A developer implemented a game with two foxes and twenty sheep and explored its balance through search and self-play. Foxes initially won about 85% of games, but deeper search and a learned evaluation function challenged that conclusion. The sheep’s pen objective, movement constraints and draw rules also shaped strategy. The article describes building evaluation puzzles and tests alongside the game. Its results concern particular implementations, not a mathematical solution: search depth and evaluation design can change what an experiment appears to show.