Publications-NSC Projects
Article View/Open
Publication Export
-
Google ScholarTM
NCCU Library
Citation Infomation
Related Publications in TAIR
題名 貝氏評分系統(II) 其他題名 Bayesian Rating System 作者 翁久幸 貢獻者 國立政治大學統計學系
行政院國家科學委員會關鍵詞 數學;貝氏評分系統 日期 2009 上傳時間 30-Aug-2012 09:59:10 (UTC+8) 摘要 許多球類,棋藝,遊戲比賽,都有對各個參賽者實力作排名,目前有線上(online)與非線 上(offline) 排名方法. 線上排名(online rating)方法與非線上排名(offline rating)方法的主要差別在於前者不用儲存過去大量比賽結果. 由於許多比賽(特別 是線上遊戲)不僅參賽者多,而且比賽場次也多,每天可能就有上百萬參與者和場次. 線上排名方法不需要儲存過去大量比賽結果的特性,使得它廣受使用. 目前最有名的排名方法是Elo. 它被相當成功地用在許多兩人比賽之球類,棋藝等. 之後, Glickman 提出Glicko System. 它與Elo 的主要差別在於前者在"實力’這個 參數上引進變異(variability)的觀念. Glicko 也被成功地用在許多兩人比賽 隨著線上遊戲時代的到來,許多多人多隊比賽的遊戲越來越普遍. 如何評定各個參賽 者之實力也成為需要研究的問題.對此問題,最近Microsoft Research 提出一個評分 系統TrueSkill. 本計畫擬討論TrueSkill 之優缺點,然後提出一個簡易有效之評分 系統. 並且我們要以Halo 2 dataset (the beta testing of the Xbox)來測試我們 評分系統之表現.
There are some rating systems for sports, chess, and games. Some systems are online and some are offline. The main difference between online and offline is that the former does not need to store all the past data. Since many games (especially online games) involve a huge number of players and everyday there could be millions of games played, the online rating system is more suited to be used in such situation. Many have proposed online algorithms for paired comparison experiments. These For ranking of many sports, possibly the most prominent ranking system in use today is ELO, originally invented by Arpad Elo, as an improved chess rating system and now it has been used successfully by a variety of leagues organized around two-player games. Another famous updating algorithm is the Glicko system, developed by Mark E. Glickman, chairman of the US Chess Federation (USCF) ratings committee. The main difference between Elo and Glicko is that the later introduced “variability”into the skill parameter. Though the ELO and Glicko ranking system have been successfully, they are designed for two-player games. In video games many of these leagues have game modes with more than two players (and/or more than two teams) per match. To support such games, recently Microsoft Research developed the TrueSkill ranking system. In this project, we propose to use an approximate Bayesian method together with a generalization of the Bradley-Terry model to obtain simple update formulas in cases where there may be multiple teams and/or multiple players.關聯 基礎研究
學術補助
研究期間:9808~ 9907
研究經費:522仟元資料類型 report dc.contributor 國立政治大學統計學系 en_US dc.contributor 行政院國家科學委員會 en_US dc.creator (作者) 翁久幸 zh_TW dc.date (日期) 2009 en_US dc.date.accessioned 30-Aug-2012 09:59:10 (UTC+8) - dc.date.available 30-Aug-2012 09:59:10 (UTC+8) - dc.date.issued (上傳時間) 30-Aug-2012 09:59:10 (UTC+8) - dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/53397 - dc.description.abstract (摘要) 許多球類,棋藝,遊戲比賽,都有對各個參賽者實力作排名,目前有線上(online)與非線 上(offline) 排名方法. 線上排名(online rating)方法與非線上排名(offline rating)方法的主要差別在於前者不用儲存過去大量比賽結果. 由於許多比賽(特別 是線上遊戲)不僅參賽者多,而且比賽場次也多,每天可能就有上百萬參與者和場次. 線上排名方法不需要儲存過去大量比賽結果的特性,使得它廣受使用. 目前最有名的排名方法是Elo. 它被相當成功地用在許多兩人比賽之球類,棋藝等. 之後, Glickman 提出Glicko System. 它與Elo 的主要差別在於前者在"實力’這個 參數上引進變異(variability)的觀念. Glicko 也被成功地用在許多兩人比賽 隨著線上遊戲時代的到來,許多多人多隊比賽的遊戲越來越普遍. 如何評定各個參賽 者之實力也成為需要研究的問題.對此問題,最近Microsoft Research 提出一個評分 系統TrueSkill. 本計畫擬討論TrueSkill 之優缺點,然後提出一個簡易有效之評分 系統. 並且我們要以Halo 2 dataset (the beta testing of the Xbox)來測試我們 評分系統之表現. en_US dc.description.abstract (摘要) There are some rating systems for sports, chess, and games. Some systems are online and some are offline. The main difference between online and offline is that the former does not need to store all the past data. Since many games (especially online games) involve a huge number of players and everyday there could be millions of games played, the online rating system is more suited to be used in such situation. Many have proposed online algorithms for paired comparison experiments. These For ranking of many sports, possibly the most prominent ranking system in use today is ELO, originally invented by Arpad Elo, as an improved chess rating system and now it has been used successfully by a variety of leagues organized around two-player games. Another famous updating algorithm is the Glicko system, developed by Mark E. Glickman, chairman of the US Chess Federation (USCF) ratings committee. The main difference between Elo and Glicko is that the later introduced “variability”into the skill parameter. Though the ELO and Glicko ranking system have been successfully, they are designed for two-player games. In video games many of these leagues have game modes with more than two players (and/or more than two teams) per match. To support such games, recently Microsoft Research developed the TrueSkill ranking system. In this project, we propose to use an approximate Bayesian method together with a generalization of the Bradley-Terry model to obtain simple update formulas in cases where there may be multiple teams and/or multiple players. en_US dc.language.iso en_US - dc.relation (關聯) 基礎研究 en_US dc.relation (關聯) 學術補助 en_US dc.relation (關聯) 研究期間:9808~ 9907 en_US dc.relation (關聯) 研究經費:522仟元 en_US dc.subject (關鍵詞) 數學;貝氏評分系統 en_US dc.title (題名) 貝氏評分系統(II) zh_TW dc.title.alternative (其他題名) Bayesian Rating System en_US dc.type (資料類型) report en
