Data sources

Counting the records…

• Every number is calculated from the Knesset's records: plenum votes, bills, transcripts and positions.
• The same rules apply to all 120 members.
• Next to each figure is a small mark that leads to its source.
• Numbers, votes and quotes never pass through an AI model.
• Quotes are copied as they are, without editing.

How many of the plenum votes held while the member served they took part in. The 25th Knesset held 7,557 plenum votes.

For each vote we check how the majority of the member's party voted that day. Only for and against count, and only when at least two party members voted.

Bills the member initiated that passed a third reading in this Knesset.

The share of votes in which two members voted the same way, out of the votes where both voted for or against. Only pairs with at least 150 shared votes are shown.

Calls to order and removals from meetings, from 9,331 committee transcripts. Quotes are copied as they are, with no score or ranking.

Only for and against answers count. A member who did not vote on a law is left out of that question, neither for nor against them. Anyone compared on fewer than half the answers is not ranked. A party's figure combines all its members' votes.

Night vote: between 23:00 and 06:00. Rushed: up to 7 days from the first reading to the final vote. Omnibus: amends 5 or more laws. Few hands: 10 or fewer members in the final vote. Pile-up: 15 or more laws passed the same day. Temporary provision: by the Knesset's flag or the law's name.

A party counts as coalition on a given day if one of its members held a government post that day, minister or deputy minister. That is how each vote is split into coalition and opposition.

Lists on the site are ordered by a relevance score from 0 to 100. On the home page, laws are ordered by their importance (60%) and how recent the final vote is (40%), a weight that halves every 45 days, and members by the average importance of the laws they voted on. A law's score is built from its importance (45%), how recent its final vote is (25%), how close the vote was (15%), its "under the table" signals (10%) and how many members took part (5%). Importance is the highest of three: the 1 to 5 reading, how close and how full the final vote was, and a Basic Law, which always counts as most important. A member's score is built from their office (50%), the size of their party (15%), the laws they got passed (15%), their attendance at votes (10%), their votes against the party majority (5%), and calls to order and interjections (5%). A member's vote takes the law's score, plus 20 points when it went against the party majority and 10 when the law was their own bill.

Law summaries, and who gains and who pays, are written with a language model that reads the text of the law. Every line points to a section, and a line whose section is not in the text is dropped. Numbers, votes and quotes never pass through a model.

Every law and bill gets an importance score from 1 to 5. An AI model reads what the law does and compares it with the same fixed examples every time, so the scores stay consistent. The levels:

1. Mundane things everyone can agree on, with no real impact on people
2. Policy changes with an indirect effect on people
3. Decisions that directly affect a group of people
4. Decisions with a direct impact on the average person
5. Critical, country-changing decisions