How your match with the candidates is calculated

This test belongs to a family of tools known as voting advice applications. The idea is always the same: take the candidates’ positions on a list of issues, ask for your opinion on the same issues and measure the distance between you and each of them.

It is worth being clear about what it is not. A poll asks a representative sample what people think or how they will vote, and extrapolates to the whole population. This test extrapolates nothing: it only tells you, individually, how far you are from each candidate. The result is neither a forecast nor a voting recommendation.

Why counting matches is not enough

The simplest method would be to count the questions on which you agree with each candidate. The problem is that it treats two very different situations the same: being one point apart (you “strongly agree”, the candidate “agrees”) and being four points apart (you at one end of the scale, the candidate at the other).

The test therefore applies three corrections. The first penalises large distances more. The second gives more weight to your firm opinions than to lukewarm ones. The third, in the full test only, takes into account how much each issue matters to you.

The calculation, question by question

For each statement you answer:

Step 1. The distance. The candidate’s position (1 to 5) is subtracted from your answer (1 to 5) and the absolute value is taken. The result is between 0 and 4.

Step 2. The base score. It is calculated as (4 − distance)². Squaring makes each extra point of distance cost more than the one before:

Note the jump: being just one point apart already costs almost half the score.

Step 3. The conviction factor. It is 0.5 + (distance from the centre × 0.25). If you answer at either end of the scale (“strongly agree” or “strongly disagree”) it is 1; if you answer “agree” or “disagree”, 0.75; if you stay in the middle (“neutral”), 0.5. A neutral answer therefore weighs half as much as a firm one, and an “agree” or “disagree” answer three quarters as much.

Step 4. The weight. The confidence weight of the candidate’s position (currently 3 for all positions, estimated ones included) is multiplied by the importance you give the issue (1 to 5 in the full test; always 3 in the quick one) and by the conviction factor.

Step 5. The percentage. For each candidate, the points from all questions (base score × weight) are added up and divided by the maximum that could have been obtained. That ratio, times one hundred and rounded to one decimal place, is your match.

A real example

Take the statement “There should be a minimum tax of 2% on fortunes worth more than 100 million euros.” This is the so-called “Zucman tax”: the National Assembly passed it at first reading in February 2025, then rejected it during the debate on the 2026 budget. Suppose you answer 4, agree, and give it an importance of 4 (“quite a lot”).

Your conviction factor is 0.75, because you are one step from the centre. The weight is 9 for every candidate (confidence 3 × importance 4 × conviction 0.75), and the maximum on this question is 16 × 9 = 144 points.

Had you answered 5, strongly agree, the conviction factor would be 1, the weight 12 and the maximum 192. The order would change: Jean-Luc Mélenchon and Marine Tondelier 100%; Fabien Roussel and Raphaël Glucksmann 56.3%; Nicolas Dupont-Aignan 25%; Gabriel Attal, Marine Le Pen, David Lisnard, Bruno Retailleau and Éric Zemmour 6.3%; Édouard Philippe 0%. The same opinion, held more firmly, separates the candidates more and weighs more in your result.

All eleven positions in this example are documented, and their sources are listed at the end of this analysis. Some come from the candidate personally (their programme, statements or vote as a member of the National Assembly); others, as the methodology provides while not all 2027 programmes have been published, come from the vote of their party’s group in the National Assembly or from an earlier programme of their party. Each candidate’s page says which.

Two questions together

Points add up across all questions, which is why the final percentage is not “the share of questions on which I agreed with the candidate”. Let us add a second statement to the previous answer: “The legal retirement age should go back to 62 or lower.” The 2023 reform raised the legal retirement age from 62 to 64; it is suspended until January 2028. You answer 1, strongly disagree, but the issue matters only 2 to you (“a little”). The conviction factor is 1, the weight 3 × 2 × 1 = 6 and the maximum 16 × 6 = 96 points. Together with the first question, the maximum is 144 + 96 = 240.

Fabien Roussel and Bruno Retailleau each match you exactly on one question out of two, and yet Fabien Roussel scores 5 points more: his exact match is on the question you consider more important. Raphaël Glucksmann moves ahead of Fabien Roussel because, on retirement, he is three points away from you rather than four: even a large distance is worth a little more than the maximum one. Édouard Philippe matches you on retirement but is three points away on the tax, hence his 43.8%. We checked these figures with the same code that calculates the test result.

The quick test and the full test

The full test has 46 statements, and for each one you can set an importance from 1 (“not at all”) to 5 (“a lot”); if you leave it unchanged, it counts as 3. The quick test has 20 statements, and importance counts as 3 for all of them, so only distance and conviction matter. Otherwise the calculation is the same.

All 11 candidates in the test are included in the calculation. Each has a position on every statement; of the 693 positions in the database, 562 link to the document they are based on and the other 131 are marked as estimated on the candidate’s page. In the calculation they all carry the same weight. Your age, your sex and the candidate you would vote for, if you choose to give them, have no effect on the result.

Topic by topic, and the compass

Your result also shows your match with the candidates on each topic. It is calculated with the same formula, using only that topic’s statements: if, say, you answered the two statements on health, the percentage for that topic is based on those alone and on the importance you gave them.

The political compass is calculated differently. It does not compare your answers with the candidates’ positions; it turns each answer into a number from −100 to +100, depending on which way the statement pushes, and averages them on each axis. Neither importance nor the conviction factor plays a part, and statements on Europe, foreign affairs and defence, and institutions do not count: 35 of the 46 statements move it (15 on the economic axis, 20 on the social axis). So the candidate with the highest match may not be the one closest to you on the compass: your match takes all 46 statements into account, the compass only 35.

Three practical consequences

Answering “neutral” is not neutral. It halves the weight of that question, so you give up half of its power to tell candidates apart. If you have an opinion, say so.

Skipping questions is perfectly fine. Questions you leave unanswered (“Don’t know / Skip”) are excluded from both the numerator and the denominator: they do not count as disagreement. The same would apply if a candidate had no position on a question.

Ties make sense. If two candidates share a position on most of the statements where you took a clear stand, their percentages will be very close. That is not a flaw in the calculation: on those issues they are alike. That is why the test has 17 tie-break questions that are not part of the main test. If the second or third candidate in your result is 3 points or less behind the first, the test offers you up to 8 of them: the ones that separate the tied candidates most. Only questions on which all the tied candidates have a sourced position (never an estimated one) are chosen, and only if at least one of them separates the candidates by 2 points or more. Your answers are added to the usual calculation for all candidates, with an importance of 3. You can answer the tie-break questions only once, and they do not change your compass. The details are in the methodology.

What this number does not measure

It does not measure whether a candidate keeps their promises: positions come from programmes, votes and statements, not from an assessment of their record. Nor does it measure your trust in the person or their ability to carry out their programme: what a president can do also depends on the majority they have in the National Assembly. And the list of candidates will change: the official list will be drawn up by the Constitutional Council after checking the 500 sponsorships, and after the first round the test will keep only the two candidates who go through to the second. How the election works is explained in the analysis How the presidential election works.

If you find a position you think has been assigned wrongly, the methodology explains how to tell us. We check it against the source and correct it if needed.