A　Game　arithmetic　by　forecasting　opposed　robot　state　in　danger　area　to　select　optimal　strategy　is　proposed　in　this　paper.　Generally,　only　the　home　robot,　the　ball　and　the　goal　are　considered　in　traditional　attack　or　defence　action　to　select　appropriate　strategy,　but　it　is　unsuitable　in　realization　of　area　intention.　This　arithmetic　conquers　the　disadvantage　above.　The　evaluations　of　minatory　robot　and　its　possible　action　are　integrated　to　make　optimal　decision　in　the　Game　course.　The　experiments　conducted　on　simulation　show　the　validity　of　it　in　part　attack　or　defence　decision-making.