Most mental math advice is wrong

Walk into any bookstore and you'll find the same shelf: how to multiply four-digit numbers in your head, square anything ending in 5, calculate the day of the week for any historical date. The video titles are always some variation of "the trick they never taught you." It's entertaining, and in our experience not how people actually get better.

We built /train around a different premise. Most advice circulating about mental math is wrong in four specific ways, and every deck in the catalog is a reaction to one of them.

Tricks without structure don't transfer

Take the multiply-by-11 shortcut. To do 27 × 11, add the digits and insert in the middle: 2 + 7 = 9, answer 297. Now try 89 × 11. The digit sum is 17, and if you only memorized "insert in middle," you write 8179 or freeze. Try 271 × 11 and the rule has nothing to say at all.

The reason 27 × 11 worked is that 11 = 10 + 1, so you're really computing 270 + 27. That decomposition handles 89 × 11 without flinching, and it scales to three digits. The structure is the part that generalizes; the trick is a special case dressed up as magic.

This is why we organize strategy training into families instead of one-off shortcuts. compensation-add trains the same structural move — round up, then correct down — across dozens of different number shapes. After enough exposure, you stop seeing problems as separate; you see the family.

Speed is a proxy, not the skill

The mental math you actually need in daily life is not about speed. It's about whether the number makes sense.

Your restaurant bill comes to $847 for four people. You don't need to divide precisely — you need to notice that $800 ÷ 4 ≈ $200, and decide whether $200 a head is fine for this place. Two seconds, no tricks. Or a doctor adjusts a dose: the question is not "how fast can I multiply?" but "is the answer 200, or did I drop a decimal and get 20?"

Speed becomes useful when calculations stack. For most people, in most situations, what looks like speed is really fluency — the state where numbers feel light rather than laborious. We train for that fluency directly, with decks like number-bonds-10 and x-tables-2-9, so the sub-steps stop occupying attention. Speed is what happens after fluency, not before it.

Talent is mostly compressed familiarity

Watch someone work out 347 × 29 in a few seconds and the instinctive response is: they're wired differently. Almost always, they aren't. They've internalized that 29 = 30 − 1, computed 347 × 30 = 10,410, subtracted 347, and arrived at 10,063 by collapsing the problem into a shape they've seen a hundred times.

The myth of innate talent is actively bad for adults trying to improve. They watch someone fluent and conclude: not for me. But the effortlessness is the product of the practice, not a precondition for it. Our near-round-mul family is built around exactly this move — see 29, see 98, see 102, and reach for the anchor automatically.

"Train your brain" is too vague

A whole genre of advice says mental math is really about general mental fitness — sharpen the mind and the arithmetic will follow. Brain-training apps lean into this.

It doesn't work. The representations involved in 18 × 15 are not the ones involved in memorizing digit sequences or rotating shapes. If the goal is fluency with numbers, you have to work with numbers. We make this concrete: every deck targets a specific skill, and the catalog makes the targets legible.

What we're actually training

We treat mental math as a stack of four things, in this order:

Number sense — an intuitive feel for magnitude and proportion. It tells you 30% of 90 is around 27 before you calculate.

Pattern recognition — seeing that 48 × 25 is really 12 × 100, or that 997 × 6 is (1,000 − 3) × 6. The strategy decks (compensation, counting-up, halve-and-double) train this directly.

Fact fluency — the foundation that makes everything else cheap. If you have to reconstruct 7 × 8, every problem that touches it gets slow. Decks like squares-1-32 and fraction-decimal-anchors live here.

Sanity-check reflexes — the output that matters most in daily life. A 5% raise on $60,000 is $3,000, not $6,000. You should feel the wrongness before precision. The sanity-check deck trains exactly this signal.

Why the framing matters now

The standard objection to teaching mental arithmetic is "we have calculators." This was always weak; it is now backwards.

Calculators are cheap. AI-generated answers are cheaper. What is not cheap is noticing when an output is nonsense. A language model can confidently tell you that a 15% tip on a $43 bill is $9.45 — it isn't, it's $6.45. It can produce a financial projection where costs grow faster than revenue and the bottom line still improves. A person with decent number sense catches it instantly. Without it, you proceed.

A calculator gives answers. Mental math gives resistance to nonsense. That's the skill we're building, and it's why we don't lead with tricks.

This is a product-context distillation. The long-form essay lives on mingjerlee.com →.