Condorcet ranking: Difference between revisions
→Smith set ranking
Line 41:
{| class="wikitable"
|+Pairwise comparison table (candidate on left is preferred by # of voters in their cell over candidate on top)
Pairwise victories are bolded, defeats are underlined, ties are struck through
Groups of candidates who pairwise beat all other lower groups are italicized; every odd group's numbers are made bigger and even group's numbers smaller
Line 116 ⟶ 119:
|''<big>-</big>''
|}
The generalized Smith set ranking can be found by sorting the candidates in the [[Pairwise counting|pairwise comparison table]] based on their [[Copeland method|Copeland]] scores (or in any way that follows a Smith ranking), and then starting at the top left cell comparing two candidates, checking if all the cells to the right of the cell that is being looked at show pairwise victories; if so, the candidate to the left of the cell being looked at is in the Smith set, if not, repeatedly go one cell down and one cell to the right until at the cell directly below the cell furthest to the right in the top row with a pairwise tie or defeat, covering all cells from the top left cell to the bottom right cell, and checking again. Once all the cells to the right of the cells being looked at (the "covered" cells) show pairwise victories, all candidates to the left of the covered cells are in the Smith set. Go one cell down and one cell to the right, ignore all of the previously covered cells, and repeat the process to find the secondary Smith set, then repeat for the tertiary Smith set, etc. Then rank all candidates in the Smith set 1st, all candidates in the secondary Smith set 2nd, etc.
The generalized Smith set ranking is A=B=C'''>'''D=E'''>'''F=G, or (A, B, C) tied for 1st, (D, E) tied for 2nd, and (F, G) tied for 3rd place. This is because 3 voters prefer (A, B, C) over all others (and at most 2 voters prefer any other candidates over (A, B, C)) but no smaller subset of (A, B, C) can say the same, 2 voters prefer (D, E) above all others when ignoring (A, B, C) (and at most 1 voter prefers other unignored candidates over (D, E)) but no smaller subset of (D, E) can say the same, and 1 voter prefers (F, G) above all others when ignoring (A, B, C, D, E) (and no voters prefer any other unignored candidates over (F, G)) but no smaller subset of (F, G) can say the same. Notice that when looking at the pairwise comparison table for the generalized Smith ranking that the Smith sets can be seen by looking at which groups of candidates have pairwise victories over all other candidates when ignoring the pairwise matchups between candidates in the group. ▼
As an example, here, starting at the top left pairwise comparison cell (the A>A cell), we see a pairwise defeat or tie furthest to the right in the 3rd cell in the top pairwise comparison row (the A>C cell), so we cover all cells from the top left cell to the 3rd cell down and 3rd cell from the left:
{| class="wikitable"
!
!A
!B
!C
!D
!E
!F
!G
|-
|A
|''<big>-</big>''
|'''''<big>2</big>'''''
|<u>''<big>1</big>''</u>
|'''<small>3</small>'''
|'''<small>3</small>'''
|'''<small>3</small>'''
|'''<small>3</small>'''
|-
|B
|<u>''<big>1</big>''</u>
|''<big>-</big>''
|'''''<big>2</big>'''''
|'''<small>3</small>'''
|'''<small>3</small>'''
|'''<small>3</small>'''
|'''<small>3</small>'''
|-
|C
|'''''<big>2</big>'''''
|<u>''<big>1</big>''</u>
|''<big>-</big>''
|'''<small>3</small>'''
|'''<small>3</small>'''
|'''<small>3</small>'''
|'''<small>3</small>'''
|}
and we now see only pairwise victories to the right of all the covered cells (all the cells from A>A to C>C, here shown in larger font), so A, B, and C are in the Smith set. Going one cell down and one cell to the right, we ignore all previously covered cells and so now the top left cell of the covered cells is the D>D cell. We see a pairwise defeat or tie furthest to the right of this cell in the D>E cell, so we go one cell down and one cell to the right, making the E>E cell the bottom right cell of the covered cells.
{| class="wikitable"
|-
|D
|<u>''<sup><small>2</small></sup>''</u>
|<u>''<sup><small>2</small></sup>''</u>
|<u>''<sup><small>2</small></sup>''</u>
|''<sup><big>-</big></sup>''
|''<sup><s><big>1</big></s></sup>''
|'''<small>2</small>'''
|'''<small>2</small>'''
|-
|E
|<u>''<sup><small>2</small></sup>''</u>
|<u>''<sup><small>2</small></sup>''</u>
|<u>''<sup><small>2</small></sup>''</u>
|''<sup><s><big>1</big></s></sup>''
|''<sup><big>-</big></sup>''
|'''<small>2</small>'''
|'''<small>2</small>'''
|}
Now, all cells to the right of the covered cells (the covered cells are all cells from the D>D cell to the E>E cell, which are the cells from the 5th cell in the top row to the 6th cell in the bottom row) show pairwise victories, so D and E are in the secondary Smith set. Going one cell down and one cell to the right, and again ignoring all previously covered cells, the top left cell of the now-covered cells is F>F. The only cell to its right shows a pairwise tie, so we automatically know that whichever candidate is in the row below F is also in the tertiary Smith set with F.
{| class="wikitable"
|-
|F
|<u>''<small>1</small>''</u>
|<u>''<small>1</small>''</u>
|<u>''<small>1</small>''</u>
|<u>''<small>1</small>''</u>
|<u>''<small>1</small>''</u>
|''<big>-</big>''
|''<big><s>0</s></big>''
|-
|G
|<u>''<small>1</small>''</u>
|<u>''<small>1</small>''</u>
|<u>''<small>1</small>''</u>
|<u>''<small>1</small>''</u>
|<u>''<small>1</small>''</u>
|''<big><s>0</s></big>''
|''<big>-</big>''
|}
F and G are in the tertiary Smith set.
▲
{| class="wikitable"
|+
Line 141 ⟶ 228:
|<big>-</big>
|}
So for example, if 3 candidates are ranked 1st in Copeland, they are part of the Smith set, and if there are several candidates ranked lower than them by Copeland who pairwise beat or tie any of them, take the lowest ranked of them based on Copeland (suppose that candidate was ranked 5th by Copeland) and add all candidates ranked 5th or above by Copeland into the Smith set. Supposing no candidates pairwise beat or tie any of the candidates newly added into the Smith set, all candidates ranked 6th by Copeland are part of the secondary Smith set, and the process repeats to see who is part of the secondary Smith set, then tertiary Smith set, etc.
== Notes ==
Methods that produce only Condorcet rankings may include [[Kemeny-Young]]; methods that produce Smith rankings and thus also Condorcet rankings (when they exist) include [[Copeland's method]], [[Pairwise Sorted Methods]], [[Schulze]], and [[Ranked Pairs]];
The generalized Smith ranking can also be found by finding all candidates in the Smith set, equally ranking each of them, then eliminating all of them and finding the secondary Smith set (the Smith set now that the original Smith set has been eliminated) and equally ranking each of them lower than the candidates in the Smith set, and repeating until all candidates are ranked.
Smith or Condorcet rankings can be visualized in the following form:
Line 275 ⟶ 362:
This makes it more ambiguous as to how to record the exact margins of some [[Pairwise counting#Terminology|pairwise matchups]] where groups of candidates are involved, as even if all candidates in a group of candidates pairwise beat all candidates not in the group, they may each do so with different margins.
Theoretical note:
Note that a variant of generalized Smith rankings can be created to address the unanimity/Pareto criterion. For example:<blockquote>34 A>B>C>D
| |||