戦艦ゲームにおいて相手モデルを立てない場合の攻撃及び配置戦略の評価

書誌事項

タイトル別名
  • Evaluation of Attacking and Placing Strategies in a Classic Battleship Game without Considering Opponent Models

抄録

ポピュラーな不完全情報ボードゲームの代表として,サイズ2〜5の船5隻と10×10ボードを使う戦艦ゲームを選び,相手モデルを考慮しない場合の有効な攻撃戦略及び配置戦略について調べた.攻撃戦略においては,市松模様攻撃を基本として攻撃がハズレになった場合の可能配置数を最小にする枡を攻撃する,すなわち,当たる確率が最大の枡を攻撃する戦略が優秀であった.その際,早く可能配置数え上げを行うほど攻撃精度が上がることがわかったが,時間の制約を考慮して,残り船数が4隻以下でかつ未確定枡数が70以下のとき可能配置数え上げを行うものを良い攻撃戦略として選んだ.攻撃戦略を上記に固定し数種の配置戦略の評価を行ったところ,船を辺に沿って配置する戦略が優秀で,それに加え船の連接配置を禁止したものが最も優秀な成績を収めた.

A battleship game that uses 10 × 10 boards and five ships of the sizes from 2 to 5 has been chosen as a target domain for a representative testbed of the popular board games with incomplete information. The effective strategies for the attack or the placement of ships have been investigated without considering opponent models. As to the attacking strategies, a strategy, which is based on the checkerboard strategy and selects a square such that the number of possible placements is minimized when the attack is assumed to be missed, that is, selects a square such that the hit probability is maximized, has shown the most excellent results. The accuracy of the attacking strategy has turned out to get better as the enumeration of possible placements is performed earlier. Considering the time constraint of experiments, we have selected the attacking strategy that starts the enumeration when the number of remaining ships is fewer than or equal to 4 and the number of unidentified squares is fewer than or equal to 70. The examinations of some placing strategies, with the above attacking strategy fixed, have shown a placing strategy that places ships along the edges is superior. In addition, the above placing strategy with prohibiting the connected placements of ships has shown the most excellent results.

収録刊行物

詳細情報 詳細情報について

  • CRID
    1050292572120139008
  • NII論文ID
    170000080167
  • Web Site
    http://id.nii.ac.jp/1001/00097500/
  • 本文言語コード
    ja
  • 資料種別
    conference paper
  • データソース種別
    • IRDB
    • CiNii Articles

問題の指摘

ページトップへ