評估玩家及團隊影響力來平衡線上遊戲的自動配對系統

Abstract

線上遊戲中的互動要素是玩家十分重要的樂趣來源，而部分遊戲更是以玩家間的競技為主軸，遊戲中的自動配對系統會配對出實力相近的隊伍來進行對抗，但是當配對的結果雙方的實力不平衡時，勝方會感到無聊沒有樂趣，敗方則會感到焦慮、挫折甚至辱罵自己的隊友。

現行的配對系統多半是衍生自ELO這套積分系統，每位玩家都會有一個代表自身實力的ELO積分，配對系統就是利用這ELO積分來配對出實力相近的隊伍，但不幸的，在ELO這套積分系統的限制下玩家必須進行足夠的場次，ELO積分才會真正的收斂符合自身實力，而配對系統並不知道玩家的積分是收斂或是未收斂，也因此常常產生不平衡的配對結果。

本研究提出利用推估玩家及隊伍的影響力來彌補未收斂的ELO積分所帶來的問題，影響力的推估能推測出玩家們的真正實力，在實驗中，除了原本配對系統根據ELO積分配對出的結果，再將兩隊玩家的影響力控制在相近的區間，這樣能大幅改善配對出來的場次平衡度，除了能有效避免不平衡場次的產生，也能讓玩家收斂積分的過程更加順利。

The interaction between players is a very important element of fun in online game. Some online games are based on team competition, with an automatic match making system to help players forming two teams with similar skill level. But if the match result is unbalanced, the players of winning team will feel bored and the players of losing team will feel anxiety, disappointed even blame teammates.

Almost all match making systems are derived from ELO ranking system. Every player has an ELO score which represents the player skill in this game. Most match making systems use ELO score to match two teams. Unfortunately the ELO ranking system has a limitation that player must play enough games to converge effective ELO score. Otherwise the score may not present real skill of player. The match maker mechanism does not know whether player’s ELO score is converged or not. So that match making system often generates unbalanced match.

In this research, we evaluate individual and team influences to solve the problem caused by unconverged ELO score. Evaluating influences can reveal the real skill of players. In our experiment, after match making system match a game, we evaluate team influences and control the value in similar interval. This method can improve balance of match a lot, so as to avoid unbalanced match effectively and help players to have an ELO score converged more smoothly.