在贓物所有權歸屬問題之下，本文承繼 Rose (2010) 的設定，使用賽局模型分析贓物不適用善意取得 (即無條件歸還原持有者) 及適用善意取得的法律規定，並討論我國實際贓物所有權的制度，並比較三種制度下的參賽者的聯合效用大小。透過本研究可發現，原持有者擁有物品所有權的效用相對較高時，社會上應偏向贓物不適用善意取得較有效率；當潛在買方擁有物品所有權的效用相對較高時，社會上應偏向贓物適用善意取得較有效率。而我國現有制度不管在何種情況之下，皆不是最有效率的，故我國對贓物所有權歸屬的法律從效率面而言有重新檢討之必要。 / This paper which continues the study from Rose (2010) focuses on the question whether a buyer of a stolen good should obtain title to the good if he/she has purchased it in good faith. We use game model to analyze three different regulations - a rule of law where good faith is irrelevant, a rule where good faith may protect an innocent buyer, and the real regulation in Taiwan. Finally, in discussing social utility, we compare efficiency among the three regulations, and reach our conclusion that when the potential buyer’s utilities of having the ownership are higher than the owner’s, a rule of law where good faith is irrelevant is the most efficient. On the contrary, when the potential buyer’s utilities of having the ownership are lower than the original owner’s, a rule where good faith may protect an innocent buyer is the most efficient. Considering efficiency, the regulation in Taiwan is not the best to the society so the government may amend the regulation.