Skip to content

← Math Further 6–8

Is the Answer Roughly Right?

Topic 38 · Grades 6–8 · Estimation and casting out nines

What this builds on from class
  • Grade 6: Decimal (rounding off)
  • Grade 6: Whole Number (divisibility tests)
In this unit 7 sections
  1. Three quick checks
  2. Why the nines never lie
  3. What the nines can’t see
  4. A second net: casting out elevens
  5. Where this is used
  6. Questions
  7. Where to go next

It is the week before school starts, and Sunita didi’s stationery shop near Bhaktapur’s Durbar Square has been busy all day. Her cousin Bikash worked out the bills on a calculator and wrote each answer on the bill pad. Now didi is checking them before she closes, and she has no time to do every multiplication again. “A calculator is never wrong,” says Bikash. “Calculators aren’t,” says didi. “Fingers are.”

She runs her eye down the pad and taps three of the bills almost at once. “This one, this one and this one.” She has not multiplied anything. How did she know?

Try it before you read on. Mark each bill ✓ (right) or ✗ (wrong) as fast as you can, without a calculator and without working any of them out in full, then press Check.

Try it

Sunita didi’s bill pad

Mark each bill ✓ or ✗, then press Check. The page times you from your first tap.

  1. 24 notebooks × Rs 65 24 × 65: right. = Rs 1,560
  2. 38 pens × Rs 15 Wrong: 38 × 15 is Rs 570, the 7 and the 0 swapped. = Rs 507
  3. 12 geometry boxes × Rs 185 Wrong: 12 × 185 is Rs 2,220, a 0 too many. = Rs 22,200
  4. 4.5 m of cloth × Rs 240 4.5 × 240: right. = Rs 1,080
  5. 64 rulers × Rs 29 Wrong: 64 × 29 is Rs 1,856, a 4 for a 5. = Rs 1,846
  6. 46 glue sticks × Rs 32 46 × 32: right. = Rs 1,472

Mark all six bills ✓ (right) or ✗ (wrong), then press Check.

All six marked. Press Check.

All six right, in 0 seconds! Three bills were wrong. How did you spot them?

0 of 6 right, in 0 seconds. Three bills were wrong; the notes say how.

Guess first. Didi caught three wrong bills in a few seconds. What do you think she looked at?

Pick an answer to see why.

Three quick checks

The last digit

38 pens at Rs 15 each. You don’t need the whole product to know how it ends: only the last digits matter. 8 × 5 = 40, which ends in 0, so 38 × 15 must end in 0 as well. The bill says Rs 507, which ends in 7. It is wrong, whatever the rest of it says. (The right answer is Rs 570. Bikash swapped the 7 and the 0.)

Why do only the last digits matter? Write 38 as 30 + 8 and 15 as 10 + 5. Multiplying out gives 30 × 10, 30 × 5, 8 × 10 and 8 × 5. The first three each have a 10 in them, so each ends in 0. Only 8 × 5 can change the last digit.

The last digit check is very fast, but it only looks at one place. A slip anywhere else gets past it.

The rough size

12 geometry boxes at Rs 185 each. Round each number to its first digit (to one significant figure), as Grade 6 rounds off decimals to a place: 12 is about 10 and 185 is about 200. So the bill should be about 10 × 200 = Rs 2,000. The pad says Rs 22,200, which is about ten times too much. Somewhere a 0 was pressed once too often. (The right answer is Rs 2,220.)

This is called estimating: working out a rough answer from rounded numbers. A rough answer can’t tell you the exact bill, but it tells you its size: how many digits it should have, and roughly what it starts with. That is exactly what you need to catch a slip of a whole place, the most expensive kind.

How rough is rough? You can do better than “about 2,000”. Round both numbers down and you get an answer that is surely too small; round both up and you get one that is surely too big. For 12 × 185, the true answer is somewhere between 10 × 100 = 1,000 and 20 × 200 = 4,000. These two numbers are bounds: the answer can’t be outside them, 1,000 ≤ 12 × 185 ≤ 4,000. The bill’s Rs 22,200 is far outside, so it is wrong for certain, not just “probably”.

Without multiplying, between which bounds must 64 × 29 lie?

Pick an answer to see why.

Decimals and the point

Estimating is at its best with decimals, because a decimal point in the wrong place is a slip of a whole place. Didi’s cloth is 4.5 meters at Rs 240 a meter. About 5 × 200 = 1,000, so Rs 1,080 is believable. If Bikash had missed the point and typed 45 × 240, the calculator would have said Rs 10,800, and the estimate would have caught it at once.

The same works with fractions and percents. Three quarters of Rs 1,196 is about three quarters of 1,200, which is 900 (exactly Rs 897), and 48% of Rs 205 is about a half of 200, so about Rs 100 (exactly Rs 98.40).

The rough size can’t see small slips. The rulers’ bill, 64 rulers at Rs 29, says Rs 1,846. The estimate, 60 × 30 = 1,800, looks fine. Its last digit, 4 × 9 = 36, ends in 6, and so does 1,846. Both checks pass. And yet didi tapped it. For this one she needed the third check.

Casting out nines

Add up the digits of each number until only one digit is left:

  • 64: 6 + 4 = 10, then 1 + 0 = 1.
  • 29: 2 + 9 = 11, then 1 + 1 = 2.
  • Multiply those: 1 × 2 = 2. So the answer’s digits must also come down to 2.
  • The bill, 1,846: 1 + 8 + 4 + 6 = 19, then 1 + 9 = 10, then 1.

The answer should come down to 2, and the bill comes down to 1. It is wrong. (The right answer is 1,856: 1 + 8 + 5 + 6 = 20, which comes down to 2. Bikash typed a 4 for a 5.)

This is casting out nines. To “cast out” is to throw away, and what it throws away is nines: a 9 in a number can simply be crossed out, and so can any digits that add up to 9, because they make no difference to the result. In 1,856, the 1 and the 8 make 9, so cross them out, and 5 + 6 = 11 comes down to 2. When the digits come down to 9, count it as 0.

Here is the casting-out-nines check for anything you like. Type a sum, a difference or a product and the answer you got (or one you made up), and it lays the check out as a cross, the way old arithmetic books did: the two numbers’ nines on the left and right, what their answer’s nines must be at the top, and the answer’s own nines at the bottom.

Try it

The nines cross

Type two numbers, choose +, − or ×, and type an answer. The top of the cross says what the answer’s nines should be; the bottom says what they are.

=
1221
Left
64 → 10 → 1
Right
29 → 11 → 2
Top (should be)
1 × 2 = 2
Bottom (the answer)
1,846 → 19 → 10 → 1

Type both numbers and an answer.

Put the bigger number first for a difference.

Top and bottom disagree, so the answer is surely wrong.

Top and bottom agree: no slip the nines can see. The answer may be right, or the slip may be one the nines miss.

Why the nines never lie

Topic 4, Why Do the Divisibility Tests Work?, showed one fact: a number and its digit sum leave the same remainder when divided by 9. (Each 10, 100 or 1,000 is one more than a number made of nines, so each digit leaves itself over.) Adding the digits again and again until one digit is left just keeps doing that. So the single digit you end with, counting 9 as 0, is the remainder on dividing by 9: 64 leaves 1 (64 = 9 × 7 + 1), and 29 leaves 2 (29 = 9 × 3 + 2).

So casting out nines is really a check on remainders. It works because remainders behave well under adding and multiplying.

For a sum

Take 1,250 + 875 + 2,460. Each number is some nines plus its remainder: 8, 2 and 3. Adding the numbers adds the nines (still nines, with nothing over) and adds the remainders. So the total must leave the same remainder as 8 + 2 + 3 = 13, which is 4. And the total, 4,585, comes down to 4 + 5 + 8 + 5 = 22, then 4. They agree.

For a product: a picture

A product is an area. Draw 64 × 29 as a rectangle 64 wide and 29 tall, and cut it where the nines end: 64 = 63 + 1, and 29 = 27 + 2, with 63 = 9 × 7 and 27 = 9 × 3.

63 = 9 × 7127 = 9 × 3263 × 27 = 1,701a multiple of 92763 × 2 = 12621,701 + 126 + 27 + 2 = 1,856left over after the nines: 1 × 2 = 2
Figure 38.1 64 × 29 cut into four pieces (not to scale: the thin strips are drawn wider). Every gray piece has a side made of nines, so it shares out evenly among nine. Only the red corner, 1 × 2, has no nines in it: it is the remainder of the whole product.

The big piece, 63 × 27, is a multiple of 9. So is the strip 63 × 2 (63 is), and so is the strip 1 × 27 (27 is). Only the corner, 1 × 2 = 2, is made of the two remainders. So the whole product leaves the same remainder as 1 × 2, which is 2. That is why the rule is “multiply the remainders”: the rest of the rectangle is all nines. It works for any two numbers, because any number can be cut into nines and a remainder.

With letters: if one number is 9a + r and the other is 9b + s, the four pieces are 81ab, 9as, 9br and rs. The first three have a 9 in them. So the product leaves the same remainder as rs.

A calculation passes the casting-out-nines check. What do you know?

Pick an answer to see why.

Subtracting and dividing

Subtraction works the same way, with one care: if the second remainder is bigger, add 9 first. For 5,000 − 1,875 = 3,125: 5,000 comes down to 5 and 1,875 to 21, then 3; 5 − 3 = 2, and 3,125 comes down to 11, then 2.

For division, check it backwards: dividend = divisor × quotient + remainder. Is 7,345 ÷ 23 = 319, remainder 8, right? 23 comes down to 5, and 319 to 13, then 4; 5 × 4 = 20, which comes down to 2; add the remainder 8 to get 10, which comes down to 1. And 7,345 comes down to 19, then 10, then 1. They agree.

What the nines can’t see

The check looks only at the digit sum, so any slip that keeps the digit sum the same gets past.

  • Two digits swapped. 1,856 and 1,586 have the same digits, so the same digit sum. A swap is invisible to the nines, wherever the two digits are. Swapping is one of the commonest typing slips.
  • A 0 and a 9 mixed up. 9 is counted as 0, so 1,056 and 1,956 look the same to the nines.
  • A 0 too many or too few. 22,200 and 2,220 have the same digit sum. Didi’s geometry-box bill passes the nines check.
  • A decimal point in the wrong place. Moving the point doesn’t change the digits at all, so 1,080 and 10,800 and 108.0 all look the same.

And a random wrong answer, such as a whole number misread, still leaves one of the nine possible remainders. About 1 time in 9 it is the right one by chance, so about 1 in 9 random mistakes, or 11%, get through.

Estimating can’t see a slip in the last digits, and the last digit check sees only the last place. So the three checks cover for one another:

Slip Last digit Rough size Nines
One digit wrong, in the last place catches misses catches (unless it is 0 for 9)
One digit wrong, in the middle misses misses catches (unless it is 0 for 9)
One digit wrong, the first digit misses often catches catches (unless it is 0 for 9)
Two digits swapped only if the last digit moves misses, unless the first digit moves misses
A 0 or a decimal point in the wrong place sometimes catches misses

Estimating catches the costly slips of a whole place; the nines catch a wrong digit in the middle. The slip that gets past all three is the quiet swap in the middle: Rs 1,586 for 1,856, or Rs 750 for 570.

A swap leaves its fingerprint

The nines can’t catch a swap, but they leave a clue. Swapping two digits never changes the remainder, so the difference between the right number and the swapped one is always a multiple of 9: 570 − 507 = 63 = 9 × 7, and 1,856 − 1,586 = 270 = 9 × 30. If a shop’s cash and its book disagree by an amount that divides by 9, a swap is a good first guess. (Swapping the ones and tens digits changes the number by 9 times the gap between the two digits: 9, 18, … up to 81. Swapping the tens and hundreds changes it by 90 times the gap: 90, 180, … up to 810.)

A second net: casting out elevens

Nine is not the only number you can cast out: any number works, if you can find its remainders quickly. The best partner for nines is eleven.

Topic 4 showed the test for 11: take the digits from the right, and add, subtract, add, subtract. For 1,856: 6 − 5 + 8 − 1 = 8. So 1,856 leaves 8 when divided by 11. (If you get a negative number, add 11; if you get 11 or more, take 11 away.) It works because 10 is one less than 11, 100 is one more than 99, 1,000 is one less than 1,001, and so on: each place counts as +1 or −1.

The check is the same as with nines. 64 leaves 4 − 6 = −2, which is 9; 29 leaves 9 − 2 = 7; 9 × 7 = 63 leaves 3 − 6 = −3, which is 8. And 1,856 leaves 8. They agree.

Now the swap. 1,586 gives 6 − 8 + 5 − 1 = 2, not 8. The elevens catch it. Swapping two neighboring digits moves one onto a plus place and the other onto a minus place, so the alternating sum changes, by twice the difference between the two digits. That is never a multiple of 11 unless the two digits are the same, and then swapping them changes nothing. So casting out elevens catches every swap of two neighbors and every single wrong digit, 0 for 9 included.

Elevens have their own blind spot: swapping two digits two places apart, such as the 1 and the 5 in 1,856, keeps them both on plus places or both on minus places. And no check that works on remainders can catch everything. Try to fool all four checks below.

Try it

Make a slip

Tap a digit of the answer to change it, or switch to Swap two digits and tap two digits to swap them. Each check says whether it is fooled. Can you complete the three challenges?

Right answer: 64 × 29 = 1,856

Your slip:
CheckShould beYour slip
Last digit66–
Rough size1,200 to2,1001,856–
Nines22–
Elevens88–
  • Fool the nines
  • Fool the last digit, the size and the nines at once
  • Fool all four (try 347 × 268)

Tap a digit to change it (each tap adds 1).

That would put a 0 first. Swap other digits.

1,856 is wrong, and 0 of the 4 checks catch it.

1,856 is wrong, and it fools all four checks!

Hint
Answer

Where this is used

Questions

A calculator says 19 × 21 = 3,999. What does a rough estimate say?

Pick an answer to see why.

a.★

Estimate each answer by rounding: 6.8 × 41, 0.52 × 395, and a quarter of 1,612. Then say whether 27.88 is a believable answer to the first.

Answer

Which slip can casting out nines never catch?

Pick an answer to see why.

b.★★

Check by nines that 8,527 ÷ 34 = 250, remainder 27. Then show that the check catches the slip “260, remainder 27”.

Answer
c.★★

A shop’s book says the day’s sales were Rs 5,836, but the cash in the drawer, counted twice, is Rs 5,386. Could one swap explain it? Which digits?

Hint
Answer
d.★★★

Explain why a number with its digits in any other order differs from the original by a multiple of 9. Use it to explain why the bill Rs 507 for 38 pens at Rs 15 passes the nines check.

Answer
e.★★★

Casting out threes uses the remainder on dividing by 3 (the digit sum works for 3 too). Why is it a weaker check than casting out nines?

Answer
f.★★★

Explain why every wrong answer to 64 × 29 is caught by at least one of the four checks (last digit, rough size between 1,200 and 2,100, nines, elevens).

Hint
Answer

Where to go next

Where these facts come from
  • Casting out nines as a check by digital roots for addition, subtraction, multiplication and division, crossing out 9s and digits that make 9, its blind spot for transpositions and for wrong results with the same remainder, and casting out elevens by alternately adding and subtracting digits: Wikipedia, “Casting out nines”, read October 2026.
  • Aryabhata II’s rules in the Mahasiddhanta (about 950) for multiplication, division, square and cube roots; Arab writers’ check for multiplication and division; Ibn Sina (about 1020) and the “Hindu method”; Maximus Planudes; Narayana’s rule (1356) for testing by any chosen number, the first Indian mathematician to give it; the check first used only for multiplying and dividing and later completed separately: Bibhutibhushan Datta and Avadhesh Narayan Singh, History of Hindu Mathematics (1935; Asia Publishing House, 1962), Part I, “Checks on operations”, pp. 180–184, through the Internet Archive’s copy, read October 2026.
  • Fibonacci’s Liber Abaci describing casting out nines; Hippolytus (170–235) and Iamblichus reducing numbers to a one-digit root without using it as a check; the Mahasiddhanta as the earliest known surviving work to use it as a check: Wikipedia, “Casting out nines” (citing David Wells, Heath and Datta and Singh), read October 2026.
  • Narayana Pandita’s dates (about 1340–1400), the Ganita Kaumudi, and Plofker’s judgment that his were the most significant Sanskrit mathematics treatises after Bhaskara II’s, the Kerala school apart: Wikipedia, “Narayana Pandita (mathematician)”, read October 2026.
  • The J-Com order (December 2005, 610,000 shares at ¥1 instead of one share at ¥610,000; a loss of ¥40.7 billion): Kyodo News, “Mizuho Securities to sue for J-Com loss”, The Japan Times, 18 October 2006 (the Internet Archive’s copy), read October 2026.
  • The J-Com order’s date (December 2005; Wikipedia’s “Fat-finger error” says 2006, which is wrong) and the court case over the order that couldn’t be canceled: Wikipedia, “Mizuho Securities”, read October 2026.
  • The Dentsu order (2001, 610,000 shares at ¥6 instead of 6 shares at ¥610,000, about £71 million): Helen Dunne, “Trader’s slip leaves UBS Warburg £71m poorer”, The Telegraph, 1 December 2001 (the Internet Archive’s copy), read October 2026.
  • “Fat-finger error” and automated systems that catch such orders before they reach the market: Wikipedia, “Fat-finger error”, read October 2026.
  • Leading zeros required and trailing zeros forbidden in written doses, to prevent misdoses by a factor of ten or more: Wikipedia, “Leading zero” (citing Neil M. Davis, Medical Abbreviations, 2005), read October 2026.
  • Grade 6 rounding off (decimal numbers rounded to a place): CDC’s Mathematics, Grade 6, section 5.3, “Rounding off of decimal numbers”, read October 2026.
  • What the international curricula ask: Common Core 7.EE.B.3, “assess the reasonableness of answers using mental computation and estimation strategies”; England’s national curriculum for key stage 3, “use approximation through rounding to estimate answers and calculate possible resulting errors”, read October 2026.
  • Every number in this topic computed in checks/textbooks/math-further-6-8/38/tmp/numbers.py.