Skip to content

chrisconley/hangman

Repository files navigation

Hangman strategies in python

Until I get a proper README up, check out the following the lightning talk for a quick overview:

Setup

Set up a virtual environment and activate it

python3.5 -mvenv ~/venv/hangman
source ~/venv/hangman/bin/activate

Install dependencies

pip install -r requirements.txt

Prepare ./build/ with processed dictionary files

make split

Running Different Strategies

Random

time cat build/splits/9 | python play.py - --config games/01-random.cfg --reset-memory --limit 25011
Average Score:  15.4326896166 22.1743632802

real	12m42.964s
user	12m42.115s
sys	0m0.870s

Naive

time cat build/splits/9 | python play.py - --config games/02-naive.cfg --reset-memory --limit 25011
Average Score:  9.28483467274 16.3595218104

real	14m58.874s
user	14m57.871s
sys	0m1.049s

Feedback Distinct

time cat build/splits/9 | python play.py - --config games/03-feedback-distinct.cfg --reset-memory --limit 25011
Average Score:  7.02450921594 14.1903562433

real	13m47.919s
user	13m47.074s
sys	0m0.877s

Feedback Positional

time cat build/splits/9 | python play.py - --config games/05-feedback-positional.cfg --reset-memory --limit 25011
Average Score:  1.40973971453 5.97277198033

real	1m29.261s
user	1m29.064s
sys	0m0.201s

Maximize Information

time cat build/splits/9 | python play.py - --config games/06-entropy-mismatched-scorers.cfg --reset-memory --limit 25011
Average Score:  1.66422773979 5.09367878134

real	1m29.895s
user	1m29.674s
sys	0m0.227s

Optimize

time cat build/splits/9 | python play.py - --config games/07-entropy-same-scorer.cfg --reset-memory --limit 25011
Average Score:  1.38179201151 5.37903322538

real	1m28.110s
user	1m27.906s
sys	0m0.209s

About

Hangman strategies in python

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published