by James-Livesey on 1/26/22, 7:58 AM with 52 comments
by probablypower on 1/28/22, 4:15 PM
- Each cell can only take one of two values, therefore any guess is sufficient to gather enough information to select the right answer on the second try (as long as you understand binary).
- The act us converting base2 <-> base10 is trivial for computers, but laborious for humans. The majority of time on this game is spent doing this mental labour (or using a converter in a separate app).
- The optimal strategy is immediately apparent, so there is no reason to play more than one round. It is solved.
How Wordle compares:
- Each cell takes one of 26 values, but one can only guess 6 times, so brute force information gathering is not a vaiable strategy, per cell. However, one can potentially eliminate 5 values per round, so one can at least brute force 25 different letters with a round to spare (plus some additional hints on their positioning). The probability distribution of each of these letters varies over each space, and is conditional on what is occupied in other spaces. For example, 'C' is a middling letter, but if you have an 'H' in position 2 or a 'K' in position 5 the letter 'C' shoots up the charts. Same for a 'G' in position 5 affecting the probability of an 'N' in position 4.
- The game's domain knowledge is not base conversions, but rather the English language.This requires recall, mental filtering, and shuffling hinted letters around, which are cognitively simple but also hard enough to be rewarding. I think the dependence upon the user's vocabulary adds some uncertainty that is exciting.
- The optimal strategy is not immediately apparent. It has been solved by bots but manual strategies vary greatly and there is room for inspiration/innovation. What opening word do I use? Do I prioritise trimming values from the alphabet, or determining the location of hinted letters? Should I check for letters occurring more than once?
How the Bytle game could be adjusted to be more entertaining:
- Potential values per cell need to increase. For example, hexadecimal instead of binary.
- Avoid loading the user with the cognitive challenge of converting bases. If you show values in hexadecimal, let the user write in hexadecimal.
- The domain space needs to change. A uniform distribution where each square is uncorrelated with others is not fun. Maybe add a hint like: "is divisible by 7" to add some intrigue and dependencies between the spaces.
by kevincox on 1/28/22, 3:55 PM
Simply guess all 0's. Then wherever you were wrong guess a 1 the next time.
I can also solve it first guess 1/256 times, but that is less surprising.
by guareber8 on 1/28/22, 2:29 PM
by pimlottc on 1/28/22, 4:23 PM
by ice_cold_etoh on 1/26/22, 10:12 AM
Probably can impress your non-programmer friends?
by trhoad on 1/28/22, 4:52 PM
Got a big LOL from me. I think the irony of Bytle has been lost on the crowd here! Great work by the way.
by gen3 on 1/28/22, 4:51 PM
by edmcnulty101 on 1/28/22, 2:55 PM
by gojomo on 1/28/22, 6:29 PM
But trying actual mechanics I now see it as "ha, ha, but serious" – a fun little illustration of the different contours of binary-strings vs words, with a minigame of in-the-head decimal-binary conversion. Bravo! For people at a certain stage of learning computer science, it's perfect.
Love the FAQ too. (Sp: 'tommorow` -> `tomorrow`.)
Further along the joke/idea, you could add:
• a timer, which starts on 1st guess, for an extra competitive dimension.
• a 'random guess' button (as in Absurdle), so people don't start with the superficially-easiest-to-reason-from `0` or `255` guesses every time – & if deterministically pseudorandom per day, everyone who starts with the same 'random' guess could still compare timings. (I'd instinctively started with the same serial number as today's puzzle, just in case that was part of the joke. But the serial number could be the 'random' 1st guess!)
• a 'hard mode', ha-ha
Potential advanced variants:
• 16-, 32-, & 64-bit versions – 'Binarywordle', 'Longle', 'Longlongle', etc – join the 21st Century!
• 'Hammle', which somehow mixes in either a hamming-weight hint or guess-requirement, of either the target-number or some product of guess & target
• 'Bitwisle', which displays a randomly-chosen logic-operator over each column, & the hints returned in that slot aren't mastermind-style but some other bitwise combo of guess & target. ('advanced bitwisle': user has to figure out the operators, too; 'super bitwisle': hint squares are produced by varied mixed-operations on nearby answer-bits - any "Rocky's Boots" fans out there?)
• 'Noisle', which gives hints with some error-rate, either fixed or also part-of-the-game to estimate - maybe there's a parity/checksum bit in the message, or hints, too.
Thanks!
by MilnerRoute on 1/28/22, 3:25 PM
I put in something like 10,010,001 -- and it accepted it. This got me thinking that I was supposed to be inputting my guesses in binary, which I now realize is not how the game works.
by personjerry on 1/28/22, 2:44 PM
by dawnerd on 1/28/22, 5:24 PM
by arbitrage on 1/28/22, 4:24 PM
Not a great effort.
by prepend on 1/28/22, 5:02 PM
by compsciphd on 1/28/22, 4:25 PM
by lindwhi on 1/28/22, 3:55 PM
by umvi on 1/28/22, 5:53 PM
by raldi on 1/28/22, 3:17 PM