Hey there, I'm the creator of Survivle and stumbled upon this video by chance, and I really enjoyed it. It is so amazing to me to see that survivle got this popular :) What might have also been interesting as well would have been to maximize the path to a given word -- in survivle you can reveal the target word (unless you're playing in daily mode). Anyway, thank you so much -- this is a really cool video.
Glad to see your finally getting some recognition with your Wordle videos :D - I’ve been following you for ~year, and you’ve been highly highly underrated this whole time, so I’m hopeful that these more topical videos may bring people in to see all the other wonderful stuff you’ve done. If you’re looking for a game to try to master next, I’d recommend Euclidea. It is significantly more challenging than all games you’ve covered, but the game is fun on its own, it would be a really fun mega-project, and I think you’d love it.
does survivle use the same common vs uncommon list as regular wordle? or a similar one? if so could you improve the algorithm by preferring words that are less common in general english, thereby making it so that at least you dont lose in one word.
Yes, pretty much the same list (there can be 2-3 different words that were removed, but apart from that - all the same). I was thinking about having it always choosing from the "allowed" rather than "secret" list, whenever possible, but decided against it. My ratioal is: in some cases, it is possible that the best (that is, with highest average remaining list) word would be from "secret", not from "allowed" - this way I may be shooting myself in the foot. So I went with "just whatever has the highest average". If you look at the words the bot chosses - looks like it tend to favor rare words despite implicitly instructed to do so. "Secret" words are always produce 1 extra answer (all green), so maybe this is enough for the bot to tend to choose "allowed".
Unfortunately the Absurdle explanation in the video was rather hard to follow, even though I already knew what it was. It would have been much easier to understand by showing 2 all-grey guesses up front, then saying that it's "changing the target word behind the player's back to give the least information possible", then clarifying that it actually keeps a list of all possible target words to change to.
Well, I can't say I'm too proud of how I explained it. (Even replaced it with slides during editing, as originally it was even worse). But I'm not sure about first two "all grey" guesses being a good place to start - for example, it is not always 2 all greys -- sometime even the very first guess would have 1 yellow (EXITS). Also, I find "changing the target word" notion also a bit misleading: it doesn't change it - as it has no singular target word to begin with. I feel you need to get the "remaining possible words" idea through, and how program tries to maximize the size of that. But again - I've been looking at these distributions for a few weeks now, so may be a bit biased about what would be a good way to approach the explanation.
@@GamesComputersPlay I guess the video needs to explain what Absurdle is to someone who's never heard of it, before it can explain what the bot does, so starting with "the game tries to maximize the word search space" is harder to grasp compared to "the game gives you the worst possible legal answer". The "all greys" example would just be a way to explain it quicker initially, and a not-all-greys example could show up shortly afterward.
@@zowayix You do have a point, and I will give it to you: "the game gives you the worst possible legal answer" is a better way to put it then I did. Looks likeI would not be putting "This guy explained Absurdle perfectly" onto my tombstone. :)
Thanks. I tried to play semantle when I researched Wordle spinn offs. I haven't seen pimantle, just checked it out. The graphic representation is quite fascinating. I can't say I was impressed by how that semantic things work. Maybe I was just unlucky, or had wrong expectations, but it felt really off - like the underlying model wasn't trained enough or something. Maybe I just need to play some more to get into its way of thinking.
@@GamesComputersPlay It's confusing, because it defines semantic similarity by how frequently those words co-occur in news articles. So for instance, "equality" and "racism" are extremely closely linked according to that algorithm.
@@EebstertheGreat Alright, I played some more, even got to within a hundred of the answer, I think I am starting to get the hang of it. It is quite exciting, but quite challenging too. To solve it with a bot, looks like one would have to create and train their one word2vector model, ideally more or less on the same materials that the game had. Once it's done getting to the answer should be relatively easy, I am guessing. I still need to grow a bit to tackle such a problem, I think - but it is an interesting project for the future.