A conversation amongst friends turned towards oxymorons and everyone started providing examples. A friend came up with the expression “reasonably accurate” and I had to disagree that it was indeed an example of an oxymoron. “Doesn’t accurate mean correct? How can something be reasonably correct?” was his contention.
It reminded me of a story about the height of Mount Everest. When the mountain was first measured, the calculation supposedly came to exactly 29,000 feet. But would anyone believe that the surveyors had measured it carefully if they announced such a suspiciously round number? So, the story goes, they added two feet and declared the height to be 29,002 feet.
There is another interesting detail. The story itself may not be true. A technical account from the Survey of India gives the measurements obtained from six different locations and explains that their average, rounded to the nearest foot, was 29,002 feet. No additional two feet were required.
Still, we readily believe the first story. Apparently, in that story, even a carefully measured height of 29,000 feet needed two extra feet to sound believable.
But the degree of accuracy changes based on context. Let us start with a simple question.
How old are you?
My answer would be that I am fifty-four. Is that accurate? For that conversation, it is accurate enough. In India, I might also say that I am in my fifty-fifth year. I haven’t aged a year while answering. I have merely changed the way I count.
For a baby, we usually give the age in months, or even days. For someone we don’t know, “He must be forty or forty-five” may be sufficient. Some forms want our age in years and months. Others require the date of birth. An astrologer preparing a horoscope may ask for the place and time of birth down to the exact second. The more exact the information, the better the prediction, they would claim.
But imagine meeting someone and announcing that I am fifty-four years, seven months, twenty days, eleven hours, twenty-three minutes, and ten seconds old. Sorry, thirteen seconds by the time I finished saying it.
The person would probably give me a peculiar look and move away. Completed years would have served that conversation quite well. So accuracy is relative to context.
My favorite example, though, is pi.
You may remember learning at school that pi is the ratio of a circle’s circumference to its diameter. You may also remember 22/7 or 3.14. Both are approximations. Pi begins 3.14159265… and continues forever, without settling into a repeating pattern.
For many everyday calculations, 3.14 is sufficient. When necessary, we can use more digits after the decimal point. NASA’s Jet Propulsion Laboratory uses fifteen decimal places for its highest-accuracy interplanetary navigation calculations: 3.141592653589793.
Let us use a real-life example.
The formula is C = 2πr. Since twice the radius is one meter, the answer is conveniently pi meters.
Use 3.14, and we get 3.14 meters. Use 3.14159, and we get 3.14159 meters. The difference is 1.59 millimeters.
Now use all fifteen decimal places from the JPL calculation. Our answer becomes 3.141592653589793 meters. Compared with the answer using 3.14, the difference is now 1.592653589793 millimeters.
We have made the difference considerably longer to read. It is still less than 1.6 millimeters.
Let us try a slightly larger circle.
Voyager 1, which has traveled beyond the heliosphere, is roughly 26 billion kilometers from Earth. Imagine a circle with that distance as its radius. If we calculate its circumference using pi to fifteen decimal places, the error caused by leaving out the remaining digits is about 1.2 centimeters.
A circle reaching all the way to Voyager 1, and we are debating a centimeter.
Still not big enough?
The observable universe has an estimated radius of 46 billion light-years. According to NASA/JPL, thirty-seven decimal places of pi would be sufficient to calculate the circumference of an imaginary circle of that radius with an error smaller than the diameter of a hydrogen atom.
There is a condition attached to these calculations. We are treating the radius as exact and considering only the error introduced by our approximation of pi. Is the radius of our table exactly half a meter? Voyager is about 26 billion kilometers away, the radius of our observable universe is estimated to be 46 billion light years. So in all these examples, the uncertainty in the result depends not only on how many decimal places of pi we use, but also on how accurately we know the other inputs.
Meanwhile, pi has been calculated to 314 trillion decimal places. StorageReview’s computation took about 110 days on a powerful server. Their purpose is rather different from finding the circumference of a table. These exercises test the performance and reliability of computers and software, among other things.
Pi has one value. How much of it we need depends on what we are doing.
The same question arises when someone asks how long a project will take. A spacecraft docking with a space station requires careful timing. Installing software may be discussed in weeks. A metro rail project may be discussed in months or years. The task, the purpose of the answer, and what we actually know determine how specific we can reasonably be.
Once, a client asked me to provide an exact estimate. I could not resist and had to ask him, “Would you want me to be exact or would you prefer me to estimate?”




