In multiplayer zero sum game, it still works to Nash.
Intuitively, in a zero sum game, running CFR on one player will find him best responses for the strategies he's playing against. Doing this for multiple players finds corresponding best responses, which is the definition of a Nash Equilibrium.
In non-zero sum game CFR only converges to correlated equilibrium, which is a much weaker equilibrium than a Nash equilibrium.
Intuitively, in a zero sum game, running CFR on one player will find him best responses for the strategies he's playing against. Doing this for multiple players finds corresponding best responses, which is the definition of a Nash Equilibrium.
In non-zero sum game CFR only converges to correlated equilibrium, which is a much weaker equilibrium than a Nash equilibrium.