virtualenv venv source venv/bin/activate

pip install -r requirements.txt

python Hangman_Engine.py

Several trategies listed below are used:

• choose the most frequent character in all potentially matched words each time
• first guess based on the word-length based character frequency table
• adjust the weight based on universe word frequency table: When obsscured chars become fewer, it is hard to select simply based on the character frequency. For example: 'THE EMBRACE O_ THE BREASTS': It will return a character frequency table: {'g': 1, 'f': 1, 'k': 1, 'l': 1, 'n': 1, 'p': 1, 'u': 1, 'w': 1, 'v': 1, 'y': 1, 'x': 1, 'z': 1} it is hard to choose the correct word in this case. This problem can be fixed by adjusting the frequency weight based on the frequency of the potentially matched words. Larger weight is given to the specific char if the potentially matched word with this char has larger word frequency.
• increase the weight if the potentially matched word list is small

Apart from what mentioned above, the strategy also adapts dynamically. For example, it will not increase the weight when the number of obscured chars has not arrived certain threshold even if the potentially matched word list is small.

This only gives a big idea of how to guess. There are several hyper-parameters need to be fine-tunes in practice, such as the weight and threshold.