evil hangman is the usual word-guessing game with a puzzle-setter who reserves the right to secretly change the sought word mid-game to another one that's still consistent with all your guesses so far. we put up a version you can try here (as the guesser) http://labs.counterwave.com/evilhangman/ there's a recent paper about evil hangman in the arxiv, too (not our work), but we simply wanted to play it, so we created this thing. -- Thane Plambeck tplambeck@gmail.com http://counterwave.com/
That's fun! But I think there's a bug. On length 6, guess eariotnslcm. It says there are two words left, presumably "humbly" and "dumbly". Guess bly, still two words left. Why if I guess 'h' does it not say "humbly" is wrong and leave 1 word left? On Mon, May 4, 2020 at 10:35 PM Thane Plambeck <tplambeck@gmail.com> wrote:
evil hangman is the usual word-guessing game with a puzzle-setter who reserves the right to secretly change the sought word mid-game to another one that's still consistent with all your guesses so far.
we put up a version you can try here (as the guesser)
http://labs.counterwave.com/evilhangman/
there's a recent paper about evil hangman in the arxiv, too (not our work), but we simply wanted to play it, so we created this thing.
-- Thane Plambeck tplambeck@gmail.com http://counterwave.com/ _______________________________________________ math-fun mailing list math-fun@mailman.xmission.com https://mailman.xmission.com/cgi-bin/mailman/listinfo/math-fun
-- Mike Stay - metaweta@gmail.com http://math.ucr.edu/~mike https://reperiendi.wordpress.com
There is a good video about hangman at https://www.youtube.com/watch?v=le5uGqHKll8. On Tue, May 5, 2020 at 12:28 PM Mike Stay <metaweta@gmail.com> wrote:
That's fun! But I think there's a bug. On length 6, guess eariotnslcm. It says there are two words left, presumably "humbly" and "dumbly". Guess bly, still two words left. Why if I guess 'h' does it not say "humbly" is wrong and leave 1 word left?
On Mon, May 4, 2020 at 10:35 PM Thane Plambeck <tplambeck@gmail.com> wrote:
evil hangman is the usual word-guessing game with a puzzle-setter who reserves the right to secretly change the sought word mid-game to another one that's still consistent with all your guesses so far.
we put up a version you can try here (as the guesser)
http://labs.counterwave.com/evilhangman/
there's a recent paper about evil hangman in the arxiv, too (not our
work),
but we simply wanted to play it, so we created this thing.
-- Thane Plambeck tplambeck@gmail.com http://counterwave.com/ _______________________________________________ math-fun mailing list math-fun@mailman.xmission.com https://mailman.xmission.com/cgi-bin/mailman/listinfo/math-fun
-- Mike Stay - metaweta@gmail.com http://math.ucr.edu/~mike https://reperiendi.wordpress.com
_______________________________________________ math-fun mailing list math-fun@mailman.xmission.com https://mailman.xmission.com/cgi-bin/mailman/listinfo/math-fun
Nice game, Thane! And the YouTube video is good, but I’m disappointed that he says the best strategy is the letter that occurs in the most still-possible words. I suspect that is rarely the case. With hangman, you get two kinds of information: is the guessed letter present, and if so, where does it occur. You should guess the letter that, against all the still-possible words, gets you the greatest expected information. Especially near the end, you may want to limit it to letters that do occur in at least one still-possible word, but that’s a very minor tweak. Also, there is an interesting question of whether a computer guesser can and should figure out likely “words” that are not in its dictionary, being kind to a poser who may have chosen a not-quite-real or overly obscure word. As Bernie points out, an evil Jotto poser can prolong that game. But not by terribly much. My program that played either side had a bit with each word to flag “foreign/obscure/obscene/questionable”. It used these in guessing, but never as its secret word as a poser; about half the words had the bit set, but the machine still won a lot. [ “Information Theory and the Game of JOTTO”, MIT AI Lab Memo #218, 1971, by me. ] — Mike
On May 5, 2020, at 4:16 PM, Allan Wechsler <acwacw@gmail.com> wrote:
There is a good video about hangman at https://www.youtube.com/watch?v=le5uGqHKll8.
On Tue, May 5, 2020 at 12:28 PM Mike Stay <metaweta@gmail.com> wrote:
That's fun! But I think there's a bug. On length 6, guess eariotnslcm. It says there are two words left, presumably "humbly" and "dumbly". Guess bly, still two words left. Why if I guess 'h' does it not say "humbly" is wrong and leave 1 word left?
On Mon, May 4, 2020 at 10:35 PM Thane Plambeck <tplambeck@gmail.com> wrote:
evil hangman is the usual word-guessing game with a puzzle-setter who reserves the right to secretly change the sought word mid-game to another one that's still consistent with all your guesses so far.
we put up a version you can try here (as the guesser)
http://labs.counterwave.com/evilhangman/
there's a recent paper about evil hangman in the arxiv, too (not our
work),
but we simply wanted to play it, so we created this thing.
-- Thane Plambeck tplambeck@gmail.com http://counterwave.com/ _______________________________________________ math-fun mailing list math-fun@mailman.xmission.com https://mailman.xmission.com/cgi-bin/mailman/listinfo/math-fun
-- Mike Stay - metaweta@gmail.com http://math.ucr.edu/~mike https://reperiendi.wordpress.com
_______________________________________________ math-fun mailing list math-fun@mailman.xmission.com https://mailman.xmission.com/cgi-bin/mailman/listinfo/math-fun
_______________________________________________ math-fun mailing list math-fun@mailman.xmission.com https://mailman.xmission.com/cgi-bin/mailman/listinfo/math-fun
On Tue, May 5, 2020 at 6:33 PM Mike Beeler <mikebeeler2@gmail.com> wrote:
Nice game, Thane! And the YouTube video is good, but I’m disappointed that he says the best strategy is the letter that occurs in the most still-possible words. I suspect that is rarely the case. With hangman, you get two kinds of information: is the guessed letter present, and if so, where does it occur. You should guess the letter that, against all the still-possible words, gets you the greatest expected information.
This isn't quite right, because your score isn't the number of guesses, it's the number of *incorrect* guesses. Given two guesses with the same expected information gain, you would choose the one that is more likely to be in the word. And you'd prefer a guess with slightly less expected information if it had a sufficiently higher chance of being in the word. Andy
On 4 May 2020 at 21:34, Thane Plambeck wrote:
evil hangman is the usual word-guessing game with a puzzle-setter who reserves the right to secretly change the sought word mid-game to another one that's still consistent with all your guesses so far.
I did that for "jotto" a fair bit ago: I wrote a jotto player that solved by trying the test word that'd get the most information from the number of "jots". And when it was setting the problem, it didn't pick a word, but when a guess was made, it'd find all the words that are consistent with that guess, divide them into sets by jot-results and picked a random word from the set that had the *most* possible words left. It remembered that set, but did the same decision making as the guesser made subsequent guesses. So until there was exactly ONE word that was consistent with all the guesses and all the jot-answers the program had given, it played on.. when there was only one left it said "you got it". /Bernie\ Bernie Cosell bernie@fantasyfarm.com -- Too many people; too few sheep --
It's also fun to pick a word and try to force it to appear. On Mon, May 4, 2020 at 10:35 PM Thane Plambeck <tplambeck@gmail.com> wrote:
evil hangman is the usual word-guessing game with a puzzle-setter who reserves the right to secretly change the sought word mid-game to another one that's still consistent with all your guesses so far.
we put up a version you can try here (as the guesser)
http://labs.counterwave.com/evilhangman/
there's a recent paper about evil hangman in the arxiv, too (not our work), but we simply wanted to play it, so we created this thing.
-- Thane Plambeck tplambeck@gmail.com http://counterwave.com/ _______________________________________________ math-fun mailing list math-fun@mailman.xmission.com https://mailman.xmission.com/cgi-bin/mailman/listinfo/math-fun
-- Mike Stay - metaweta@gmail.com http://math.ucr.edu/~mike https://reperiendi.wordpress.com
participants (6)
-
Allan Wechsler -
Andy Latto -
Bernie Cosell -
Mike Beeler -
Mike Stay -
Thane Plambeck