How we measure
Every figure on Elevenfold carries a plain sentence saying what it means, a line saying how much it rests on, and the method behind it. This page is the method in full, in the same words the “How this is worked out” panels use.
The four levels
A level is one of four words: Minimal, Developing, Proficient, Advanced. It is read for each part of a topic, each topic, each area and overall, and drawn as the word with four bars beside it, one bar for Minimal and four for Advanced. On the pages a parent reads, the three difficulties of a question are said as the simpler questions, standard exam questions and the hardest questions, and a difficulty that passes the rule below is called reliably right. This page keeps the method's own words: Easy, Medium, Difficult and secure.
Only first tries count: a typed answer the site marked, a Got it or Not yet given after seeing the answer, a worksheet's answers at hand-in, and a mark on a paper, finished or not, timed or not, and a re-sit as it stands now. A part of a paper the child saw and left blank counts as wrong when the child handed the paper in, and not at all when the clock or the site ended the sitting. A second try and a reveal with nothing said never count. A reading uses the last 90 days and, on each difficulty, the latest 30 first tries.
The rule can be checked by hand. A difficulty is secure when it has at least 8 first tries in a topic (6 in a part of a topic), at least three quarters of them right, and a statistical lower bound (the Wilson 80% interval) of at least 60%. The fewest right that is secure: 7 of 8, 8 of 10, 10 of 12, 12 of 15, 15 of 20, 23 of 30. So 7 of 8 is secure and 6 of 8 is not. A secure difficulty makes every easier one secure too.
No difficulty at the floor is Not enough yet. Some at the floor and none secure is Minimal. Easy secure is Developing, Medium secure is Proficient, Difficult secure is Advanced. When Easy has too few answers to read and Medium is at the floor with at least 60% right and a lower bound of at least 40%, the level is Developing all the same, because the standard answers show the basics are in place, and the same holds one level up. The sentence under a level names the difficulty that would move it up, with its counts: "The simpler questions are reliably right (14 of 16). Standard exam questions are not yet: 9 of 14 right, and 11 would make them reliably right."
A level goes up the moment the rule gives a higher one. It comes down only when the difficulty that made it has fewer than the floor left in the window or its share right has fallen under 70%, and then one step at a time. When answers simply age out, the level lapses rather than falls, and the page reads "Not practised enough lately to say · was Proficient in June" for 90 days.
Speed and help count too. A question should take the time the paper allows, until 10 real children of the child's year group have got it right, and then 1.5 times their middle time. When at least 5 right answers at the difficulty that made a level carry a time and the middle one took at least the time it should, the level is held one step lower and carries the word slow. When at least 3 of the right answers there, and at least 33% of them, came after the working was shown or with the worked example open, the level is held one step lower and carries the words with help. Both together still hold it one step, never below Minimal, and neither ever enters a comparison.
Before a level, the site gives an early reading in words from 3 first tries in a topic: "Early reading, from 4 answers: Ada is probably Developing in Fractions." It reads the child's ability from every first try, with right answers given at pace counting fully and slow or helped ones counting less, and says a word only when it is at least 60% likely (75% for Minimal). From 8 first tries it reads "Firming up". The overall early reading needs 12 first tries across 3 topics. An early reading is never recorded as a level, never counted in Since a month ago, never a comparison and never in an email.
A topic's level is the rule on all its answers. An area's level is the highest level that two thirds of its topics with a level reach, from 2 such topics, and the overall level the same across every topic, from 3, always with its coverage: "Ada is Developing. In the 12 topics with a level, the simpler questions are reliably right in 10, and standard exam questions in 2."
Subtopic levels and the limiting subtopic
Each topic is split into its parts (the method calls them subtopics), and questions with no part named sit in the topic's general questions (the method's foundations row). A part is read by the same rule with a floor of 6. The counts under a part are the ones its level was read from. A parent's lists show the parts simplest first: each after the parts it builds on, and among those ready, the one whose questions are easiest.
Under a topic that is not yet Advanced, the site looks inside the same 30 answers that set the topic's level for the part with the most wrong answers. It is named as what holds the topic back when its own share right is under 60% and, with its answers taken out, the rest would be secure: "Reverse fractions is what holds Fractions back: 9 of the 11 wrong answers to standard exam questions were there. Without them the rest of Fractions would be Proficient." When no part passes that test, none is blamed. When a part reads higher than its topic, the topic's line says so.
The diagnosis and the map
Behind the topics is a map of 143 rows and 287 links, drawn from the curriculum: what each subtopic rests on, and what rests on it. Nothing in it is learned from children.
A subtopic is struggling when at least 3 of the last 5 first tries were wrong, when under 60% are right at its deciding difficulty with the floor reached, when it is Minimal, when it lost at least 3 marks and a third of its marks on a timed paper, or when it is slow. A subtopic with fewer than 5 first tries and no paper marks is never called struggling.
For a struggling subtopic the site walks down the map, at most 3 steps. Each thing it rests on is secure, weak or not yet tried. The flaw is the earliest weak step whose own foundations are secure. When everything below is secure, the wrong answers themselves are read for a pattern, in a fixed order: the reverse form, a unit slip, the part taken for the whole, only the Difficult ones wrong, rushing, a return after a gap, verdicts mostly given after seeing the answer, or the method. A subtopic got right only slowly, or only with help, is struggling too. When something below has not been tried, the flaw is unknown and a quick check of 6 questions comes first.
Every diagnosis is four lines with their counts, stored once and printed the same on every page: What happened, What the site checked, What is going wrong, What to do. A subtopic not yet tried whose foundations are weak is called likely to be hard, never weak. It is never recommended and never counted as down.
Recommendations and their kinds
There are four kinds. Go back over the basics: one worksheet, then 8 real questions on the subtopic once the sheet is passed at 80%. More practice: 8 questions at the deciding difficulty, keep going until that many are right. Ready to move on: the next difficulty, or the next subtopic on the map. A quick check first: 6 questions on what has not been tried.
The count is 8, unless secure is within reach, when it is the fewest right that would make it secure plus 2, at most 12. At most 3 are open at once, each with a reason of two sentences at most, and "Set this practice" opens the practice already filled. One not done in 14 days is quietly replaced. Nothing is called missed.
Since a month ago and three months ago
Every time a level changes, the site writes it down with the date. Since a month ago compares the level written on that day, the same day of the previous month, with the latest one, so no past level is ever recomputed. Topics are counted as up a level, stayed the same, down a level, or now has a level, a topic that had no level then because too few of its questions had been answered. A level that lapsed is counted apart, never as down.
Improving but not yet up a level means the same level with the share right at the deciding difficulty up by 15 points, from at least 8 answers on each side, with the two 80% intervals apart. A family whose first answer is inside the month reads that there is nothing to compare yet, and the day the first comparison appears.
The time estimate
"About 25 minutes" is worked out from the time allowed on each question, the child's own pace, and how many questions a keep-going practice will probably run to, plus 2 minutes. The pace is the middle ratio of time taken to time allowed over the child's last 20 timed answers at that difficulty, from at least 8 of them, kept between 0.5 and 2. A new child is estimated at the site's own pace. The estimate is rounded to the minute under 20 and to five minutes from there.
Keep going until 8 are right, and bring back 3
A practice plays its questions easiest first. When the last is answered and fewer than 8 were right first time, more are added from the same pool, never a repeat, up to twice the size and 40 at most, and then it ends as it is. Up to 3 questions from the mistake bank due by that day, from the practice's topics, are placed in it and count towards the 8.
Where the child stands on a paper, and the Elevenfold SAS
Where the child stands on a paper. The group is every real child's first finished timed sitting of that paper. Demo children, re-sits, untimed or unfinished sittings and work done together never count. Once enough children have sat it, the page shows each one as a dot on a line from 0 to 100%, the child's dot in blue, the middle score as a tick, and says "Position 6 of 16 who sat this paper timed", with the middle score and the middle half of the scores. From 100 children the dots become bars and the position gains a band, "in the top 40%".
When more than half the group is within 2 marks of full marks, nobody is placed, because the paper does not separate children. Sittings marked on screen and at home are one group. A paper marked at home can earn marks for the working that a screen marking cannot, so the site watches the gap between the two and names the marking mode in the sentence if the gap on a typical paper ever passes 3 marks.
The Elevenfold SAS. A score that places the child among children of a similar age preparing for the 11+ on this site. Each timed paper score is first allowed for the paper's difficulty, so that different papers share one scale. The middle of the group is set to 100 and each step of 15 points is one standard deviation, the usual measure of spread, and the scale is kept between 69 and 141. About 70 in 100 children fall between 85 and 115.
The group for the Elevenfold SAS is the children in the child's year group who have sat a timed paper here, one figure for each child, once enough children in the year have sat papers here, and it narrows to children within a few months of the child's age when enough of those have sittings. Every Elevenfold SAS is printed with its likely range, the 80% range, which comes mostly from the fact that one paper is one day. The figure over the last 3 papers is steadier. Elevenfold's own score, from children on this site. It is not a GL or CEM score, and no school's pass mark is set on it.
Pass marks and other scales. No school's pass mark is on this scale. Each school sets its own pass mark on the exam board's scale, GL, CEM or its own paper, and it moves from year to year with the children who apply. An Elevenfold SAS is worked out among children preparing here, who are not the same group as one school's applicants or as all children of the age, so 100 here is not 100 from GL or CEM. What the score can tell you is where the child stands among children preparing here, and whether that is moving. For a school's own figures, ask the school.
Children who prepare for the 11+ are, as a group, usually ahead of the national average, so the same child would usually read higher on a national scale. Elevenfold has not measured that gap, so it offers no conversion.
Schools. With a school chosen, the position is read again among that school's applicants on this site, once enough of them have sat the paper, and the child's Elevenfold SAS is printed beside the middle applicant's and the top quarter's.
Children at each level in a topic. The monthly report shows how many children here hold each of the four levels in a topic, from 8 children as counts and from 25 as shares in 100. It is not a bell curve of scores, and it is not called one. No position or comparison is shown on a screen the child sees.
Difficulty-adjusted scores and readiness
Every paper starts with a difficulty read from its questions. Each timed sitting refines it: the share of marks children on this site miss, corrected for how each sitter usually does, blended with the starting figure as if it were 8 sittings. A score is then moved by the difference between this paper and a typical one, which misses 40% of its marks: "This paper was easier than most. Allowing for that, Ada's 55% counts as about 47% on a typical paper." The score allowing for difficulty always comes after the score itself, never instead of it.
The range on Reports (the method's readiness) is the lowest to highest score allowing for difficulty of the last 6 timed full-length papers, first sittings only, finished. Topic practice never moves it, and neither does an untimed or unfinished paper, a re-sit or work done together.
What the site marks, and the hold
In Topic practice a typed answer is marked by the site. After a reveal, the parent's or tutor's Got it or Not yet is the verdict and is kept as given. Worksheet answers are marked by the site at hand-in.
On a paper, every typed answer with a checked answer on file is marked by the site at Hand in. A part with none, such as a drawing, a construction or a table, is answered with the pen and marked afterwards against the school's mark scheme, one tap each (Right, Part or Wrong), by whoever is with the child or by the child. Marks given at the end count exactly as marks on a paper sat at home count, and a sitting with a question left unmarked is never compared with other children. A paper sat at the table is marked at home against the scheme in the same way.
After a timed paper the site names at most 2 topics to work on: a subtopic that lost at least 3 marks and a third of its marks, or at least 2 questions. Their wrong questions are held for 7 days, tied to the practice recommended for them, and never enter the mistake bank when that practice is done. A recap sheet handed in under 80% extends the hold once, by 7 days. Wrong questions in any other topic go to the mistake bank at once.
The mistake bank
The mistake bank is every question a child has got wrong, kept so the child can try it again. Each one comes back after 3 days, then again after 10, and leaves once the child gets it right twice in a row. A question saved for a paper's practice is not in the bank until the hold ends. The questions ready to try again and those in the bank are counted, never scored. An answer whose working was shown before any try never enters it.
The clock, help and lessons
The clock waits at 00:00 until the child presses Start or makes the first mark of work, then counts up. Pause it and the paused time is left out. The seconds are kept with a first try and read beside the time the paper allows, so a child who gets questions right slowly, or wrong in a rush, can be told so. When typing the answer starts the clock, the time counts from when the question appeared, and none is kept past four times the time allowed. Slow or helped right answers can hold a level one step lower. In an untimed practice there is no clock and nothing about time is read. Work marked as together is kept and shown and never counted as right or wrong.
We ask once whether the child answered before the answer was shown when four answers in a row are shown within 8 seconds of the question appearing (or a fifth of the time allowed, when that is sooner), before the clock is started or anything is written on the card. We ask once whether someone is working through the questions with the child when, twice in a sitting, a right answer at a difficulty above the one the child is working at, with nothing written on the card, comes within a fifth of the time allowed after a pause of 2 minutes, or within a quarter of the time allowed after a pause of 90 seconds once three answers like that have come.
A sitting is one child's first tries on one device, each within 20 minutes of the one before. Once you have said the work was the child's own, we do not ask again on that device for 14 days. On the family's account Working together lasts for one sitting, so after a gap of more than 20 minutes it is off again.
If nobody answers, the right answers shown before any try are left out and printed as left out, and you can restore them. A Not yet is never left out, and nothing else is left out on a pattern.
A first try made after How to work it out, with the worked example open, is the child's own work and counts as an answer. A right answer made with the example open counts as help, as one made after the working was shown does: enough of them at the difficulty that makes a level hold it one step lower, with the words with help, as The four levels says, and never raise it. When at least 5 right first tries, and at least half of the right first tries at the difficulty that decides the level and the one below it, came with the example open, the practice recommended is 8 questions without the examples.
Work marked as together at Medium or Difficult, with Working together or in a practice set as together, is used only so that we do not send a quick check on a foundation it has already shown, and look for the difficulty in the topic itself instead. It is never a level, it never takes away a warning that a topic rests on a weak foundation, and it never outweighs the child's own answers. Work marked as together counts for this when it had at least 6 questions on one subtopic, at least half of them right before the answer was shown, and only for a foundation with fewer than 5 answers of the child's own and none on a paper, for 60 days.
Several right answers after the working was shown hold a level one step lower, and a tutor's lesson at Medium or Difficult still spares a quick check on what it rests on.
The demo family
The Demo family is made up: two children with invented history, and the children they are compared with, so every page can be shown filled in. Every page that shows them says so. Nothing about them is ever counted in a real child's reading or comparison.
Reports, the written reports and the Sunday email
Reports is one page: the level with three paragraphs under it (the level in words, what is going well and what is struggling with the clearest cause, and the basis: how many answers since when, what the next level needs and how near the child is), where the child stands among the children who sat the same papers timed, at most 5 things to work on next in the site's order with their reasons, what moved since a month ago (read from the levels the site wrote down at the two dates, never recomputed), the mistake bank, the papers with one graph and one table, and every topic with its parts. Every sentence on it is one the engines above wrote, from a fixed template. The page arranges them.
A written report is the same page as it stood on one date, stored once and never changed. While everything is free, every family is written one every Sunday at 6 pm. When free access ends, it is weekly on Max, written every Sunday, and monthly on Pro and Max, written on the first Sunday of the month. An older report renders exactly as it was stored. The monthly report adds, per topic, Children at each level, a strip of four bars with how many children on this site hold each level in that topic (counts from 8 children, shares in 100 from 25), and a comment that repeats the paragraph's blocks after the level, signed, so the two can never disagree.
The Sunday email carries the top of the page and one link. The link is signed for that child and opens Reports read only for 7 days without a sign-in. A forwarded link shows the report and can set nothing, and every link can be ended from Settings, Notifications. When free access ends, on Pro the weekly email is six lines and one link, and the written weekly report stays part of Max.
What the site never does
The site never shows a percentage of a child's own accuracy, only counts and one figure on a finished piece of work. It never records an early reading as a level, never moves the range of timed papers with practice or with an untimed or unfinished paper, never compares a child with children who sat different work, never shows a position or a comparison on a screen the child sees, never writes a free-text judgement, and never awards points, streaks or badges.