Thanks for the comments. I noticed the reverse sorting of items in the sorted list: Delta, Epsilon, and Theta all show the same value, but they are listed with with Theta first, because ROW()/1000 gives a larger value for a higher row. I could have changed the added term to something like (100-ROW())/10000, which would restore the order of the items in the original list. Your formula looks nicer, though, and precludes the user having to decide what large number to divide the row number by.

You also mentioned that my chart is actually showing slightly distorted data, with numbers exaggerated by ROW()/1000. I should follow one of my cardinal rules, which is, if you should use another column, use another column. The added column would use the sorted numbers in column F to look up the actual data from column C, much as the labels in column E are looked up. The lookup formula is:

=INDEX($C$2:$C$9,MATCH(F2,$A$2:$A$9,0))

In fact I use this in my own web site tracking workbook, where I track statistics backwards for six months. I use the lookup for all six months of the tracking data, based on the sorted values for the most recent month.

Maybe when I get a minute I’ll fix the protocol above to address both of your suggestions.

In cell A2, use
=RANK(C2,$C$2:$C$9)+COUNTIF($C$2:C2,C2)-1
This formula creates an artificial sort allowing for duplicates with the use of the COUNTIF function. When duplicates values arise the first label in the list takes precedence.

In cell E2, use
=INDEX($B$2:$B$9,MATCH(ROW()-ROW($E$1),$A$2:$A$9,0))
This formula works in the same way as described in the blog but makes reference to the updated sort column.

In cell F2, use
=INDEX($C$2:$C$9,MATCH(ROW()-ROW($F$1),$A$2:$A$9,0))

Each of these formulae should be copied down.

The result is slightly different in that (a) my sort function is based on first listed data (as opposed to introducing ROW() as a tolerance variable), and (b) returns actual results (rather than close estimates).