A runner training for a spring marathon sets a goal of three hours thirty minutes, does the division in her head, and programs 5:00 per kilometer into her watch. She executes almost perfectly. She crosses the line at 3:30:58. Nearly a minute slow, on a day she paced flawlessly. The problem was not in her legs. It was a single rounding step she took before the start: she treated the marathon as 42 kilometers. It isn't. It's 42.195, and that stray 195 meters is the difference between the number she wanted and the number she got.
The odd part is how ordinary the mistake is. Every distance in the sport is a clean figure except the two that matter most on race day. A 5K is 5,000 meters. A 10K is 10,000. Then the half marathon lands on 21,097.5 meters and the full on 42,195, and suddenly the math has a decimal tail that punishes anyone who trims it.
A distance nobody would have chosen on purpose
The marathon commemorates a run that, by the usual telling, was about 40 kilometers. When the modern Olympics revived the event in 1896, the Athens course from the town of Marathon to the Panathenaic Stadium ran roughly 40 km, near 24.85 miles. For the next several Games the distance floated. Paris in 1900 and St. Louis in 1904 each measured something different, all in the neighborhood of 40 km, none identical.
The number we use today comes from one specific race: the 1908 London Olympics. The course started at Windsor Castle and finished inside the White City Stadium, and when it was measured it came to 26 miles and 385 yards. The extra 385 yards is the stuff of running folklore, usually credited to extending the finish so it sat in front of the royal box, though historians argue about which royal adjustment did what. What is not in dispute is the total: 26 miles is 41,842.94 meters, 385 yards adds 352.04 meters, and the sum rounds to 42,195.
That figure could have stayed a one-off quirk of a single London course. Instead the International Amateur Athletic Federation, the body now called World Athletics, ratified 42.195 kilometers as the official marathon in 1921, and the number stuck. The half marathon is defined as exactly half of it, which is why it carries the even uglier 21,097.5 meters. Neither distance was designed for arithmetic. Both were inherited.
What the last 195 meters does to your pace
Rounding feels harmless because the error looks small. Forty-two versus 42.195 is a gap of less than half a percent. But pace is time divided by distance, and a plan is that division run backward: you take the time you want and spread it over the distance you think you are covering. Feed the wrong distance in and the pace that comes out is quietly wrong in the same direction, every kilometer, for the entire race.
Take the 3:30 marathon. That is 12,600 seconds. Spread over a rounded 42 km, the math asks for exactly 5:00 per kilometer. Spread over the real 42.195, it asks for 4:58.6. Those two paces look interchangeable. They aren't. If you run the 5:00 you calculated, but the course is the 42.195 it actually is, you finish at 3:30:58. The 1.4 seconds per kilometer you gave away by rounding did not stay small. It compounded across 42 splits into most of a minute.
The half behaves the same way, scaled down. A 1:45 goal over a rounded 21 km wants 5:00 per kilometer; over the true 21.0975 it wants 4:58.6; run the rounded pace and you come in around 29 seconds late. Not race-ending, but real, and entirely self-inflicted before you laced up.
The cost of each rounding, side by side
Not every shortcut costs the same. The distance you feed your plan matters more than whether you think in miles or kilometers, and one common rounding is far more forgiving than the others. Here is the same 3:30 marathon goal, computed against several ways of writing the distance, then run over the true 42.195 km:
| Distance you plan against | Pace it asks for | Finish if you run that pace on the real course | Gap to goal |
|---|---|---|---|
| 42 km | 5:00.0/km | 3:30:58 | +58 s |
| 42.2 km | 4:58.6/km | 3:29:58 | −2 s |
| 42.195 km (exact) | 4:58.6/km | 3:30:00 | on target |
| 26 mi | 8:04.6/mi | 3:31:46 | +1:46 |
| 26.2 mi | 8:00.9/mi | 3:30:09 | +9 s |
| 26.219 mi (exact) | 8:00.6/mi | 3:30:00 | on target |
The pattern worth noticing: rounding a marathon to 42 km costs you nearly a minute, but rounding it to 26.2 miles costs only nine seconds. That is not a coincidence. The traditional "26.2" already carries two decimal places of the imperial distance, so it lands much closer to the truth than the tidy-looking "42 km" does. Meanwhile "26 miles" flat is the worst of the set at almost two minutes off, because it throws away that fractional tail entirely. If you are going to round, round toward more precision, not less. Better yet, use the exact constant and stop guessing which rounding is safe.
Where splits tables quietly lie
The rounding trap has a cousin that lives one level deeper, in the splits sheet itself. A marathon at 8:00 per mile is not 26 clean miles. It is 26 full miles plus a trailing 0.219 of a mile, and in metric it is 42 full kilometers plus a final 0.195. Plenty of splits tables simply drop that partial. They hand you 26 tidy rows, the last one reading 3:28:00, and let you believe that is your finish. It isn't your finish. It's your split at mile 26, with roughly two more minutes of running still to account for.
A table that omits the trailing segment isn't just incomplete, it's actively misleading on the one number you care about most. A correct splits sheet shows the partial explicitly, labeled as the fraction it is, so the final cumulative time actually lands on your goal rather than a marker short of it.
There is a second, subtler failure in how the cumulative column is built. If a tool rounds each split to the nearest second and then adds those rounded values together, the rounding errors accumulate. Round 26 splits that were each half a second high and your total drifts more than ten seconds off the pace you asked for. The fix is to accumulate from the unrounded times and round only at the moment of display, so the last row reconciles to the finish within a second instead of wandering away from it. When a splits table and a finish-time field disagree, this is usually why.
Even, negative, or positive: spending one time budget three ways
All of the above assumes even pacing, where every mile gets the same share of the clock. That's the sensible default and the right starting point, but it's a choice, not a law, and the alternatives change the numbers in the table meaningfully.
A negative split spends the first half slower than the second. Run the first half of that 3:30 marathon at 4:59.6/km and the back half at 4:57.6 and you still average out to the goal, but the early rows of your splits sheet read slower and the late rows faster. Most coaches consider this the disciplined way to race a marathon, because the physiological cost of going out too hard is nonlinear and tends to arrive around mile 20. A positive split is the mirror image, faster early and fading late, and it is what actually happens to the runner who built her plan on a rounded-down distance: she banks time she thinks she has, then bleeds it back over the closing kilometers.
The reason to know your exact even-split pace first is that it is the anchor every strategy bends around. You can't plan a deliberate 20-second negative split if the baseline you're adjusting from is already 58 seconds wrong from rounding. Get the honest even-pace number, then decide how to distribute it.
The meters don't move when the units do
One last source of confusion is worth clearing up, because it is the flip side of the rounding trap. A marathon is 42.195 kilometers and 26.219 miles at the same time. Those aren't two different distances, they're two labels for one distance of 42,195 meters, and the meters don't change when you toggle the display. This is why a well-built calculator keeps everything in meters and seconds internally and treats miles, kilometers, minutes-per-mile, and minutes-per-kilometer as views over that canonical pair. Switch units and your entered distance should convert cleanly, your finish time should not budge, and your splits should re-index from mile rows to kilometer rows without the total drifting.
When a tool instead stores the distance in whatever unit you happened to type and reconstructs it on the fly, small conversion errors creep in exactly where the rounding errors already live, and the two compound. The distance is a physical fact fixed in meters. Everything else on the screen is a rendering of it.
None of this requires memorizing 42,195 or doing long division at the start line. It requires a tool that holds the exact constants, keeps the trailing partial honest, and reconciles the splits to the finish. That is the whole design brief behind Milemarq: pick the marathon preset and it loads 42.195 km exactly, not a rounded stand-in, and the splits table below your goal pace carries the final partial all the way to the line.