useful to calculate that all ahead of time so that you don't get stuck in something counterintuitive like Marilyn Vos Savant's Monty Hall Problem.
 
2020-08-13 3:36:39 PM  

Glorious Golden Ass: Sasquach: dittybopper: Amazing what math can do.

Remember "Deal or No Deal"?  I figured out that the average you can expect to walk away with if you play to the very end was somewhere in the neighborhood of $38,000.   Meaning, from the very start, if you total up all the amounts in play and divide by the number of cases, that's the average expected payback.

The way to win was to play until the Banker offered more than that average expected value, and then take it.

But but but....what if I have the million?!?!?

That's the stupid way to play.

The right way is to continuously adjust the expected value with each case of new information that is opened.  If you open a few of low value cases first, the expected value of the case goes up.  When faced with a choice between a million dollars and one dollar, take the $500,000 offered, but when face between a choice of five cases, four of which are worth more than $100,000, and one that's worth $100; and you're offered $40,000, take the chance.  You'll be able to say you took the best option available with the information you had and shouldn't have any regrets.


The problem is that the Banker inherently low-balls the figures.  Until they got to the higher numbers, it was always significantly less than average left.  *HOWEVER*, once you get past that average value, you should take it.  Because you still stand a significant chance of going home with much, much less.

Remember, you don't have any money invested here, it's all "free money", even if you only win $1.   If you get an offer that's past the average amount, you're ahead of the game.  I noticed that in practically every game I watched, there would be at least one offer that high or higher

Of course, if you've already knocked out all of the lower values (lucky you!), that might be a good exception to the general rule.  In that case, I might abandon the conservative strategy and go for it.

However, I also noticed that they seemed to always have contestants who wanted to push their luck as far as possible.  None of them seemed the "do the math in my head" sort of type.  So a person like you or I who might have a math-based strategy would probably never have been picked to compete anyway.
 
2020-08-13 3:42:18 PM  
Akinator?

/DNRTFA
 
2020-08-13 3:42:26 PM  

Sim Tree: MusicMakeMyHeadPound: Russ1642: I disagree with that logic. None of those techniques mean anything if you only get to play once. Same goes for almost any game of odds. For example, there's really no point calculating the long-term odds if you're only going to play one hand of poker.

Disagree all you want. He's got the simple version but he's right - that method gives you a significant advantage versus playing completely naively.

But it ignores the probability that you're more likely to choose a "stupid" case than one that has any real value. 3/4 of the cases have a trivial amount of money. If you remove 8 cases, including 6 bad cases and 2 marginal cases, your expected value is now over a hundred thousand.

There is an additional factor over deciding whether to keep playing versus not to keep playing: is a random case elimination more likely to increase or decrease the expected value of the remaining cases?


I never did the math, but 38K does make sense as a ballpark figure.  And as you said, if I've eliminated a lot of lower value cases, it seems silly to quit based on that.  If there are two 100K+ cases left and I only have to pick 1, why not continue if the next offer will be higher than 38K regardless?
 
2020-08-13 4:10:53 PM  

johnny_vegas: Akinator?

/DNRTFA


I just tested it, and nope. I tried to make him guess Joe, the curly-blond-haired guy with the glasses and he never got close in 80 questions.
 
2020-08-13 4:18:47 PM  

MagicBoris: johnny_vegas: Akinator?

/DNRTFA

I just tested it, and nope. I tried to make him guess Joe, the curly-blond-haired guy with the glasses and he never got close in 80 questions.


*shrug* Worked for me.  It guessed 'subby's mom' everytime
 
2020-08-13 4:23:05 PM  

bloobeary: Then, you finagle your way around the rules of the game by turning three questions into the one question you get per turn: "Does your character have red hair, or glasses, or a big nose?"

So, basically the winning method is to cheat.

Great plan, Ray. Real scientific.


Seriously, from the rules:

https://www.hasbro.com/common/instruc​t​/GuessWho.PDF

...ask your opponent one question per turn.  Each question must have either a "yes" or a "no" answer.  For example, you may ask: "Does your person have white hair?"

A compound "or" question would seem to be cheating.

TFA is right, getting as near to a binary search (split the population evenly) as you can is good.  The characteristic with a the greatest population that is closest to half the remaining choices is your best chance for success on average.

I would be somewhat surprised if the rule abiding optimal is anything other than whatever the characteristic is with the population that is closest to half (over or under) is afterwards.
 
2020-08-13 4:26:21 PM  
Why limit yourself to facial characteristics? I'm guessing a faster method could be found using the names - contain an e? more than 4 letters and so on
 
2020-08-13 5:41:46 PM  
I used to cheat my younger brother in Battleship all the time.
B7
-miss
E7 is a miss?
-Oh, I thought you said...it's a miss
You're telling me G7 missed?
-Wait, didnt you say E?
No *G*
-oh.....hit.
 
2020-08-13 5:43:51 PM  

monkeypapa: Why limit yourself to facial characteristics? I'm guessing a faster method could be found using the names - contain an e? more than 4 letters and so on


Faster in what way? If the number of individual characteristics is still divided evenly amongst the choices, it's just a different way to cut down the list, otherwise you're risking being on the wrong side of a lopsided distribution.

The ultimate fast way to win is to guess correctly on the first try.
 
2020-08-13 6:14:48 PM  
My brothers played Guess Who so much they stopped using the pieces and just went by memory. It was very strange to watch.
 
2020-08-13 6:31:32 PM  

hogans: "Is your character an American Woman?"

"Nope."

"Does your character have These Eyes?"

"No."

"Well, I'm fresh out of ideas."


Will you be there to shake my hand?

Will you be there to share the land?

"God, you're old."
 
2020-08-13 6:51:34 PM  
Guess Who becomes a more interesting game if you re-title it Find the Child Molester.
 
2020-08-13 9:24:38 PM  
shiat article without actually pointing out some of the better splits

Facial hair 8 yes, 16 no
Assymetric lips 10 yes, 14 no
Pursed lips 12 yes, 12 no
Hair with yellow as a principle component (yellow or orange) 10 yes, 14 no

Or just assume everyone lined up alphabetically

1.  Would your character stand  in front of a guy named Harry; 12/12 split between George and Herman
2a.  Would your character stand in front of a guy named Bob; 6/6 split between Bill and Charles
2b.  Would your character stand in front of a guy named Petros; 6/6 split between Peter and Philip
etc. etc.
Fark user imageView Full Size
 
2020-08-13 11:09:08 PM  
My sister and I used to play that. She would always pick a female character, and it was always the question I asked first.
That was 20 fewer options to deal with. Guess Who needs a massive overhaul.
 
2020-08-14 2:57:23 AM  
Fark user imageView Full Size
 
2020-08-14 3:42:29 AM  
Yeah that's garbage and I've already seen and played with better strategies.

Step one. Ignore the pictures.

Step two. Does your character have an "A" in their name.   No matter the answer it clears half the board on the first move.

Rinse and repeat for a much easier win than the BS borderline cheating of phrasing a question to have three questions and hope they don't just answer yes or no.
 
Displayed 33 of 33 comments

View Voting Results: Smartest and Funniest

This thread is closed to new comments.

Continue Farking




On Twitter



  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.