Mental Math Practice: How Many Can You Solve in 40 Seconds?

How to play

One round of this mental math practice lasts 40 seconds. Think of it as a set: as many clean reps as you can before the clock runs out.

1. Pick a mode. Test mode uses the same fixed problem mix every time and counts toward your records. Training mode starts at your level and moves you up or down after each round.
2. A problem appears in big numerals. Type the answer on the keypad below it. There is no submit button: the moment your answer is right, the next problem appears.
3. If you’ve typed as many digits as the answer has and it’s wrong, the answer box flashes red with an ✕ and clears. The problem stays, so try again.
4. Stuck? Tap Skip. The right answer shows for 0.6 seconds, tagged Answer, and the problem counts as wrong.
5. When the 40 seconds are up, Time! flashes over the problem and your score is the number of correct answers.

From training level 9, where an answer can be negative, a minus key appears on the keypad.

Test mode always uses the same recipe. About 80% of the problems are basics, split 40 : 30 : 30 between two-digit ± one-digit (47 + 8, 62 − 5), times tables from 2 to 9 (7 × 8) and two-digit minus two-digit with a positive answer (83 − 27). The other 20% or so are harder kinds from Training mode, in equal shares: two-digit × one-digit (23 × 4), two-digit + two-digit (38 + 47), division with no remainder (56 ÷ 7) and three terms with an answer of zero or more (14 + 9 − 6). The first three problems of every run are always basics. Because the recipe never changes, today’s score and next month’s measure the same thing.

What mental math practice measures

Researchers call this skill arithmetic fluency: how quickly and accurately you do basic calculations without paper. It leans on two things. One is stored knowledge, like knowing 7 × 8 = 56 without counting. The other is working memory, the mental sticky note that holds “carry the one” while you finish the sum.

Timed arithmetic tests are a standard research tool. A recent example is Math4Speed, a free, standardized test that gives you two minutes for each of the four basic operations. It was developed with 1,165 young adults in six European countries.

A few honest limits:

• Practice improves the practiced task. A large review of brain-training research confirmed that, but found little evidence that gains carry over to other tasks or daily life (Simons et al., 2016). Expect to get faster at these problems, not a higher IQ or better grades.
• It’s a workout log, not an assessment. Not a math placement test, a school exam or a medical check.
• Phones add delay. In robot-timed lab tests, touchscreen browsers added roughly 58–70 ms to every tap (Pronk et al., 2020). Each answer takes one to three taps, so over 40 seconds it adds up. Compare runs on the same phone.

Averages and ranges

There is no published norm for this exact format: 40 seconds, a phone keypad and this problem mix. Math4Speed uses two-minute blocks with one operation at a time, so its numbers don’t transfer. Peer comparison may come later. Until then, the benchmark is you.

What you see while you play and after:

Training mode has ten levels:

Behind the scenes the game also counts your accuracy: correct answers ÷ everything you tried, with wrong entries and skips included. It isn’t shown, but it decides your level. Get 13 or more right at 85% accuracy or better and you move up a level. Fewer than 7 right, or accuracy under 60%, and you move down one. Otherwise you stay. With Large text & slower pace on, the counts drop to 10 and 5. The 85% line borrows from a 2019 paper that found learning algorithms improve fastest at about 85% correct (Wilson et al.). That comes from models, not people, so treat it as a rule of thumb.