Rudder

• Sort values from small to big
• Have a counter for the rounding error (= displayed value - real value)
• All values <1% are displayed <1% and counted as 1%
• For each other value, round depending on the counter (above i negative, below if positive)

> this should be as close to 100% as possible, more or less 1% (except if there are too many <1%
> some value are not rounded to the closest

• Idem except:
• Round to the closest all values interdependently of the counter except the last one
• Substract the counter from the last value before rounding

> This can make sure we hit 100%
> only the biggest value is imprecise

my algorithm is still valid

I tested really small status (1 error for 1k reports to 1billion)

1k to 20k => Ok! (a small portion appear containing the error)

20k to 21M => 100% green (no error section)

21M and above => Overflow !

Above 20k reports, this breaks because we have less that 0,005 % and since we round to 2 digits, it's rounded to 0,00%

This is causing grave problem in the case where one among a big number of reports are in error, and that component is critical.