Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.java.programmer > #14994

Re: Slightly off-topic: Determining the strength of "Hangman" word.

Show all headers | View raw

On Fri, 1 Jun 2012 12:45:21 -0700 (PDT), Lew <lewbloch@gmail.com>
wrote:

>Gene Wirchenko wrote:
>> Roedy Green wrote:
>>> Daniel Pitts wrote, quoted or indirectly quoted someone who said :
>>>> The problem is that some words are false positives.  They seem ambiguous 
>>>> for the most part, but one letter absolutely gives it away.
>> >
>>> This is a messy problem. For every guessing strategy you need a
>>> different algorithm to determine difficulty. But if you have enough
>>> algorithms that each say produce a number 0..1 you can take the max as
>>> the guessability or the sum, or something in between that gives extra
>>> weight to high components (sum of cubes?)
>
>Not really. A statistical approach, that is, analysis of a large number of games
>for frequency of words chosen, number of guesses by the opponent and if 
>the opponent guessed the word, should account for strategies that include 
>picking "easy" words and how well that works.
>
>You don't even need to know the guessing strategies involved.
>
>>      Suppose someone games your choice of algorithms by choosing words
>> to give bad results by those algorithms.
>
>They'd have to know what the algorithms are, and there has to be such a 
>thing as a "bad result". A statistical approach uses facts - what actually 

     You mean like getting hanged?

>happened. This is hard to game. If a player picks a word that is easy to 
>guess, then their opponent is likely to guess it quickly. This is especially true 
>if the results from games are fed back into the learning system so that it 
>can adapt to sneaky opponents.

     I tend to play weighting my guesses by frequency ("etoainshrdlu"
and all that).  Someone could pick words that that does not work well
for.  If someone knows your strategy, they can game it.

Sincerely,

Gene Wirchenko

Back to comp.lang.java.programmer | Previous | Next — Previous in thread | Find similar | Unroll thread

Thread

Slightly off-topic: Determining the strength of "Hangman" word. Daniel Pitts <newsgroup.nospam@virtualinfinity.net> - 2012-05-31 09:26 -0700
  Re: Slightly off-topic: Determining the strength of "Hangman" word. Lew <lewbloch@gmail.com> - 2012-05-31 09:44 -0700
  Re: Slightly off-topic: Determining the strength of "Hangman" word. Roedy Green <see_website@mindprod.com.invalid> - 2012-05-31 10:43 -0700
    Re: Slightly off-topic: Determining the strength of "Hangman" word. Daniel Pitts <newsgroup.nospam@virtualinfinity.net> - 2012-05-31 11:34 -0700
      Re: Slightly off-topic: Determining the strength of "Hangman" word. Roedy Green <see_website@mindprod.com.invalid> - 2012-06-01 02:55 -0700
        Re: Slightly off-topic: Determining the strength of "Hangman" word. Gene Wirchenko <genew@ocis.net> - 2012-06-01 09:34 -0700
          Re: Slightly off-topic: Determining the strength of "Hangman" word. Roedy Green <see_website@mindprod.com.invalid> - 2012-06-01 11:58 -0700
            Re: Slightly off-topic: Determining the strength of "Hangman" word. Lew <lewbloch@gmail.com> - 2012-06-01 12:46 -0700
          Re: Slightly off-topic: Determining the strength of "Hangman" word. Lew <lewbloch@gmail.com> - 2012-06-01 12:45 -0700
            Re: Slightly off-topic: Determining the strength of "Hangman" word. Gene Wirchenko <genew@ocis.net> - 2012-06-01 13:46 -0700

csiph-web