Explained clearly 6 min read

Why Did My Teacher Mark 2.5 as 2? The Half-Even vs Half-Up Confusion

A precise explanation of why 2.5 can become 2 under round-half-even conventions, how that differs from round-half-up, and what ASTM E29, ISO 80000-1, and NIST recommend.

Short Answer

A precise explanation of why 2.5 can become 2 under round-half-even conventions, how that differs from round-half-up, and what ASTM E29, ISO 80000-1, and NIST recommend.

Short answer: If your teacher rounded 2.5 to 2, they were almost certainly using round half to even (banker’s rounding), not the schoolroom round half up rule. Under round-half-even, a tie at 5 goes to the nearest even digit; 2 is even, so 2.5 → 2. Under round-half-up, 2.5 → 3. The disagreement is not an arithmetic error; it is a convention conflict. This site is a precision-and-rounding reference, not just a calculator. The significant figures calculator implements multiple rounding conventions, but the articles explain when and why each convention applies.

Rule Statement

Rounding to a specified number of significant figures or decimal places requires deciding what to do when the first discarded digit is exactly 5 and no nonzero digits follow. Two common tie-breaking rules are:

  • Round half up (half away from zero): increase the last retained digit by 1. Example: 2.5 → 3; -2.5 → -3 (away from zero).
  • Round half to even (banker’s rounding): if the retained digit is even, leave it; if odd, increase by 1. Example: 2.5 → 2; 3.5 → 4; -2.5 → -2.

The tie rule applies only when the discarded portion is exactly one half. If the discarded portion is greater than half, round up; if less, round down. For example, 2.5001 → 3 under both rules; 2.4999 → 2 under both rules. When rounding 2.5 to one significant figure, the retained digit is 2 and the discarded digit is 5 with no further nonzero digits. Half-up gives 3; half-even gives 2 because 2 is even.

Convention Comparison Table

Convention 2.5 to integer 3.5 to integer -2.5 to integer Typical context
Round half up (away from zero) 3 4 -3 School arithmetic, many calculators, Excel ROUND
Round half up (toward +∞) 3 4 -2 JavaScript Math.round
Round half to even 2 4 -2 ISO 80000-1, ASTM E29, NIST SP 811, Python, R
Round half down (toward zero) 2 3 -2 Some financial truncation contexts
Round half to odd 3 3 -3 Rare, used in some specialized algorithms

Notice that half-even is not simply ’round down on 5.’ It rounds 3.5 up to 4 because 3 is odd. The rule removes bias by making ties alternate between up and down.

Worked Examples

Example 1: 2.5 to one significant figure

  1. Identify the retained digit: 2.
  2. Identify the discarded portion: 5, with no nonzero digits after it.
  3. This is an exact tie.
  4. Half-up: 2 + 1 = 3.
  5. Half-even: retained digit 2 is even, so keep 2.

Result: 2.5 → 2 under half-even; 2.5 → 3 under half-up. Your teacher’s mark of 2 is consistent with half-even.

Example 2: 3.5 to one significant figure

  1. Retained digit: 3.
  2. Discarded portion: 5, exact tie.
  3. Half-up: 3 + 1 = 4.
  4. Half-even: 3 is odd, so increase to 4.

Result: 3.5 → 4 under both rules. This counteracts the idea that half-even always rounds 5 down.

Example 3: 2.451 to one decimal place

  1. Retained digit: 4 in the tenths place.
  2. Discarded portion: 51, which is greater than 50.
  3. Round up: 2.451 → 2.5.

Result: 2.5 under both half-up and half-even. The tie rule does not apply because the discarded portion is not exactly half.

Example 4: 2.45 to one decimal place

  1. Retained digit: 4.
  2. Discarded portion: 5, exact tie.
  3. Half-up: 4 + 1 = 5, giving 2.5.
  4. Half-even: 4 is even, so keep 4, giving 2.4.

Result: 2.45 → 2.5 under half-up; 2.45 → 2.4 under half-even.

Counter-Examples

  • Always rounding 5 up: This is the most common schoolroom shortcut, but ISO 80000-1 Clause 6.4 and ASTM E29 do not use it for exact ties. Treating 2.5 as 3 in a specification test can bias results upward.
  • Applying half-even to non-ties: 2.5001 is not a tie. It rounds to 3 under both half-up and half-even because the discarded portion exceeds one half.
  • Double rounding: Rounding 2.4445 to three decimals gives 2.445 (half-up) or 2.444 (half-even), and then rounding again can produce a different answer than rounding directly to two decimals. See Double Rounding Error.
  • Confusing significant figures with decimal places: 2.5 to one significant figure and 2.5 to the nearest integer both involve the same tie, but the significant-figure context changes how you count retained digits. See Rounding vs Significant Figures.
  • Forgetting negative numbers: Half-even rounds -2.5 to -2, not -3. Half-away-from-zero rounds -2.5 to -3. The sign changes the direction of the tie-break.

Standards Citation

ISO 80000-1:2009, Clause 6.4 (Rounding of numbers) states that when the digit following the last digit to be retained is exactly 5, the last digit retained is increased by one if it is odd and left unchanged if it is even. This is the round-half-even convention. ISO 80000-1 is the general metrology standard for quantities and units.

ASTM E29, Standard Practice for Using Significant Digits in Test Data to Determine Conformance with Specifications, specifies a rounding procedure that also uses round-half-even for exact ties. The exact clause varies by edition; in ASTM E29-13, the rounding rules appear in Section 7 and Annex A1. It is widely used in materials testing and specification conformance.

NIST SP 811, Guide for the Use of the International System of Units (SI), Section 9.8 and Appendix B, recommends rounding by the same half-even rule when a tie occurs. NIST also emphasizes retaining enough digits to avoid false precision.

JCGM 100:2008 (GUM), Clause 7.2.6, advises that uncertainty be reported with no more than two significant digits. When rounding those values, ISO 80000-1 Clause 6.4 provides the tie-breaking rule. The GUM does not create a separate tie rule; it relies on the ISO convention.

Common Mistakes

  • Assuming the teacher is wrong. The teacher may be following ASTM E29, ISO 80000-1, or a course-specific convention. Ask which rounding rule the course uses.
  • Mixing rules within one calculation. If you round intermediate results with half-up and final results with half-even, you can create inconsistencies. Keep one convention throughout unless a standard requires otherwise.
  • Rounding before the final step. Carry extra digits through calculations and round only the final reported value. This avoids double rounding and false precision.
  • Ignoring exactness. A tie requires the discarded portion to be exactly one half. In measurement, 2.5000000001 is not a tie.
  • Assuming all software agrees. Python, Excel, JavaScript, R, MATLAB, and SQL databases use different tie-breaking rules. See the Software Behavior Note below.

Software Behavior Note

Tool 2.5 rounded to integer Tie rule
Python round(2.5) 2 Round half to even (binary floating-point caveats)
Excel ROUND(2.5,0) 3 Round half away from zero
Google Sheets ROUND(2.5,0) 3 Round half away from zero
JavaScript Math.round(2.5) 3 Round half toward +∞ (Math.round(-2.5) is -2)
R round(2.5) 2 Round half to even (IEC 60559)
MATLAB round(2.5) 3 Round half away from zero by default
SQL databases Varies Check the specific RDBMS; SQL Server and PostgreSQL numeric types often round half away from zero, while binary floating-point may use half-even.

Calculators such as the TI-84 and Casio fx-991 typically use half-up in their basic rounding displays, but statistical and engineering modes may differ. Always consult the manual for high-stakes work.

Discipline Note

Different fields adopt different defaults. In school mathematics, half-up is common because it is easy to teach. In science, engineering, and metrology, half-even is preferred because it reduces cumulative bias in large data sets and is recommended by ISO 80000-1, ASTM E29, and NIST SP 811. In finance and accounting, half-up or half-away-from-zero is often used for monetary amounts, while some banking systems use banker’s rounding to avoid systematic bias. In measurement uncertainty, follow the GUM and ISO 80000-1: report uncertainty to two significant digits and use half-even for exact ties. The key is to state the convention explicitly whenever a tie can affect conformance or reported precision.

FAQ

Why did my teacher mark 2.5 as 2?

Your teacher likely used round-half-even (banker's rounding). Because the retained digit 2 is even, an exact tie at 5 stays at 2. Under round-half-up, it would become 3.

Is 2.5 rounded to one significant figure 2 or 3?

Both answers are defensible under different conventions. ISO 80000-1 and ASTM E29 give 2; the common school half-up rule gives 3. The correct answer depends on the required convention.

Does ASTM E29 use banker's rounding?

Yes. ASTM E29 specifies round-half-even for exact ties in its significant-digit rounding procedure. Check the edition for the exact clause.

What does Python round(2.5) return?

Python returns 2 because it uses round-half-even for exact binary ties. However, many decimal fractions are not exact in binary, so Python's behavior can surprise you for values like 2.675.

Is round-half-even always rounding down on 5?

No. 3.5 rounds to 4 because 3 is odd. Half-even rounds to the nearest even digit, so it alternates between rounding up and rounding down.

Verified sources

References

  1. ISO 80000-1:2009, Quantities and units — Part 1: General, Clause 6.4 (Rounding of numbers).
  2. ASTM E29-13, Standard Practice for Using Significant Digits in Test Data to Determine Conformance with Specifications.
  3. NIST SP 811, Guide for the Use of the International System of Units (SI), 2008 Edition, Section 9.8 and Appendix B.
  4. JCGM 100:2008, Evaluation of measurement data — Guide to the expression of uncertainty in measurement (GUM), Clause 7.2.6.
  5. IEEE 754-2019, IEEE Standard for Floating-Point Arithmetic (for software tie-breaking behavior).

Leave a Reply

Your email address will not be published. Required fields are marked *