Solving Wordfeud with Artificial Intelligence
Niklas Arnell and Johan Wadenholt
Abstract
Background: With the introduction of Wordfeud on the mobile device market the general interest in
word puzzle games has increased and hence the interest in programs designed to solve Wordfeud. Most
Wordfeud solving programs designed today are in many ways limited since they do not look further
ahead than what combination of letters that returns the highest score. This can be a huge limitation
since these programs do not take in consideration what possibilities that might open up for the
opponent.
Results: Wordfaid does sustain a positive win/loss-ratio while playing against the original program it is
based on, Scrabaid. It does not always win by large amounts but all moves that Wordfaid makes have as
much purpose of protecting itself against retaliation as playing high scoring words and hence not playing
the riskiest words that could potentially result in huge wins.
Conclusions: Estimating your opponent’s possible moves by randomizing his/her letters remaining in
the bag several times and testing them on the board for each word that you could potentially play
proves to be a valuable asset in the conquest for Wordfeud victory since it helps you avoid opening up
big plays after your turn is over.