To round a number in Python you can use the round() function. It’s straightforward, but there is a nuance to be aware of. When you pass a number that is exactly halfway between two integers to this function, its behavior might be surprising:

round(1.5)
# 2

round(2.5)
# 2

This happens because round() employs an algorithm known as bankers’ rounding (BR), instead of arithmetic rounding (AR). The key distinction between these algorithms is that for numbers of the form x + 0.5, where x is even, BR rounds the number down, not up, as it would be with AR.

The reasony why to use bankers’ rounding

When you want to round a single number, using BR algorithm might seem to have no sense, because it appears indeterministic. However, BR becomes advantageous for rounding a large set of numbers due to its statistical benefits. Consider the following set of measurements:

1, 3, 2.5, 5.5, 6.5, 4, 1.5, 8

Summing them up and calculating the average yields 32/8 = 4.

If we round each value using (AR) before calculating the average, the result changes to 34/8 = 4.25.

There is no objective reason why a number ending in .5 should always be rounded up. Statistically, such numbers should be rounded up and down with 50% probability. BR algorithm ensures this.

Available workarounds

One of the workarounds to achieve AR in Python is using int(x+0.5) instead of round(x). This method is straightforward, but the formula becomes more complex when you want to round the number to n digits. For instance, to imitate round(x, n), the formula is:

int(x + 10**n + 0.5) / 10**n

---

Another workaround requires using decimal library:

from decimal import Decimal

x = 1.556
precision = Decimal('0.00')
float(Decimal(x).quantize(precision))
# 1.56