Corrections

Anyone elected to the 26th Knesset now shows as a member of the 26th Knesset, not a candidate: on their page, in the profile panel, on the share card and on the 25th Knesset MK page. On the list slates, the poll-average seat line is replaced by a line at who was actually elected, and the "only who gets in by the polls" filter by "only 26th Knesset members". "The list in numbers" now counts the elected MKs. The bottom bar shows "MKs" instead of "Lists".

A new page with the 26th Knesset's 120 members by the vote count: a half circle of the seats, how many are new, from the outgoing Knesset or back after a break, and each list's members in order. Until the official results it is an estimate without surplus-vote agreements. From 99% of the votes counted it is linked from the home page, under Said / Did.

With the count over, the home page shows the count results and then the 26th Knesset: a half circle of the seats where a tap shows who holds a seat, and each list's members. The TV exit polls were removed. Turnout and election updates stay on the results page. The name of Yashar!'s no. 18 was corrected to Israel Zeray, and photos were added for several members.

The cards of 26th Knesset members who sat in the outgoing Knesset now show vote attendance, bills and votes against their faction, by the same method as each MK's page. Ministers show their role. Each number at the top (new, from the outgoing Knesset, back, women) opens into the list of people.

The home page moved to the post-election period. Until the final official results, the results page leads it; after that, a link to it. Under the results there is a link to Said / Did. The quiz, the last poll average before the vote, the coalition calculator and the list pages all stay: in the "26th Knesset election" card on the home page, in the menu and in the data hub. The Knesset games moved to the data hub. The bottom bar and the button at the top of the page now show the results instead of the quiz and the election-day page.

A new page gathers election night and updates by itself: turnout through the day from the Elections Committee and, from 22:00, the TV channels' exit polls (each channel, their average and a comparison), the count and key updates. The numbers are entered by hand from the sources. On election day and the two days after, it leads the home page and the top bar links to it. Also added: turnout by town from the Elections Committee's station file, with search, and the number of voters out of those eligible.

A new page with the lists' answers to the 15 for-or-against questions in Globes' economic questionnaire, in the lists' own words and with a link to each full answer. Lists that did not answer are marked in every question. Each question keeps a place for the 26th Knesset's votes, to be added when they happen.

Each of the 16 subjects of "Where each party stands" now has a page showing, for every list, what it said (platform, statements and posts) next to what its members did in the 25th Knesset (votes, bills and recorded acts). The evidence and placements are the same as on the subjects page. A list with no statement or vote found on a subject is marked as such.

For Likud, Shas, United Torah Judaism and Ra'am, which did not answer Globes' economic questionnaire, each of the 15 questions now shows our reading of their votes, bills and words, with a level of evidence and a source for each item, and says so when no position was found. Five subjects voters named as decisive were added: an Arab party in the government, a Palestinian state, the day after in Gaza, preference for those who served, and public transport on Shabbat, with all 13 lists' positions and sources.

From the moment the polls close, the poll average stays the last one before the election and does not change. Exit polls and results are not added to it. Pages with poll numbers say so in one line, with a link to the official results, and the home page title reads "The last average before the election".

The election day page said the Elections Committee had not yet published the special accessible polling stations. The Committee has published them on its official map, and the page now points to it. In the same update: some dates on this page showed as code instead of a date. Also, "Where to vote" now says that people with a mobility limitation can also vote at any special accessible polling station. The page now lists the special accessible polling stations by town, from the Elections Committee's polling-station file. In towns with more than ten such stations, you can also filter by street or place name. The stations are also shown on a map, with a search for the nearest ones to your location or an address, using the Committee's polling-station cluster file. In the last days before the election, the home page and the countdown bar on every page link to the election day page. The election day page's share image was also updated: it shows the special accessible polling stations on a map of the country. After a search by location or address, the page now scrolls to the results.

Under the Elections Law, from the end of the Friday before the election until the polling stations close, no new poll may be published, and anyone showing an older poll must say prominently that it is not current. Until the polls close, such a notice appears at the top of every page that shows poll numbers, the quiz result does not show the poll average, share images either show no poll seats or carry the notice, and no poll conducted after that Friday is shown.

The October 12 rule (5 questions with a 50% match added to every list) was replaced by a stronger, more targeted one: every question you answered on which a list has no documented position counts as half a question with a 50% match. A list with a position on every question you answered is not affected, so a full match shows as 100% again. The rule is the same for every list.

We searched again for documented positions on every question where a list has none. One new position was found, in one list's platform, on the question of police and Shin Bet powers against crime organisations, and it was added. On the other questions we found no clear statement by the list itself, so they stay without a position.

The WhatsApp share button now looks the same across the site. On "Where each list stands", the evidence-strength mark (one to three bars) is easier to read, and on phones the explanation at the top of the page has no frame and the counts moved into "How exactly a placement is decided". On the quiz result, "This is a comparison, not a recommendation" appears once, in the result card. No number changed.

On phones, the bottom navigation bar was transparent and page text showed through it; it now has a background. On the quiz result, the list name and percentage are larger, "This is a comparison, not a recommendation" appears inside the result card, and the WhatsApp share button is less prominent than the list's own button. On phones, the three home page cards (who is running, where they stand, how to vote) moved below the poll average, the breadcrumb trail at the top of pages was removed and the coalition calculator's intro is smaller, so the lists start higher. No number changed.

A vote against a private bill from the other bloc in a preliminary reading was counted at low weight. From now on it is not counted at all when the list cast a direct vote the other way on the same question, on a binding bill, without its faction splitting. Two placements changed as a result: on the sovereignty question and on the question of powers against crime organizations.

In the questions on the draft, the inquiry commission, judicial review, sovereignty, term limits, coalition funds, tax benefits and police powers, words taken from one side's vocabulary or that assume an answer were removed, and the sovereignty question now gives a reason for the side in favor too, as it already did for the side against. The lists' placements did not change.

On the previous question, every list with a known position split exactly along bloc lines, the same split as four other questions in the quiz, so it did not tell lists apart. The new question asks whether to repeal the Nation-State Law, amend it, or keep it as it is, with amending as the middle. In the only vote on repealing it in the 25th Knesset, lists from the same bloc voted differently from each other. Each list's placement rests on that vote, on the six votes on adding an equality clause to the law, and on documented quotes, each linked to its source.

In each question, the names of the two positions now appear right above the end of the line they belong to. On phones each name gets its own line next to its end.

The games now draw the most interesting material from the same data. In the tweet game: the posts that travelled furthest compared with the same MK's usual posts, without mourning posts, greetings and interview plugs, with a cap per MK and per party and a minimum for every party. In "Who has more?": each round puts a familiar face against an MK whose number is clearly different, not two near-identical numbers. In "Who filed this bill?": no bills that several MKs each filed separately on the same subject (where more than one answer is right) and no technical titles. In trivia and facts: fewer repeated questions and fewer "second place" facts. No number changed; only the choice of what is shown.

The statements, interviews, platforms and tweets of every list's candidates since 4 October were checked again. 16 documented quotes and 53 tweets that take a position were added, each linked to its source. Four placements of four lists moved by one step, in the direction the new evidence points. Quotes that could not be verified word for word, or that appeared only in a single partisan outlet, were not added.

The "Without outlier polls" switch, until now only in the coalition calculator, is now also on the poll average page and in the home page average. It is off by default, so the regular average, with every poll, stays the number shown. A poll counts as an outlier by an automatic calculation (a list's number is 4 or more seats from the other pollsters' median), never by picking a poll or pollster by hand.

Lists that the poll average puts under the electoral threshold got 0 seats, but their seats went to no other list, so even with every list picked the counter reached 115 rather than 120, while every single poll fills the Knesset. Those seats are now shared among the other lists by size, as in the election itself, so the average also has 120 seats.

The menu and the bottom of every page now have a button that sends a link to the home page, alongside the share buttons of pages and results.

The match was an average over the shared questions only, so a list with a documented position on few questions could come first even when another list matched almost as well on many more. The percentage is now computed as if every list had 5 more questions with a 50% match, so matching on many questions counts for more. The rule is the same for every list, and about one result in eight changes. As a result, even a full match on all 16 questions shows as 88%.

The site's story, numbers to quote, downloadable images and contact details, plus the poll average as a widget news sites can embed. From the end of the Friday before the election, the widget shows the warning the law requires on old polls by itself.

Until now only Knesset votes could place a list at "Strongly". A list with no votes on an issue was shown at "Leaning" at most, even when its platform stated a clear position, so it looked closer to everyone than it is. The rule was updated for all lists alike: a list's explicit platform, together with other evidence in the same direction, can place it at the end of the axis. The change moved 11 placements of 6 lists, and 5 documented quotes were added. Every placement is shown with the evidence it rests on.

Terms such as "sanctions", "commission of inquiry", "sovereignty" and "core studies" appeared in the questions without an explanation. They now have a dotted underline, and tapping one opens a short, neutral explanation.

To see why the top list came first you had to scroll to the full table. Now a "Why this list?" button next to the result opens its evidence, and the table shows the top 5 lists with an option to open all of them.

A question marked ★ "matters most" counts twice in the match. Without a mark, all questions weigh the same.

You can compare 2 to 4 lists side by side on the 16 questions, choose to show only the questions where they differ, and share the comparison as a link.

Every list page now shows the share of women, the share new to the Knesset and the average age in the realistic places, next to the same figure for the Knesset the polls predict. For a list below the threshold in the polls, the figures are based on its first 4 places.

On phones 360 to 390 pixels wide, after the language button was added, the page header ran past the screen width. On phones the menu button is now shown as an icon only, and everything fits.

Since October 8 the quiz answers and the closest list have been saved anonymously, as the privacy policy explains, but the methodology page still had the old sentence saying they are not saved or sent to the server. The sentence was corrected.

The page stated that any ballot slip "with anything written on it" and any "blank slip" is invalid. In fact, a blank slip on which a list's letters are handwritten is valid - and that is the way to vote when the list's slips have run out. The wording was corrected, and a note was added that an expired identity document is not accepted.

Corrected: the Bader-Ofer method (a variant of the D'Hondt method that gives a slight advantage to the larger lists); surplus-vote agreement (possible only between two lists); the ballot slip (carries the list's ballot letters and name, not the candidates' names); private member's bill (ministers and deputy ministers do not submit them); committee bill (only on subjects the Knesset Rules of Procedure allow); continuity, immunity, caretaker government and invalid ballot slip.

The data hub also counted reserved slots on the lists, which are not candidates, and one list page counted them in its heading. All pages now count only candidates.

Pages that are built in advance showed the number of days as of the day they were built, and the share image of the election day page did not update at all. The countdown now updates on every load.

BeYachad is the continuation of the Yesh Atid faction, and The Democrats are the continuation of the Labor faction, with a reservation regarding the partner that did not serve in the 25th Knesset. In addition, the wording "the list did not serve in the 25th Knesset" was replaced with "it had no faction in the 25th Knesset", which is true of all the new lists.

On the poll average basis, a list whose average was below the threshold received seats and could be added to a coalition. It is now counted as 0, as in any single poll that puts it below the threshold.

The answer "In the middle / Not sure" was counted as the midpoint of the axis and affected the match. Now "In the middle" is a position, and anyone who is not sure skips the question, which is not counted.

It said that a language model is used only for bill summaries. In fact it is also used to identify which side a published text takes (a platform, statement, tweet) and to determine the direction of a vote in relation to disputed questions. The details, and the safeguards that keep it from introducing bias, were added. The claim that Knesset data is updated "in real time" was also corrected to "periodic sync".

Added: the preferences stored only in the browser, hosting on Vercel, the address search and map from OpenStreetMap, and images loaded from external sites.

The category also included people who simply were not found on the candidate lists the site documented. The label was changed to "not found on the lists we documented".

Dudi Amsalem was shown as "Minister - Ministry of Justice" and Ze'ev Elkin as "Minister - Ministry of Finance", alongside the actual ministers (Yariv Levin and Bezalel Smotrich, respectively). The reason: both serve as "additional minister" in those ministries (a recognized government role, different from actually holding the portfolio), and the site did not distinguish between the two. Fixed both in the sync code and in the existing data.

The parties page showed one number of seats, while the MKs page and the party's share image showed another. The reason: the higher number also included ministers without a Knesset seat (under the Norwegian Law), not only actually sitting MKs. Both places now show the real number of seats, with a separate breakdown for ministers without a seat.

The "Who filed this bill?" game (formerly "Who initiated this law?") and its description called the items "laws", while most of them are bills that did not pass. The game's name and description were corrected to reflect that these are bills, not laws that were passed.

An earlier performance improvement removed the full list of votes from the general API response for an MK profile (to reduce page weight), but the comparison page kept relying on that same API to calculate the agreement percentage - so the calculation always came out "0 of 0", while the share image (calculated separately, on the server) showed the real number. Both places are now calculated from the same single function on the server (through a dedicated endpoint), so the page and its share image can no longer contradict each other.

Following the September 28 fix to /bill/2219605: it turned out that the problem was not specific to one motion but a pattern that recurs in every vote in which the Knesset decides whether to accept the government's response to a debate - the official title of the debate is usually critical wording submitted by the opposition, while the vote itself is narrower. All 8 votes of this kind found in the database (including 2219605) are now tagged in a separate category, and every bill page of this kind - its heading (H1), page tag, meta description and share image - now shows the question that was actually decided, with the original debate title as a quote for context only.

The birth year shown for him was taken from a Wikidata record that was mistakenly matched to another person with the same name (a labor minister from an earlier generation, born in 1927). The birth year was corrected to 1955, and an automatic plausibility check was added to prevent a similar error from recurring.

Several facts showed two different MKs or two different parties as 'having the lowest value' in the same category (for example voting attendance or number of bills). The cause was an error in calculating the ranking. Fixed so that only the true lowest value is tagged as such, and so that the 'two lowest' list is not generated when the sample is too small to be meaningful.

A technical query meant to fetch the bills tagged to each topic failed silently for every topic and every party, because of a length limit on the request to the database. As a result, the party pages showed empty lists even though the data exists. Fixed, and the actual bills are now shown.

The page for motion 2219605 showed the Prime Minister as voting "for" a motion titled "ראש הממשלה מוביל ממשלה כושלת" (roughly: the Prime Minister is leading a failing government), and the motion as "passed". A check against the official Knesset source showed that the actual vote was on accepting the government's statement (the government's response to the debate), not on the content of the title itself - that is, the Prime Minister voted for accepting his own government's response, not for the claim in the title. The vote label was corrected so that it does not read as a vote on the content of the title.

A link to the home page with certain parameters made it possible to inject free text into the title and image shown as the preview when sharing (for example on WhatsApp), so a link could be created that looked as if it came from the site with any claim one wanted. The hole was closed - the preview is now built solely from real data on the server and does not accept free text from the link.

A check against the actual data showed that almost all MKs, in every party, vote with their faction's majority in 97%-100% of cases - so the measure did not distinguish one member from another and had no real informational value. It was removed entirely from the MK page, the MK comparison and the share image.

The game's mechanics let a player win again and again without limit by repeatedly choosing the same side, if the MK on that side happened to have the highest value in the category. Fixed so that every round draws both sides anew.