Electoral systems
A seat projection is only as good as the rule that turns votes into seats, and that rule is different almost everywhere. Poliwave does not apply one generic model to every country and relabel the output. Each jurisdiction is counted the way its own electoral law counts it, with the divisor, the threshold, the tier structure and the transfer rules that law specifies.
That covers 218 countries and 10,957 jurisdictions, across 21,331 elections going back to 1789. The systems below are the ones actually implemented, each with the number of countries in the archive that use it.
One seat per district, won by whoever leads it. No threshold, no transfer, no second round.
Used in: Canada, the United Kingdom, India, the United States House.
The riding winner is read straight off the projected vote. This is the system the Canadian model was built around, and the one the methodology page describes in full.
Seats are shared out inside each district in proportion to the vote, and voters influence which candidates on a list get them.
Used in: Finland, Brazil, Indonesia, much of Latin America and the Nordics.
Allocation runs on the district's own magnitude with the country's divisor, D'Hondt or Sainte-Laguë, and its legal threshold applied before any seat is handed out.
The same proportional maths, but the party fixes the order of its candidates.
Used in: Spain, Portugal, South Africa, Israel.
Spain is the worked example in the code: per-province D'Hondt with a 3% district threshold, applied province by province rather than nationally, which is why a party can poll well across Spain and still take nothing in Soria.
Two tiers counted side by side. Some seats are won in single-member districts, the rest are allocated proportionally, and the two do not compensate each other.
Used in: Mexico, Japan, South Korea, Italy.
Mexico runs 300 single-member distritos plus 200 proportional seats drawn from five circunscripciones, with the constitutional over-representation cap applied to the coalition rather than to each member party.
If nobody clears the bar in the first round, the leaders go to a runoff.
Used in: France, Brazil and most presidential systems.
The first round is projected, the qualifying rule for that country decides who advances, and the runoff is projected between the survivors. A real recorded runoff is used in preference to a modelled one wherever the archive has it.
District seats are topped up from party lists until the chamber matches the party vote. A party that wins more districts than its share keeps them, which is the overhang.
Used in: Germany, New Zealand, Scotland and Wales under the related additional member system.
Direct mandates are settled first, then the list tier levels the result. Overhang and levelling seats are both computed, and a party below the threshold still keeps districts it won outright, which is how the German Grundmandat rule works.
Multi-seat districts counted by preference. Anyone over the quota is elected and their surplus transfers on at a fraction of a vote.
Used in: Ireland, Malta, the Australian Senate.
The count runs in rounds: quota, election, surplus transfer, then elimination of the lowest candidate, repeating until the seats are filled. Later counts decide real seats, which is why the round-by-round output is kept rather than just the final total.
One seat, counted by preference, eliminating from the bottom until somebody holds a majority.
Used in: The Australian House of Representatives.
Preferences flow through a transfer matrix rather than a fixed guess, so a change in the primary vote changes where the preferences land.
Multi-seat districts where each voter has one vote that cannot transfer, or as many votes as there are seats.
Used in: Hong Kong, Kuwait, and a number of at-large local councils.
Seats go to the leading candidates in order. Vote splitting is the whole story in these systems, so candidates are projected individually rather than by party total.
Alongside the real systems there is a deliberate thought experiment. The simplified simulator has a voting system control that re-runs a single-member election under proportional representation: take a contest actually fought under first past the post and ask what the chamber would look like if the same votes had been counted proportionally. Seats are allocated from the national pool and handed back to the districts where each party polled best, so the map still sums to the national totals.
This is not any country's law and it is not presented as one. It exists because "what would this parliament look like under PR" is a question people genuinely ask about their own election, and answering it with that election's real riding-level votes is more useful than answering it with a national swing.
The methodology page documents how votes are projected: poll aggregation, vote transfers, tactical voting, demographic adjustment and the Monte Carlo error model. That work produces a vote share for every party in every district. Everything on this page happens afterwards, and the two are deliberately independent. Improving the projection does not change how Ireland transfers a surplus, and adding a new country's counting rules does not disturb anyone else's forecast.
Each system is a separate handler with its own tests, including per-country checks against certified results: New Zealand and German MMP, Irish STV, Taiwanese and Hungarian parallel counts, Macau's modified D'Hondt, Hong Kong's SNTV and Sweden's levelling seats. When a country's rules turn out to be more specific than the general case, the rule goes in rather than the result being rounded to fit.
How the projection model works Why Poliwave is different Open the world simulator