How the 2027 Date Vote Will Be Counted

Walkthrough

How the count works, in seven steps

A guided version of the method below, with worked examples. Use the arrows (or your keyboard) to step through; the Present button makes it full-screen for a call.

The short version

The winning week is the one that would beat every other week in a head-to-head vote.

That is the whole idea. Everything else on this page is the machinery for two questions: how do we run those head-to-heads from a ranked ballot, and what if no week beats all the others?

Not this

“Which week got the most first choices?”

This

“Which week would win a two-way vote against each of the others?”

Step 1 · Your ballot

You rank the weeks. As many or as few as you like.

Rank as many as you like. Only the order matters, not the numbers themselves.

Equal ranks are fine. Two weeks you like the same can share a place.

Anything you leave out is joint last. If you rank only one week, you have no say in the head-to-heads between the others. Ranking more gives you more say, never less.

The shortlist above is illustrative. The real ballot will list the shortlisted weeks plus the current early-July dates.

Step 2 · One head-to-head

Take any two weeks. Which do more voters rank higher?

Eleven voters, four weeks on the ballot. Each card is one ballot. We only look at where and sit relative to each other and ignore everything else.

Step 3 · Every week plays every other week

Like a league: each week plays each of the others. The week that beats them all wins.

First choices only

Head-to-head results

Step 4 · When no week beats all the others

Sometimes it's rock–paper–scissors. Then we trust the biggest wins first.

Nine voters, four weeks

Results, strongest first

Step 5 · If a week becomes unavailable after the vote

We don't promote the runner-up. We re-run the count without that week. The winner can change.

Same nine ballots as the last step. Say the authorities rule out a week — bird-nesting rules, say — after the vote. Nobody re-votes; the ballots are simply counted again with that week removed.

Remove a week and recount

What to remember

Rank as many weeks as you can. Ties are fine. Unranked weeks count as joint last, and ranking only one week silences you in the other contests.
The winner is the week that beats every other week head-to-head. Most first choices is not the same thing, and it is not what wins.
If there's a loop, the biggest wins are trusted first and the weakest contradicting result is set aside.
If a week is vetoed afterwards, the count is re-run without it. The runner-up does not automatically inherit, and the winner can change. That is the method working, not failing.
One method, fixed before anyone votes. Other counts may be shown afterwards for transparency; they decide nothing.

The rest of this page is the reference version: the formal description, the same examples in text, and how the alternatives compare.

Selected system

Ranked Pairs (Tideman)

Ranked Pairs looks for the date that performs best in direct, head-to-head comparisons with every other date. If one date is preferred to every alternative by a majority, it wins. If preferences form a cycle, Ranked Pairs resolves it by giving priority to the strongest head-to-head victories without creating a contradictory loop.

It looks for broad agreement

A date with wide second-choice support is not eliminated simply because another date has more first-choice votes.

Our options are very similar

Adjacent weeks can split the same group of voters. Ranked Pairs is comparatively resistant to this “clone” effect.

It handles genuine indifference

Asociados may rank dates equally. A tied pair contributes no preference to that particular head-to-head comparison.

It supports later exclusions

If a date proves unavailable after the vote, it can be removed and the same ballots recounted without rewriting anyone’s preferences.

How Ranked Pairs works

Each Asociado ranks as many dates as they wish. Equal rankings are allowed; unranked or rejected dates are treated as equally least-preferred.
Every pair of dates is compared: how many voters prefer A to B, and how many prefer B to A?
Those victories are ordered from strongest to weakest.
Victories are “locked in” in that order, except where doing so would create a cycle.
The date left at the top of the resulting preference order wins.

Any exact procedural tie-breaking rule will be declared before voting. The official counting method will not be changed after seeing the result.

Two examples with four options

These simplified examples use options A, B, C and D. The numbers represent groups of voters who submitted the same ranking.

1. The consensus option has few first choices

4 voters: A > B > C > D
3 voters: C > B > D > A
3 voters: D > B > C > A
1 voter: B > A > C > D

B has only one first-choice vote, so IRV would eliminate it immediately. But B beats A by 7–4, C by 8–3, and D by 8–3 in direct comparisons.

Ranked Pairs chooses B: the option preferred by a majority over every alternative, despite not leading on first choices.

2. A preference cycle would otherwise deadlock

4 voters: A > B > C > D
3 voters: B > C > A > D
2 voters: C > A > B > D

The group produces a cycle: A beats B by 6–3, B beats C by 7–2, but C beats A by 5–4. A simple Condorcet check therefore has no winner.

Ranked Pairs locks B over C first, then A over B. It skips C over A because that final, weakest victory would create the cycle A > B > C > A.

Ranked Pairs chooses A: it resolves the cycle by respecting the stronger victories before the weaker conflicting one.

Why changing date availability can look counter-intuitive

A preference cycle is like rock–paper–scissors: each option beats another, but is beaten by a third. There is no Condorcet winner—no single option that beats every other available option head-to-head. In the example above, A beats B, B beats C, and C beats A. Ranked Pairs resolves that loop by locking the strongest victories first and skipping any later victory that would recreate the cycle.

An option inside the cycle may therefore help determine which other victories can be locked, even when that option does not ultimately win. If that option later becomes unavailable, the cycle is broken and the remaining options are recounted directly. The winner can change even if the previous winner is still available.

For example, if B were removed from the cycle above, the remaining head-to-head comparison is C against A. Because C beats A by 5–4, C becomes the winner—even though Ranked Pairs selected A when B was included. This is not a lost or transferred vote, and the saved ballots have not changed. It is the result of applying the same method to a different set of genuinely available dates.

A missing Condorcet winner does not always prove that a cycle exists: tied head-to-head contests can also leave no option that strictly beats every other option. The Humans results page calls out this behaviour only when an actual preference cycle has been detected.

Comparison of the main systems

System	Main strength	Main weakness	Fit for this vote
Ranked Pairs	Finds a head-to-head consensus winner and handles cycles; resistant to similar options splitting support.	Less familiar and slightly harder to explain than a simple elimination count.	Selected. Best match for choosing among many similar dates.
Condorcet check	Directly identifies a date that beats every alternative one-to-one.	It is not a complete method: sometimes A beats B, B beats C, and C beats A.	Useful diagnostic, but it needs a completion method such as Ranked Pairs.
Borda count	Uses the whole ranking and often rewards broadly acceptable compromise options.	Sensitive to strategic ranking and to which alternatives happen to be included.	Reasonable comparison result, but weaker protection against tactical effects.
Baldwin	Combines Borda’s use of the full ranking with successive elimination.	More complex, and the elimination path can make the result harder to interpret.	A useful alternative analysis, but not simpler or more compelling here.
IRV	Well known and relatively easy to describe as repeated last-place elimination.	Can eliminate the broadly acceptable consensus date early because it has too few first choices.	Familiar, but poorly matched to our main risk.
Coombs	Actively removes the option receiving the most last-place opposition.	Highly sensitive to bottom rankings and unfamiliar to most voters.	Interesting where avoiding a disliked option is paramount, but less natural for date selection.

The systems in brief

Ranked Pairs

Compares every option against every other option, then locks the strongest victories while preventing cycles.

Condorcet check

Asks whether one option defeats every other option head-to-head. It may return no winner when preferences cycle.

Borda count

Awards points for every position on each ballot. The option with the highest total score wins.

Baldwin

Repeatedly calculates Borda scores and eliminates the lowest-scoring option until one remains.

Instant-runoff voting (IRV)

Counts first preferences, eliminates the weakest option, and transfers those ballots to their next remaining preference.

Coombs

Checks for a first-choice majority, then eliminates the option with the most last-place rankings until one wins.

One method, chosen in advance

Other systems may be shown after the vote to demonstrate how sensitive the result is to the counting rule. They will not be used to choose whichever outcome looks most convenient. Ranked Pairs is the official method because it has been selected before the ballots are counted.