(1) uniform points in [0,1/4]^n; fractions of the cube (standard error in brackets)
   n=3  G(B)                   0.3660  [0.0008]
   n=3  {Q1>=Q2}               0.3547  [0.0008]
   n=3  {Q1>=Q2} minus G(B)    0.0000  [0.0000]
   n=3  G(B) minus {Q1>=Q2}    0.0113  [0.0002]
   n=4  G(B)                   0.6970  [0.0007]
   n=4  {Q1>=Q2}               0.8332  [0.0006]
   n=4  {Q1>=Q2} minus G(B)    0.1362  [0.0005]
   n=4  G(B) minus {Q1>=Q2}    0.0000  [0.0000]
   n=5  G(B)                   0.8825  [0.0005]
   n=5  {Q1>=Q2}               0.9580  [0.0003]
   n=5  {Q1>=Q2} minus G(B)    0.0755  [0.0004]
   n=5  G(B) minus {Q1>=Q2}    0.0000  [0.0000]
   n=6  G(B)                   0.9616  [0.0003]
   n=6  {Q1>=Q2}               0.9916  [0.0001]
   n=6  {Q1>=Q2} minus G(B)    0.0301  [0.0003]
   n=6  G(B) minus {Q1>=Q2}    0.0000  [0.0000]
(2) max of Q2/Q1 over random points of G(B) (lambda uniform in P_n, d = lambda(1-lambda))
   n=4: 499697 sampled points of G(B); max Q2/Q1 = 0.051650; points with Q1 < Q2: 0
   n=5: 574861 sampled points of G(B); max Q2/Q1 = 0.000719; points with Q1 < Q2: 0
   n=6: 594994 sampled points of G(B); max Q2/Q1 = 0.000000; points with Q1 < Q2: 0
   exact value at (1/4,1/4,1/4,0) for n = 4: Q2/Q1 = 27/512 = 0.052734 (see verify_paper.py, Part 8)
