Two training seeds per controller row

References: oracle 0.99, wait-and-see (the memoryless bound) 0.51, stay 0.38, random 0.31.

two seeds

interceptions / floor visit

controller	seed 0	seed 1	mean	spread
`ctrl_v31_tau1`	0.73	0.97	0.85	0.24
`ctrl_v31_real`	0.65	0.75	0.70	0.10
`ctrl_v31_poshead`	0.68	0.65	0.66	0.02
`ctrl_v31_ff`	0.46	0.63	0.55	0.17
`ctrl_v31`	0.57	0.50	0.53	0.07
`ctrl_v31_z_only_tau1`	0.51	0.53	0.52	0.03

interceptions / visit, long moves (>0.35)

controller	seed 0	seed 1	mean	spread
`ctrl_v31_tau1`	0.72	0.66	0.69	0.06
`ctrl_v31_poshead`	0.71	0.49	0.60	0.23
`ctrl_v31_real`	0.37	0.51	0.44	0.14
`ctrl_v31`	0.45	0.18	0.32	0.28
`ctrl_v31_ff`	0.20	0.38	0.29	0.19
`ctrl_v31_z_only_tau1`	0.06	0.25	0.15	0.20

controller	seed 0	seed 1	mean	spread
`ctrl_v31_poshead`	0.30	0.27	0.28	0.03
`ctrl_v31`	0.28	0.24	0.26	0.04
`ctrl_v31_real`	0.34	0.17	0.25	0.17
`ctrl_v31_tau1`	0.22	0.10	0.16	0.12
`ctrl_v31_ff`	0.13	0.09	0.11	0.04
`ctrl_v31_z_only_tau1`	0.02	0.03	0.02	0.01

Above the memoryless bound?

controller	two-seed mean	above the bound on
`ctrl_v31`	0.53	ONE SEED ONLY
`ctrl_v31_tau1`	0.85	both seeds
`ctrl_v31_ff`	0.55	ONE SEED ONLY
`ctrl_v31_poshead`	0.66	both seeds
`ctrl_v31_z_only_tau1`	0.52	ONE SEED ONLY
`ctrl_v31_real`	0.70	both seeds

Which adjacent orderings survive two seeds – interceptions / floor visit

above	below	gap of means	larger spread	seeds overlap	resolved
`ctrl_v31_tau1`	`ctrl_v31_real`	0.15	0.24	yes	no
`ctrl_v31_real`	`ctrl_v31_poshead`	0.04	0.10	yes	no
`ctrl_v31_poshead`	`ctrl_v31_ff`	0.12	0.17	no	no
`ctrl_v31_ff`	`ctrl_v31`	0.02	0.17	yes	no
`ctrl_v31`	`ctrl_v31_z_only_tau1`	0.02	0.07	yes	no

Which adjacent orderings survive two seeds – interceptions / visit, long moves (>0.35)

above	below	gap of means	larger spread	seeds overlap	resolved
`ctrl_v31_tau1`	`ctrl_v31_poshead`	0.09	0.23	yes	no
`ctrl_v31_poshead`	`ctrl_v31_real`	0.16	0.23	yes	no
`ctrl_v31_real`	`ctrl_v31`	0.13	0.28	yes	no
`ctrl_v31`	`ctrl_v31_ff`	0.03	0.28	yes	no
`ctrl_v31_ff`	`ctrl_v31_z_only_tau1`	0.13	0.20	yes	no

above	below	gap of means	larger spread	seeds overlap	resolved
`ctrl_v31_poshead`	`ctrl_v31`	0.02	0.04	yes	no
`ctrl_v31`	`ctrl_v31_real`	0.01	0.17	yes	no
`ctrl_v31_real`	`ctrl_v31_tau1`	0.10	0.17	yes	no
`ctrl_v31_tau1`	`ctrl_v31_ff`	0.05	0.12	yes	no
`ctrl_v31_ff`	`ctrl_v31_z_only_tau1`	0.08	0.04	no	yes

Non-adjacent separations that hold on both seeds

Every pair (not only neighbours) whose two seed ranges do not overlap at all. These are the orderings this experiment actually established.

metric	above	below	above’s seeds	below’s seeds
interceptions / floor visit	`ctrl_v31_tau1`	`ctrl_v31_poshead`	[0.73, 0.97]	[0.65, 0.68]
interceptions / floor visit	`ctrl_v31_tau1`	`ctrl_v31_ff`	[0.73, 0.97]	[0.46, 0.63]
interceptions / floor visit	`ctrl_v31_tau1`	`ctrl_v31`	[0.73, 0.97]	[0.50, 0.57]
interceptions / floor visit	`ctrl_v31_tau1`	`ctrl_v31_z_only_tau1`	[0.73, 0.97]	[0.51, 0.53]
interceptions / floor visit	`ctrl_v31_real`	`ctrl_v31_ff`	[0.65, 0.75]	[0.46, 0.63]
interceptions / floor visit	`ctrl_v31_real`	`ctrl_v31`	[0.65, 0.75]	[0.50, 0.57]
interceptions / floor visit	`ctrl_v31_real`	`ctrl_v31_z_only_tau1`	[0.65, 0.75]	[0.51, 0.53]
interceptions / floor visit	`ctrl_v31_poshead`	`ctrl_v31_ff`	[0.65, 0.68]	[0.46, 0.63]
interceptions / floor visit	`ctrl_v31_poshead`	`ctrl_v31`	[0.65, 0.68]	[0.50, 0.57]
interceptions / floor visit	`ctrl_v31_poshead`	`ctrl_v31_z_only_tau1`	[0.65, 0.68]	[0.51, 0.53]
interceptions / visit, long moves (>0.35)	`ctrl_v31_tau1`	`ctrl_v31_real`	[0.66, 0.72]	[0.37, 0.51]
interceptions / visit, long moves (>0.35)	`ctrl_v31_tau1`	`ctrl_v31`	[0.66, 0.72]	[0.18, 0.45]
interceptions / visit, long moves (>0.35)	`ctrl_v31_tau1`	`ctrl_v31_ff`	[0.66, 0.72]	[0.20, 0.38]
interceptions / visit, long moves (>0.35)	`ctrl_v31_tau1`	`ctrl_v31_z_only_tau1`	[0.66, 0.72]	[0.06, 0.25]
interceptions / visit, long moves (>0.35)	`ctrl_v31_poshead`	`ctrl_v31`	[0.49, 0.71]	[0.18, 0.45]
interceptions / visit, long moves (>0.35)	`ctrl_v31_poshead`	`ctrl_v31_ff`	[0.49, 0.71]	[0.20, 0.38]
interceptions / visit, long moves (>0.35)	`ctrl_v31_poshead`	`ctrl_v31_z_only_tau1`	[0.49, 0.71]	[0.06, 0.25]
interceptions / visit, long moves (>0.35)	`ctrl_v31_real`	`ctrl_v31_z_only_tau1`	[0.37, 0.51]	[0.06, 0.25]
displacement while blind / required move	`ctrl_v31_poshead`	`ctrl_v31_tau1`	[0.27, 0.30]	[0.10, 0.22]
displacement while blind / required move	`ctrl_v31_poshead`	`ctrl_v31_ff`	[0.27, 0.30]	[0.09, 0.13]
displacement while blind / required move	`ctrl_v31_poshead`	`ctrl_v31_z_only_tau1`	[0.27, 0.30]	[0.02, 0.03]
displacement while blind / required move	`ctrl_v31`	`ctrl_v31_tau1`	[0.24, 0.28]	[0.10, 0.22]
displacement while blind / required move	`ctrl_v31`	`ctrl_v31_ff`	[0.24, 0.28]	[0.09, 0.13]
displacement while blind / required move	`ctrl_v31`	`ctrl_v31_z_only_tau1`	[0.24, 0.28]	[0.02, 0.03]
displacement while blind / required move	`ctrl_v31_real`	`ctrl_v31_ff`	[0.17, 0.34]	[0.09, 0.13]
displacement while blind / required move	`ctrl_v31_real`	`ctrl_v31_z_only_tau1`	[0.17, 0.34]	[0.02, 0.03]
displacement while blind / required move	`ctrl_v31_tau1`	`ctrl_v31_z_only_tau1`	[0.10, 0.22]	[0.02, 0.03]
displacement while blind / required move	`ctrl_v31_ff`	`ctrl_v31_z_only_tau1`	[0.09, 0.13]	[0.02, 0.03]