%I A064387
%S A064387 1,3,6,2,7,13,20,12,21,11,22,10,23,9,24,8,25,43,62,42,63,41,18,
%T A064387 44,19,45,72,100,71,101,70,38,5,39,4,40,77,115,76,36,78,120,163,
%U A064387 119,74,28,75,27,79,29,80,132,185,131,186,130,73,15,81,141,202
%N A064387 Variation (2) on Recaman's sequence (A005132): to get a(n), we first 
               try to subtract n from a(n-1): a(n) = a(n-1)-n if positive and not 
               already in the sequence; if not then a(n) = a(n-1)+n+i, where i >
               = 0 is the smallest number such that a(n-1)+n+i has not already appeared.
%C A064387 Variation (4) (A064389) is the nicest of these variations.
%C A064387 I would also like to get the following sequences: number of steps before 
               n appears (or 0 if n never appears), list of numbers that never appear, 
               height of n (cf. A064288, A064289, A064290), etc.
%D A064387 Suggested by J. C. Lagarias.
%H A064387 <a href="Sindx_Rea.html#Recaman">Index entries for sequences related 
               to Recaman's sequence</a>
%H A064387 Nick Hobson, <a href="a064387.py.txt">Python program for this sequence</
               a>
%p A064387 h := array(1..100000); maxt := 100000; a := array(1..1000); a[1] := 1; 
               h[1] := 1; for nx from 2 to 1000 do t1 := a[nx-1]-nx; if t1>0 and 
               h[t1] <> 1 then a[nx] := t1; if t1 < maxt then h[t1] := 1; fi; else 
               for i from 0 to 1000 do t1 := a[nx-1]+nx+i; if h[t1] <> 1 then a[nx] 
               := t1; if t1 < maxt then h[t1] := 1; fi; break; fi; od; fi; od; evalm(a);
%Y A064387 Cf. A005132, A046901, A064388, A064389. Agrees with A064389 for first 
               187 terms, then diverges.
%Y A064387 Sequence in context: A076543 A005132 A064388 this_sequence A064389 A118201 
               A113880
%Y A064387 Adjacent sequences: A064384 A064385 A064386 this_sequence A064388 A064389 
               A064390
%K A064387 nonn,easy
%O A064387 1,2
%A A064387 N. J. A. Sloane (njas(AT)research.att.com), Sep 28 2001