donkirkby / ludiverbia Goto Github PK

In a video about solving Wordle, 3Blue1Brown mentions some tools that could be useful for simulating a vocabulary size in an AI opponent. The word frequencies can be derived from Google's ngram data, specifically the list of 1-grams, or single words. The data formats are a bit obscure, but there's a promising write up. The total counts file doesn't make sense to me, but it might not matter if we're only comparing relative frequencies.

The other tool is a sigmoid curve to have a rough boundary between known words and unknown words.

When building a list of acceptable words, try to eliminate proper nouns, possibly using NLTK tagging.

More ideas:

• sort by lower-case frequency
• lower capitalization cutoff
• look at portion followed by period, like inc. and fig.
• only count results after 1900? 1950?
• only count results before 2010? Seem to be problems with foreign words like "ils" and "figliuol" after that year. Coincidence? Also problems with capitalization, like "usa"?
• normalize frequency as % of counts for that year?
• average prevalance over many years?
• compare with other data sources?

If I click the Fill button in Firefox, it submits the form, even though the button type is not submit. I guess I have to prevent the default action for that button.

It still works if you drag all the letters into the grid.

Related: don't enable the Ready button if the player name is blank, or use a default name.

Particularly on mobile, it's too easy to accidentally hit the back button or click on a link. Use local storage to remember game state, so you can resume the game.

Make the Quizl grids resize as a fraction of the viewport.

I like Cell Tower, but I think it needs more feedback while solving. Also, the rule for reading a connected group of words is a bit weird.

Try splitting a grid of letters into "ropes", like a connected group, but without branching. Maybe add a count of the words in each row.

The firebase proof of concept looks good, but it would be helpful to share basic functions like player matching between several word games.

Come up with a name for the collection, and rename this project. Ludiverbia? Players could be ludiverbalists, like cruciverbalists.

Add the game, What's My Word. It's similar to Word Mastermind, but with a clever scoring mechanic that stretches out the learning process.

On the live coding project, recreating the React project with the latest version fixed all the warnings. See if that will work here.

Inspired by NYT Connections, I wonder if I could create an automated version, using the same data source as Pimantle. It's based on a set of semantic meaning vectors, I think.

Several games will involve dragging letters around to make words. Create a prototype.

Resources:

• react-dnd - basic drag and drop
• react-flip-move - animation for dragging and sorting
• scrabblr - demo built on react-flip-move
• react-sortable-hoc tutorial - how to drag items around a grid, looks like it's been replaced by @dnd-kit
• Daniel Weck's HTML UI - another full example

Look for inspiration from Word Games of Skill for Connoisseurs.

Keep score.

• add a guess button
• add a button to give up when player is down to ten hidden letters
• highlight winner when game ends

Implement the bet buttons.

Create private games where you can invite friends. Add an option to mark a game as public.

The Quizl grid still doesn't work with a touch interface on Android Chrome. I thought the problem was touch-action, but that didn't seem to solve it. Maybe it has something to do with the DragOverlay?

Generate grids and display a new one each day. The share button should show colour blocks for the grid, with yellow for revealed letters, green for correctly guessed letters, or red for incorrectly guessed letters. The grid should be translated a random number of rows and columns, and randomly flipped. Everyone should use the same transformation on a given day.

• generate grids
• display one each day
• track win streak and score distribution
• add a share button

A player's grid didn't show up. I think the problem was that they hit ready before they hit join.

Add an Acrostic game that presents a ten-letter word, and the player has to use those ten letters to make a word that starts with each of the ten letters. Score the total length of all ten words, with a theoretical maximum of 100. Present a new word each day, and add a button to share the results. To avoid spoilers, sort the ten words by length, and show green squares for the letters. Add red squares for the most common word that would have improved the score.

• type in a ten-letter word, and then drag letters around
• display total score, updated when you submit each word
• dictionary check when you submit each word
• pick a random ten-letter word from the dictionary
• add a check button that shows a hint
• build a list of words, slightly shuffled from frequency order, then rot13 and select by date
• add a share button that copies green squares, plus the regular score, the rough frequency of the hint, and time to complete
• hard mode with 5 minute time limit and killer mode with 1 minute time limit
• pass/fail option where the total score has some relationship with the frequency of the hint
• hard or killer mode increases the required ratio between frequency and score

It isn't clear which player name is labelling which grid in a three-player game. Try putting a box around the grid and label.

• drag and drop letters in grid
• enter player name and click Ready button
• see other names and grids appear with target buttons
• click a target button to reveal all opponents' letters in that spot, labeled with the spot's coordinates
• player's grid switches to light (hidden) and primary (revealed) letters
• add history with entries like "Alex hit 53" and "Blake hit 60"
• add instructions and link to David Parlett's site and books