Mental Math Trainer
Timed mental arithmetic drills with a daily streak. Tells you which facts are actually slow — and refuses to time a question you were away for.
Mental Math Trainer
Ready?
7 + 8 =
Done
Worth another look
How to Use the Mental Math Trainer
Pick the operations
Start with one. Mixing addition and multiplication in the same drill trains switching between them, which is a separate skill from knowing either.
Choose a difficulty
The right band is the one where you get most answers right but still have to think. If you are at ninety-five per cent and instant, move up.
Answer with the keyboard
Type the number and press Enter. Keeping your hands still matters — reaching for a mouse adds a second to every answer and hides your real recall speed.
Read the weak facts, then come back tomorrow
The facts listed at the end are the ones you got wrong or were slow on relative to your own pace. Short and daily beats long and occasional.
What a Drill Can Measure, and What It Quietly Gets Wrong
The point of practising arithmetic facts is not to become a faster calculator. It is to stop calculating. A child who works out 7 × 8 by adding eight sevens will get the right answer and will still be unable to follow a lesson that uses it as a step, because the working occupies the attention the lesson needs. Recall frees that attention. This is why speed is worth measuring at all: it is the only signal available in a browser that distinguishes a fact you know from a fact you can derive.
Which makes the timing the part worth getting right
And it is the part most drills get wrong, in three ways that all point the same direction — flattering.
The first is timing wrong answers. If a wrong answer's clock goes into the same pile as the right ones, then hesitating, guessing and being wrong lowers your average response time, because a guess is fast. The number improves as the session gets worse. Only correct answers are timed here.
The second is the mean. Response times are not symmetrically distributed: they have a floor around human reaction time and no ceiling at all, so the occasional huge value is guaranteed rather than exceptional. One interruption in twenty questions moves an average by seconds. The median ignores it, which is the entire reason to use one.
The third is the interruption itself. A tab switch does not slow you down — it stops you existing, and the clock keeps running. Most drills record that as a very slow answer, which is not a description of anything.
Browsers will tell a page when it is hidden. So a question that was on screen while you took a call, replied to a message or locked your phone is scored — you were still right or wrong — but not timed, and the tool says how many it dropped. That is a small thing that changes what the median means, because the alternative is a statistic that mostly measures your notifications.
What is actually hard in a times table
A 12 × 12 grid has 144 cells, which sounds like a great deal to memorise and is not. Multiplication is commutative, so 7 × 8 and 8 × 7 are one fact rather than two, which halves the grid immediately. The 1 and 10 rows are rules rather than facts. The 2 and 5 rows arrive early and stick. The 11 row up to 9 is a pattern anyone spots. Strip those out and what remains is roughly twenty products that must genuinely be learned — and they are the ones people are still slow on decades later, with 7 × 8 the reliable worst case: no rhyme, no trick, two awkward operands.
That is why this tool reports individual weak facts rather than a single score. A drill that says "82%" tells you nothing you can act on. A drill that says you were slow on 7 × 8 and 6 × 9 tells you what tomorrow's two minutes should contain.
A regional note on drilling, and its limits
Timed arithmetic practice carries particular weight across this region. Singapore's PSLE and Malaysia's UPSR-successor assessments both reward fluency under time pressure, and the after-school culture built around them — kumon-style repetition, tuition-centre drilling — treats speed as a proxy for competence. It is a reasonable proxy at the recall level this tool operates on, and a bad one beyond it.
Worth saying plainly, then: nothing here trains problem-solving, and fluency in recall is a floor rather than a goal. Singapore's own primary mathematics curriculum is built on the concrete–pictorial–abstract progression and on model drawing — approaches specifically designed to develop reasoning about a problem rather than speed at the arithmetic inside it. A trainer like this one is a support for that, in the same way that knowing vocabulary supports reading. Practise until the facts are automatic, and then spend the time on the part that is actually mathematics.
10 Facts About Mental Arithmetic and Timing
A drill only measures recall if the answer comes from memory rather than calculation. That is why response time matters as much as accuracy — a correct answer that took nine seconds was worked out, not recalled.
This trainer reports the median response time, never the mean. One interruption adds a forty-second answer that moves a mean by seconds and a median not at all.
A wrong answer is not timed. The clock on a wrong answer measures how long someone took to be wrong, which is a different quantity — and mixing the two makes a bad session look fast.
Neither is a question you were away for. Browsers report when a tab is hidden, so a question that was on screen while you took a call is scored but not timed.
The times tables stop at 12 for historical reasons, not mathematical ones — twelve pennies to the shilling and twelve inches to the foot made the twelve times table a daily necessity in Britain.
There are 144 entries in a 12×12 table, but far fewer facts to learn: commutativity halves it, and the 1, 2, 5 and 10 rows are close to free. What remains is roughly 20 genuinely hard products.
7 × 8 is reliably the hardest single product in the table for English-speaking learners — no rhyme, no pattern, no finger trick, and two operands that are themselves awkward.
Every problem here is drawn without replacement, so a drill never asks the same question twice. Drawing 30 problems independently from a pool of 100 repeats one about 99 times out of 100.
Subtraction problems are generated backwards — pick the answer and the smaller number, then add them — which is why you will never be shown a subtraction that goes negative.
Division problems are built by multiplying first, so they always divide exactly. A drill that hands a child a remainder they have not been taught is testing the tool, not the child.
Frequently Asked Questions
- The question still counts towards your score, but it is not timed. If the page was hidden while a question was on screen, the elapsed time is a measure of how long you were away rather than how long you thought — and a single forty-second record is enough to drag a median and to label a fact you actually know as one of your slow ones. The tool notices, drops that one clock, and tells you it did.
- Because an average is wrecked by exactly the thing that happens most often in a drill: one distraction. If nineteen answers take two seconds and one takes forty, the average says four seconds and the median says two. Two is the honest description of how fast you were.
- A fact is flagged if you got it wrong, or if it took you meaningfully longer than your own typical answer — not longer than some fixed threshold. A six-year-old and an adult have very different baselines, and a fixed cutoff would tell one of them everything is weak and the other that nothing is. It also needs to have seen the fact at least twice, so one bad moment is not called a weakness.
- No — each session draws a fresh set. Within one session no question repeats, because problems are drawn without replacement rather than picked independently each time. If you want a fixed, reprintable set instead, that is what the Math Worksheet Generator is for: it prints a code that regenerates the identical sheet later.
- No. Subtractions are built by choosing the answer and the number being subtracted, then adding them to get the starting number — so the result is never below zero. The same construction is why division here always comes out exactly, with no remainder: the problem is built by multiplying first.
- One tap that sets the drill to multiplication and division with operands from 2 to 12 — the tables as they are actually taught. Including division matters: knowing 7 × 8 = 56 and being able to answer 56 ÷ 8 quickly are different skills, and drilling only one direction leaves the other slow.
- In calendar days, and a day counts once however many sessions you do. It uses your own device's date, and the arithmetic is done in a way that a daylight-saving change cannot break — the usual bug is a clock shift making two adjacent days 23 or 25 hours apart and silently ending a streak.
- No. Your settings and the list of days you practised are stored in your own browser, nothing else is kept, and no answer you type leaves the page. There is no account and no analytics on your results. Clearing your browser data clears the streak.
- In a timed session the countdown is worked out from the actual clock rather than by counting timer ticks, so leaving and coming back does not gain you time — browsers slow background timers to roughly one tick a second, and a tick-counting clock would quietly under-report. Any question that was open while you were away is scored but not timed.
- They set how big the numbers get, and the age hints beside them are a rough guide rather than a syllabus mapping. Deliberately so: tying bands to a named curriculum would mean re-checking the claim every time a ministry revised one. Pick the band where you are getting most answers right but having to think.
Related News
You may be interested in these recent stories from our newsroom.
No related news yet for this tool. Our editorial team publishes new pieces every week.
Browse all news →Pick up where you left off
Stored only in this browser — never sent to our servers.