runs complete: {'early': 12, 'stop': 12, 'native': 12}
early   runs 12  attempts 120  checks written 119  delivered early 114 (95%)  abandoned 5  gate runs/att 2.36  shown/att 2.16  fixed 111  still failing 148
        score 0.574  $/run 0.0790  uncached-in $0.0229  cached-in $0.0124  output $0.0438  requests 128
stop    runs 12  attempts 120  checks written 120  delivered early   0 (0%)  abandoned 0  gate runs/att 0.00  shown/att 1.88  fixed 78  still failing 148
        score 0.572  $/run 0.0914  uncached-in $0.0280  cached-in $0.0158  output $0.0476  requests 150
native  runs 12  attempts 120  checks written   0  delivered early   0 (0%)  abandoned 0  gate runs/att 0.00  shown/att 0.00  fixed 0  still failing 0
        score 0.562  $/run 0.0535  uncached-in $0.0195  cached-in $0.0100  output $0.0240  requests 103
early vs native: score +0.013 [-0.032, +0.062]  cost ratio 1.476 [1.312, 1.676]
early vs stop: score +0.003 [-0.051, +0.052]  cost ratio 0.865 [0.794, 0.947]
stop vs native: score +0.010 [-0.043, +0.071]  cost ratio 1.707 [1.522, 1.946]

per task score (arm means):
  008 early 0.55  stop 0.60  native 0.60
  015 early 0.57  stop 0.62  native 0.62
  018 early 0.45  stop 0.45  native 0.45
  029 early 1.00  stop 0.95  native 0.95
  031 early 1.00  stop 1.00  native 0.97
  062 early 0.63  stop 0.72  native 0.57
  066 early 0.56  stop 0.49  native 0.54
  080 early 0.08  stop 0.04  native 0.07
  150 early 0.30  stop 0.32  native 0.30
  154 early 0.61  stop 0.54  native 0.56
