←
Condorcet ranking
Tally methods
Method comparison
Each method guarantees some properties and gives up others. Our methods side by side, set against first-past-the-post and three benchmark methods.
| Offered on Condorcet.Vote | Benchmarks, not offered | |||||||
|---|---|---|---|---|---|---|---|---|
| Criterion | Schulze | Ranked Pairs | Copeland | Dodgson Quick | First-past-the-post | Instant-runoff voting (IRV) | Borda | Kemeny-Young |
| Majority An option ranked first by more than half of the voters wins. | Yes | Yes | Yes | Yes Note 1 | Yes | Yes | No | Yes |
| Majority loser An option ranked last by more than half of the voters never wins. | Yes | Yes | Yes | Not established | No | Yes | Yes | Yes |
| Mutual majority If a majority ranks a group of options ahead of all the others, the winner belongs to that group. | Yes | Yes | Yes | Not established | No | Yes | No | Yes |
| Condorcet winner The option that beats each of the others in a duel is always elected. | Yes | Yes | Yes | Yes Note 2 | No | No | No | Yes |
| Condorcet loser The option beaten by each of the others in a duel is never elected. | Yes | Yes | Yes | Not established | No | Yes | Yes | Yes |
| Smith criterion The winner belongs to the smallest group of options that beat every option outside it. | Yes | Yes | Yes | Not established | No | No | No | Yes |
| Independence of clones Adding an option very similar to another does not change the outcome among the others. | Yes | Yes Note 3 | No | Not established | No | Yes | No | No |
| Monotonicity Ranking the winner higher can never make them lose. | Yes | Yes | Yes | Not established | Yes | No | Yes | Yes |
| Local independence Removing the option that finished last (or first) in the result does not change the order of the others. | No | Yes | No | Not established | No | No | No | Yes |
| Reversal symmetry If every ballot is reversed, the former winner no longer wins. | Yes | Yes | Yes Note 4 | Not established | No | No | Yes | Yes |
| Participation Voting sincerely can never leave a voter worse off than abstaining. | No | No | No | No Note 5 | Yes | No | Yes | No |
| Consistency If an option wins in two separate groups, it wins when they are combined. | No | No | No | No Note 5 | Yes | No | Yes | No |
| Later-no-harm Ranking an additional choice can never hurt those already ranked ahead of it. | No | No | No | No Note 5 | Yes Note 6 | Yes | No | No |
| Resolvability A final tie becomes unlikely as the number of voters grows. | Yes | Yes | No | Not established | Yes | Yes | Yes | Yes |
| Fast computation The result is computed in polynomial time, even with many options. | Yes | Yes | Yes | Yes | Yes | Yes | Yes | No Note 7 |
- ✓ always satisfied
- ✕ can be violated in some votes
- ? no published source establishes it
- 1 A direct consequence of the Condorcet criterion: an option ranked first by a majority beats every other option in a duel.
- 2 Guaranteed by construction: CondorcetPHP always places the Condorcet winner first.
- 3 When two duels have exactly the same strength, full independence of clones requires the Zavist and Tideman (1989) tie-breaking rule, which CondorcetPHP does not apply: it breaks ties by the weakest opposition, then processes together any duels that remain tied.
- 4 Blank cell in Wikipedia’s general table; “yes” in its table of Condorcet methods and in the legacy condorcet.vote table.
- 5 No method that satisfies the Condorcet criterion can satisfy this criterion: this is a general result.
- 6 Trivially: first-past-the-post never looks beyond the first choice.
- 7 The exact computation is NP-hard: it becomes impractical beyond about ten options.
Reading guide
How to read this table.
- A “no” is a worst case. It means there is at least one vote, sometimes a highly contrived one, where the criterion is violated; not that the method violates it often.
- Some “no”s are unavoidable. No method that elects the Condorcet winner can guarantee participation, consistency or later-no-harm: these are theorems.
- The variants are equal here. Winning votes and margins share every cell of this table.
- Dodgson Quick is an approximation. No published table covers it: only the established cells are filled in, and the others are marked as such.
Sources
Where this data comes from.
Wikipedia’s criteria table, supplemented by its table of Condorcet methods and, for the resolvability of the benchmark methods, by the legacy condorcet.vote table. Disagreements between sources are flagged in the notes.
| Type | Reference |
|---|---|
| Encyclopedia | Comparison of electoral systems Wikipedia |
| Encyclopedia | Condorcet method — Evaluation by criteria Wikipedia |
The other methods
- The Schulze method The strongest paths of victories between each pair of options. Our default method.
- The Ranked Pairs method Duels locked in one by one, from the most decisive to the closest, skipping any that would create a cycle.
- The Copeland method One point per duel won, one taken away per duel lost: simple to count by hand.
- The Dodgson Quick method The option closest to beating everyone, following Lewis Carroll’s idea.