Last autumn I signed up for a full marathon and put down a target of 3:30:00. Everyone's advice boiled down to one line: "run at 5:00 per kilometre". That did not answer the three things I wanted to know: what time I should pass each segment; how much slowing the first half could buy me; and whether cutting segments too finely would produce a reading no human could run. I took the same 42.195 km / 3:30:00 along three routes.
1. The six even-pace readings are not the same as "exactly 5:00 per kilometre"
Once distance and total time are fixed, six statistics are settled: 4:59 per kilometre, 12.06 km/h, 8:01 per mile, 7.49 mph, 3.35 m/s, and 1:59 per 400 m lap. The trap is in the display convention: 4:59 is raw 298.614 seconds rounded, and the 400 m cell is 119.45 seconds rounded. Switching to mi changes only the display; the marathon reads as 26.22 mi, and the number of milestones drops from 12 to 11.
The 12 timing points are given at an even 4:59 per kilometre, and I copied only four:
5 kmdue at24:53,10 kmdue at49:46.- The
21.0975 kmcell is1:45:00, exactly half the finish time. 30 kmat2:29:18and40 kmat3:19:05, the two written on my wrist.- The finish
42.195 kmcomes back to3:30:00, and the whole set reconciles.
2. The nine-segment table with a +30 second step every 5 km
Segment length 5 km, step +30 seconds between segments, slow first then fast:
| Segment | Range | Pace | Segment time | Cumulative |
|---|---|---|---|---|
| 1 | 0–5 km | 6:51 | 34:13 | 34:13 |
| 2 | 5–10 km | 6:21 | 31:43 | 1:05:56 |
| 3 | 10–15 km | 5:51 | 29:13 | 1:35:10 |
| 4 | 15–20 km | 5:21 | 26:43 | 2:01:53 |
| 5 | 20–25 km | 4:51 | 24:13 | 2:26:06 |
| 6 | 25–30 km | 4:21 | 21:43 | 2:47:49 |
| 7 | 30–35 km | 3:51 | 19:13 | 3:07:02 |
| 8 | 35–40 km | 3:21 | 16:43 | 3:23:45 |
| 9 | 40–42.2 km | 2:51 | 6:15 | 3:30:00 |
| Total | 9 segments | — | 12600.0000 seconds | The even-pace total is also 12600.0000 seconds |
There are 9 segments, the last one only 2.195 km, and the centre distance c = 4.7341. The reason for keeping this table is the two 12600.0000 seconds in the last row, a difference of 0.000000: the step does not make the target one second smaller, it only moves the same total time across different segments.
The half-split hint deserves copying word for word: with a +30 step the first half (21.098 km) should be run in 2:07:12 and the second half in 1:22:48, and the page writes "first half slower by 44:23"; with a -20 second negative split the first segment is 3:44 and the last 6:24, the total is still 12600.0000 seconds, the first half 1:30:12 and the second half 1:59:48, and the line becomes "first half faster by 29:36"; with a step of 0 both halves are 1:45:00 and that difference disappears.
3. Cut the segments finely enough and you hit the floor
The bottoming-out case is one I ran by the script's convention: 10 km / 40:00, one segment per kilometre, +60 seconds between segments. On the page that step is clamped between -30 and +30 (those two numbers are the slider's min and max), so +60 cannot be entered here; I fed the value straight into the segmentation algorithm to see how it clamps. The first segment is 8:30, the last is pushed down to the 0:30 pace floor, and 1 segment bottoms out; the plan totals 2460.0 seconds against a target of 2400.0, 60.0 seconds over, which is 1:00, and the last reading is followed by an *.
The asterisk does not mean the maths broke: that segment is clamped at 0:30 per unit and lifted back to the floor, so the total no longer equals the target. For the same marathon, with nine 5 km segments the count of bottomed-out segments is 0; change the segment length to 1 km and, in the tool's own documented example, 13 of 43 segments bottom out and the finish is 36:54 over target. The finer the segments, the more easily the step pushes the total up.
4. The tier table and the Riegel prediction: treat them as a range
Target tiers: eight for the marathon, 2:45:00 at 3:55 per kilometre, 15.34 km/h, 45:00 away from 3:30:00; 3:00:00 at 4:16, 14.07, 30:00 away. 5 km and 10 km have seven tiers each, the half marathon eight; non-standard distances get six tiers at ±15%, ±10% and ±5%, and the tightest tier for 8 km / 35:00 is 38:30.
Prediction is a different matter: T2 = T1 × (D2 ÷ D1)^k. I only have a measured 50:00 for 10 km, and projecting to the half marathon with four exponents gives
| Exponent k | Predicted half marathon | Raw seconds |
|---|---|---|
| 0.9 | 1:37:54 | 5873.9 |
| 1 | 1:45:29 | 6329.3 |
| 1.06 | 1:50:19 | 6619.2 |
| 1.2 | 2:02:28 | 7348.5 |
The page limits the exponent slider to 0.9 to 1.2, step 0.01, default 1.06. Projecting the same 10 km 50:00 to the marathon, k=1.06 gives 3:50:01; a half marathon of 1:45:00 projects to 3:38:55. I do not treat these lines as promises; the swing from changing tiers is more visible than how well I trained that day.
Every change of target means rebuilding the table, and I link the three fields with Running Pace Calculator; the split table can be laid out as plain text and pasted into a notes app, and the numbers are all computed locally in the browser. Body weight has no conversion relationship with this table: BMR Calculator uses the Mifflin-St Jeor and Harris-Benedict formulas, and BMI Calculator gives weight assessments under the Chinese standard and the WHO standard; I do not put them into the pace table.
Q: What is the difference between 4:59 and 5:00?
Rounding. 4:59 is the displayed value after rounding 298.614 seconds, paired with 12.06 km/h and 8:01 per mile; 5:00 per kilometre pairs with 12 km/h and 8:03 per mile. A target of 3:30:00 needs the 4:59 tier; running at 5:00 puts every kilometre behind that line.
Q: Why is the last segment only 2.195 km?
42.195 km cut every 5 km reaches 40 km with 2.195 km left, so there are 9 segments rather than 8, and the last segment has a pace of 2:51, a segment time of 6:15 and a cumulative 3:30:00. Segment paces are spread symmetrically about the centre, weighted by length; the shorter the segment, the more extreme the reading, and this cell is the tail of the plan.
Q: The pace cell shows 0:30. Is it broken?
No. 0:30 per unit is the pace floor; when the step is pulled too far, later segments are clamped here and the reading gets an *, and the total then exceeds the target. The asterisk only says that this step does not hold at this segment length.
Q: Can the predicted half marathon be used as a target directly?
No, I only treat it as a range. The default 1.06 gives 1:50:19, not far from the even-pace 1:45:00, while k=1.2 gives 2:02:28, which is another matter.
Back to the three questions at the start. What time to pass each segment is answered by the cumulative column of the nine-segment table: 2:47:49 at 30 km, 3:23:45 at 40 km, 3:30:00 at the finish. How much slowing the first half buys: the page gives "first half faster by 29:36" and "first half slower by 44:23", while both plans total 12600.0000 seconds; what moves is the distribution, not the finish time. How finely to cut segments safely: the test is not the number of segments but whether an asterisk appears; a bottomed-out count of 0 is what equals the target, and once 0:30 and an asterisk appear, 2460.0 seconds is 60.0 seconds more than 2400.0. As for whether the step should be -20 or +30, that is not something arithmetic can decide.