JOTTO

Jotto is a code breaking game played between two players.

Each player picks a secret code word from the provided dictionary that the other player must try to guess. Every word in the dictionary is a lower-case 5-letter word in the English language, though some are somewhat obsure.

Players alternate taking turns guessing words from the dictionary until one has correctly identified his oppenent's word. After each guess the player is told how many of the letters in his word appeared in his oppent's secret word.

Examples:

---

Player1 has secret word "hello" Player2 guesses the word "world"

  w : Absent from hello   0
  o : Present in hello  + 1 
  r : Absent from hello + 0
  l : Present in hello  + 1
  d : Absent from hello + 0
  --------------------------
                          2
  world scores 2

---

---

Player1 has secret word "hello" Player2 guesses the word "skill"

  s : Absent from hello         0
  k : Absent from hello       + 0
  i : Absent from hello       + 0
  l : Present in hello        + 1
  l : Present again in hello  + 1
  --------------------------------
                                2
  skill scores 2

---

INSTALLATION

After cloning the repo, just type in the terminal.

make

To "uninstall" type

make clean

RULES

1. At the beginning of the game when promted, enter a valid 5 letter word to be your codeword.

2. On your turn when prompted, enter a valid 5 letter word in all lower case letters.

FUTURE WORK

1. Language Fluency

a. LM for AI generated guesses and codeword Many of the words in the dictionary I provided are somewhat obscure. I would like to give them weights according to their unigram frequencies in other text corpora and use these to assign probabilities of picking words, either guesses or codewords. I might disallow infrequent words all together since it seems unfair for the computer to have access to all such words.

b. Language Fluence for Different AI Levels I could also use a convex combination of unigram language model (LM) scores with a score based on maximum information in order to decide the optimal guess. The "dumber" AI will pick with a higher LM score and the best AI will ignore the LM completely.

c. Language Fluency for Final Guess The LM might also be useful in guessing the final word once 4 or all letters are known, as it seems more likely for people to pick common words than obscure ones.

2. Human-like Guess Generator

I am also almost done with a more human-like approach to guessing words which involves looking at valid combinations of letters given certain guesses, but there are some final bits of logic for generating new guesses that I need to work out and program.

3. Machine Aided Guessing

I would like to add a few features that allows the user to view candidate words from the word list given certain criteria, as well as to automatically view the potential valid letter combinations for each guess. This will require more of an event based program.

4. GUI I would like to make a basic GUI for the user. The terminal version of the game is fine for now, but it would be nice to have something more natural.