Table 2.

Results of guided halving algorithm in the case of two outgroups

R1−R2−T	n	d(T,X′⊕X′′)	Median cost	d(T,A′⊕A′′)	Median cost	ΔA	d(A, A*)	Time
AG-KL-SC	497	117	364	117	361	−3	40	131
AG-KW-SC	478	116	502	116	498	−4	41	204
KL-KW-SC	471	121	518	121	516	−2	48	217
AG-KL-CG	265	183	300	183	297	−3	124	48
AG-KW-CG	261	184	362	184	361	−1	138	55
KL-KW-CG	259	184	368	184	366	−2	136	62
AG-KL-V	283	61	278	61	275	−3	47	38
AG-KW-V	280	61	340	62	339	0	51	41
KL-KW-V	277	62	354	62	352	−2	54	54

Median cost refers to the sum of the three distances, from R₁,R₂ and the inferred ancestor X or A, to the median. The objective is d(T,X^′⊕X^″)+median cost. Δ_A is the improvement of d(T,A⊕A)+median cost over d(T,X′⊕X′′)+median cost due to local searching, allowing A to move outside the set of halving solutions. Time in minutes.