Comment by Retr0id | Hacker News Reader

Comment by Retr0id | Hacker News Reader

Everyone is missing the core game mechanic - the ability to share a "spoiler free" report of your game on social media. For example, my (human operated) game from today:

Wordle 222 3/6

⬛⬛⬛⬛⬜

⬛⬜⬛⬜⬛

⬜⬜⬜⬜⬜

However, this isn't truly spoiler free. It still leaks information. From what I posted above alone, you can infer that the last letter is probably a relatively common one.

With the help of Twitter's APIs (or in my case, a janky screen-scraping script), you can trivially gather thousands of these emojified game logs. I do not consider this cheating, because sharing these is part of the game.

You can use these, in aggregate, to perform frequency analysis. I haven't shared my code yet because I'm still working on it, but as an example, here's the output from my script for day 216:

https://twitter.com/David3141593/status/1484669351261347842

Out of all the possible words, it ranked the correct solution at rank 5 - and that's before providing a single input to the game.

Edit: Replaced green blocks with "⬜" because HN doesn't like them.

Another project that solves Wordle by reverse-engineering the emoji tweets: https://twitter.com/davidebbo/status/1482911439731912706

Code in C#: https://github.com/davidebbo/WordleReverseSolver

Damn, this is much smarter than my approach, to be honest.

Although my current implementation is O(n), compared to his O(n^2) - but he's getting much better results.

I don’t think that’s a core game mechanic, it’s the sharing mechanic - the metagame. I agree it’s also great, though!

Yet another thing that’s easily overlooked is that you don’t share a URL with a preview, as most developers would naturally think of; it’s just plain text! So it’s absolutely trivial to cheat, much simpler than extracting the word list or the solution from the code.

But why would you cheat? What’s the point?

The core mechanic is that the word is the same for everyone each day, so you can compete with friends. If you could play a different word whenever you wanted, I don't think it would have gone viral.

I enjoy the game itself and don’t have much interest in game solvers generally. But the same thought process happens when I see the emojis shared, especially now that I’m in a group where people share their results. It’s not cheating, but my temptation to gain information while playing would ruin the game for me. So my personal rule is not to look at anyone else’s emojis until I’ve completed the game for the day.

You can do much better than frequency. Just six clues you can reduce it to less than 200 and solve it 2-3 guesses

https://medium.com/@furstenheim/learning-from-the-mistakes-o...

There are a few things that I think would be really interesting to do with this data:

1. Provide a different scoring mechanism. Number of guesses is a bit simplistic. I'd like to see something where blank squares are 2 points, yellow are 1 point, and green is 0. Then a total score could be computed across all guesses, with the goal being the least number of points. It would incentivize "hard mode" I think.

2. Provide a difficulty rating of the daily word based on average number of guesses

3. Given a "spoiler free" answer, make a game out of guessing what the original player guessed.

Edit: Replaced green blocks with "⬜" because HN doesn't like them.

Thanks for clarifying, my first impression was inverted i.e. that the solid ones were "green".

> It still leaks information. From what I posted above alone, you can infer that the last letter is probably a relatively common one.

I don't get what you mean about leaking information, but you can infer that all letters are "probably" relatively common ones, that's a property of relatively common, higher probability

Let's call the last letter L4. Given no prior infomation, the probability of L4 == X, for each value of X in the alphabet, gives you a histogram - which is always the same, and can be calculated by looking at the game's dictionary wordlist.

If I give you the new information "I guessed L4 on my first guess", you can use this to update the histogram. The more-probable entries in the histogram are now even more probable, and the less-probable entries less probable.

For the specific numbers, have a look at Bayesian Inference

https://en.wikipedia.org/wiki/Bayesian_inference

oh, you meant the position contains a popular letter because that brick shade is green! I thought it was yellow

Congratulations on playing and solving Wordledles: meta Wordle!

https://twitter.com/WordleStats gathers and tweets the aggregates daily

Interesting, but discards a lot of useful information.

Reply on news.ycombinator.com