![]() It uses that curve to estimate the average expected steps. The bot played a lot of games and was able to fit a curve to the plot of information vs. "Expected steps until solution" is based on bits of information. When I've tried to figure it out, it seems to be roughly 1 skill point for an increase of 0.1 in expected steps. I can't give you a formula for "x difference in expected steps = y skill points". The worst guess gets a "0" and the best guess gets a "99". Skill is always relative - every possible guess is ranked against every other possible guess in terms of "expected steps to solution", for each specific set of "remaining possible solutions". 86?Īnd why is the difference between the Skill scores so large? The seemingly small differences in expected solutions (2.2- 1.5=0.7) and expected steps to solution (1.5-1.22=0.28) translates into a whopping 23 points difference in Skill (99- 76).įor information, in sum(p(x)*log(p(x)), what probabilities are assigned to each word? Evidently, 99 is the maximum possible of both the Luck and Skill scales (the effective range of Skill seems to be about 55-99 Luck is broader, maybe 30-99), but how does the 0.7 difference of differences between "picky" and "krill" map on to 99 vs. I get that "picky" was really lucky and luckier than "krill": both words left just one possible solution, but "picky" reduced the distance between actual and expected solutions more than "krill"-difference of differences of 0.7: (2.2-1 - 1.5-1). After a first guess of "slate," my second was "picky," which narrowed the possible solutions down to one, resulting in Skill (73), Luck (86), 1.5 expected steps to solution, and 3.2 bits of info.įor comparison, the bot guessed krill (99, unknown luck, 1.22 expected steps left, 100% info gained, and 3.8 bits). Take the puzzle for "sulky" (a few days ago). I understand these intuitively, but am unsure of how the bot actually calculates a given number. ![]() ![]() I probably would have done better if the my fifth move word was not a valid guess.Ĭould you give really specific explanations-formulas, algorithms, etc.-on how "Skill," "Luck," and "Expected steps until solution," and "Info. The bot's comment on my fourth move said that there was just one possible solution and that I should solve on the next turn. ![]() The bot said the word for my fifth move was a valid guess but was not a possible solution. What skill can possibly be involved in making a choice between two letters for which there is no rational basis? If anyone can provide an answer, I would appreciate a response.Ī more accurate response to the fifth move would have been a skill score of "-" and a luck score of 25. I have noted in the past that when a player must choose between two letters without any rational basis for the choice, NO SKILL CAN BE INVOLVED, AND THE OUTCOME IS ENTIRELY A MATTER OF LUCK. On the fifth move the bot gave me a skill score of zero and a luck score of "-". The letter that I tried on the fifth move resulted in the same response, but the letter used on the sixth move resulted in a response of five greens. I reviewed the alphabet and concluded that there were only two other letters that could be the first letter to the solution, but I had no rational basis for choosing between them. The bot's response to my fourth move in puzzle 684 was a gray for the first letter and greens for the last four letters.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |