I didn't know either. But if you think about it for a moment, you'll probably land close to the real answer.
A due date is the date on which the baby is 50% likely to have been born, on or before. In other words, it's the median of the distribution of when spontaneous labour begins.
The curve on the left shows that distribution. The higher the line is, the greater chance that the baby will be born on that day.
We can visualise the due date using this probability curve. The blue area on the left represents the 50% chance of the baby arriving on or before the due date, and the pink area on the right is the other 50% (the chance of it arriving after the due date).
The line between them (the vertical line that splits the area into two equal halves) is the due date.
Interestingly, the due date isn't the single most likely day for the baby to be born. Because of the skewed shape, the peak of the curve (the mode) falls a couple of days after the due date.
Before we go further, it's worth being clear on one idea: conditional probability. That's maths-speak for something simple: how your beliefs change once you have more information.
Imagine a die has been rolled, and you haven't seen the result yet.
Three of the six faces — 4, 5 and 6 — are highlighted. Each face is equally likely, so there's a 50% chance the roll was at least 4.
Now imagine you're told something new: someone tells you the roll wasn't a 1 or a 2. We can rule those two options out entirely.
With 1 and 2 ruled out, only 3, 4, 5 and 6 remain, each still equally likely. Now it's 3 and 4 that make up the "lower" half. So we know there's a 50% chance the roll was at least 5.
The new information changed our perception of the value of the die and shifted where the halfway point sits.
The same logic applies to pregnancy. Let's imagine we're ticking along, day by day.
Every day that goes by with no baby coming means we know a little bit more information about when the baby is due — it must be a day in the future.
Say we reach 38 weeks and the baby still hasn't arrived. Just like ruling out the 1 and 2 on the die, we can rule out everything to the left of "today."
The halfway point of what's left has shifted slightly later than the original due date.
And it keeps moving: progress to 39 weeks, still no baby, and the same thing happens again — rule out more of the curve, and the new 50% mark drifts later still.
If the baby still hasn't arrived by the original due date, the "updated due date" is now several days further into the future.
If you use the median definition of "due date" throughout a pregnancy and keep recalculating it, it keeps receding. Every day you don't go into labour pushes the "50%-likely date" a little further away.
Of course, once the baby arrives, you'll have neither the time, the energy, nor the mental faculty left to think about any of this.