advantage         final success  first update >= 0.5  first update >= 0.9  middle-guess fraction
outcome                   0.985                    9                   26                  1.000
critic lam=1              0.951                    9                   27                  0.750
critic lam=0.9            0.937                    9                   50                  0.625
critic lam=0              0.929                    9                   62                  0.618
