Rounding method, i. e. COGO-Labeling

[ChristophLampart]

Dear Civilists,

currently there is only one standard method available, Banker's rounding. Noticed during labelling COGO-Points.
This means:

“Banker’s rounding is a midpoint rounding scheme (round-to-even). 
When a value is exactly between two candidates, it is rounded to the one with an even least-significant digit.
Example: 73.5 → 74, 74.5 → 74, 75.5 → 76"

This method is completely different to what is taught in engineering schools. Therefore this standard rounding isn't suitable to engineering tasks, for what Civil 3D is made for, I guess.

It is urgently needed to add a "engineering rounding" or "commercial rounding" or "rounding to the nearest". Means if the last deleted digit is 5 or more, than the last remaining digit will be rounded up. If the last deleted one is 4 or less, than the last remaining digit will stay as it is.

Example: 73.5 -> 74, 74.5 -> 75, 75.5 -> 76, 73.4 -> 73

Please support this idea, so that it will come true.

All the best.

[AllenJessup]

Not saying it would make much of a difference. But when I was learning Surveying. I was taught the round to even method. The theory being that when dealing with a long string of numbers. That type of rounding would balance out. I've never heard that Engineering was different. But I never went to Engineering school. Can you explain why this type of rounding is unsuitable for Engineering calculations? I think it's the standard for Surveying calculations which Civil 3D also does. Possibly a split between the two methods depending on what type of calculation is being done.

[ChristophLampart]

Hi @AllenJessup,

thanks for your input.

Really never had to use the "even method", neither in school nor on university. But for me it is fine if the "commercial rounding" is only added. It should not replace the even method.

All the best

[ChrisRS]

This is a real edge case. In floating‑point numerical calculations, a displayed value like 73.5 almost never represents an exact midpoint. Internally, it’s stored as the nearest binary fraction — something slightly below or slightly above the decimal value.

For example: 

Internal Value	Display	Round‑Even	Round‑Up
73.44446	73.5	73	74
73.54444	73.5	74	74

The main rule I remember from engineering was simply not to round intermediate results. Carry full precision through the math, then round the final answer. Because calculators made exact .5 values exceedingly rare, the specific rounding method was never a major topic. Given that, I tend to favor Round‑Even.

This whole issue is very low priority compared to the need for easy, reliable formatting of left-truncated elevations without complex expressions. Example: turning 2405.68 into 05.68 should be straightforward.

@AllenJessup would also appreciate the classic surveying case: splitting a 200.00‑foot lot line into thirds should produce 66.66, 66.66, and 66.67, not three identical values. That’s a formatting and communication issue, not a floating‑point one.

[AllenJessup]

@ChrisRS  The classic surveying case: splitting a 200‑foot lot line into thirds should produce 66.66, 66.66, and 66.67, not three identical values. One Chain and One Link.