Algorithm used for matrix games too large to be solved using linear programming approach.
- Solve the subgame with T1 and T2 Ð→ (q ∗ 1 , q ∗ 2 )
- Compute the pure best responses s1 ∈ argmax s ′ 1 ∈S1 U(s ′ 1 , q ∗ 2 ) and s2 ∈ argmin s ′ 2 ∈S2 U(q ∗ 1 , s ′ 2 )
- If s1 ∈ T1 and s2 ∈ T2, then stop
- Otherwise add s1 to T1 or s2 to T2, and go to 1.