What Happened on Lap 87 at the Last Soul Ultra¶
Lap 87 of the 2026 Last Soul Ultra was Mark Dowdle's fastest loop of the race: 38:16, against a 51:42 average across 94 total laps. It happened at roughly the 83-hour mark, approximately 580 kilometers in. Dowdle finished at 94 yards (630.4km); Ayyoub El Osrouti at 93 yards (623.7km).
The Numbers Behind Lap 87¶
The Last Soul Ultra is a backyard ultra: a 6.7km loop restarted every hour on the hour, last runner standing wins. The 2026 edition started August 14 at a former military base near Duesseldorf with 169 starters. It ended early Tuesday morning, August 18, with Dowdle as the last person standing. The race was the first backyard ultra on German soil to go past 90 yards; the 2025 edition ended at 67. Caroline Maria Bolte set a German women's record at 52 yards (about 348km).
Here is what the split data looks like for lap 87.
| Metric | Value |
|---|---|
| Lap 87 time | 38:16 |
| Race average (across 94 laps) | 51:42 |
| Delta from average | 13 minutes 26 seconds faster |
| Percentage faster than average | 26% |
| Approximate distance into race | ~580 km |
| Approximate elapsed time | ~83 hours |
A 51:42 average across 94 loops needs a little context to land. Each loop gives you one hour. A 51:42 moving time means roughly eight minutes of rest per loop, compounded across more than 80 hours without normal sleep. The loop itself is flat enough that terrain variation does not account for swings of this magnitude.
Sleep deprivation in multi-day ultras does not produce faster splits. Anyone who has crewed a 24-hour race or run a stage race across multiple days has seen the physical evidence: lap times in the final third of a race that would have seemed impossible to run that slowly just six hours earlier. The standard pattern is degradation. Splits widen. Fast laps cluster early; slow laps accumulate late.
Lap 87 runs 26% faster than Dowdle's race average, at a point where his body had been in motion for the better part of four days. That delta, at that position in the race, does not come from finding extra fitness. Something else was operating on the pace.
What the Livestream Showed¶
The 2026 Last Soul Ultra was livestreamed, and lap 87 was captured on footage. That footage is what sparked the argument still running in Reddit threads, Discord servers, and the comment sections of the backyard ultra community.
What the footage shows, described plainly: Dowdle running directly behind El Osrouti through the loop. When El Osrouti slowed, Dowdle slowed. When El Osrouti accelerated, Dowdle matched it. El Osrouti attempted at several points to open a gap; the gap did not hold. Neither man was willing to lead.
That is what is visible. The rest of what has circulated online is interpretation.
The observable number is the time: 38:16. Both men ran that loop at that pace. The split clock does not distinguish between who was driving and who was responding. It records what happened.
This matters because the debate about lap 87 mostly concerns intent, which the footage cannot settle. Was the proximity deliberate? Was it sustainable? Was it something else entirely? The people arguing about this are arguing about motivations that neither man has fully spoken to publicly, in a sport where two runners sharing a 6.7km loop is normal and expected. What the footage cannot show and what no viewer can know is what either runner was thinking or planning.
What the clock did show: this was lap 87 of 94. Both men kept going for seven more loops after this one. El Osrouti completed loop 93 and could not answer the bell for 94. Dowdle ran loop 94 alone, which is what a backyard ultra asks of the winner: one yard more than anyone else, run solo. In a backyard ultra, the win belongs to the last person standing, and there can be no shared finish.
Viewers called the footage a psychological duel. Others used stronger language. The data does not take a side. The data says: two runners, one loop, 38:16.
What Happens to Split Data When You Switch From Racing a Clock to Racing a Person¶
When you are racing a clock or a distance, split behavior follows a predictable shape. Pace is constrained by what your body is doing: the accumulation of fatigue, your current fuel state, how your heart rate is tracking against the effort level. Fast splits cluster toward the front of the race, when those factors are in your favor. Slow splits cluster toward the back. The degradation curve is gradual, occasionally jagged, but directionally consistent. A runner who paces well shows a gently widening spread across the race; a runner who goes out too hard shows a cliff somewhere in the middle third.
When you are racing a specific person, the reference point for effort shifts. The primary input stops being your body's internal signals and becomes the gap in front of you. That gap is not fixed. It responds to what the other runner does, which means your effort responds to what the other runner does, often before you have consciously decided anything. It is a feedback loop with a moving target.
The data signature of this shift is recognizable once you know what you are looking for. It is not a sustained faster block: not a section of the race where several consecutive laps trend faster, which would suggest a real change in conditions or pacing strategy. It is a single loop, well outside the established pace range, appearing in the part of the race where nothing about accumulated fatigue supports it. One loop that doesn't fit. Then a return toward the surrounding range.
Lap 87 is that loop. Sitting 13 minutes and 26 seconds below a 51:42 average, at roughly the 580-kilometer mark of a 630-kilometer race, it is surrounded by laps that were already considered fast for this stage. The 38:16 does not sit next to a cluster of 42-minute or 44-minute laps that might suggest a sustained faster block. It sits inside a race being run at a fundamentally different average speed.
This pattern appears across ultra formats. It shows up in 100-mile finish-line surges where a runner who has been at 18-minute miles for the last 30 miles suddenly runs a 12-minute mile after spotting a competitor ahead. It shows up in multi-lap track ultras where a rival pulling within auditory range triggers a single fast lap with no physiological explanation behind it. Any time a format lets two runners be precisely aware of each other's position over a measured distance, this data signature is possible.
El Osrouti ran the same 38:16 loop. Both men were in the same frame of reference, responding to the same changing input. The clock recorded them equally.
The 38:16 is the data signature of a race that had shifted from hours-remaining to the runner in front of you. What produced that shift, and how each runner experienced it, is not something any of us watching can know. What we can say is that this is what the split data looks like when the target moves.
How to Find This Pattern in Your Own Race Splits¶
If you have run a backyard ultra, a 100-miler, or any timed ultra with lap data, your race file almost certainly contains at least one split that does not fit the surrounding pattern. Some are easy to explain: a long aid station stop, a headlamp that died, a section that was steeper than expected. Others need a different kind of reading.
Here is how to tell whether your anomalous fast split was driven by fitness or by something else.
1. Where does it sit in the race timeline?
Fitness-driven fast splits cluster in the first third. The body is freshest; effort is sustainable; the pace reflects actual capacity. A fast split in the final third of a long ultra almost always has an external cause: a competitor pulling within range, a pacer arriving, a finish-line draw, a section the runner remembered as significant. The further from the start your anomalous lap sits, the less likely physiology explains it.
2. What do the surrounding laps look like?
A real fitness shift produces a build: two or three laps trending faster before the anomalous one, as conditions change or a runner finds a second gear. A decision-driven spike drops suddenly from laps that were at or below average pace. No build, no gradual change. One step in one direction, then a return. Look at the three laps before your fastest split and the three laps after. Is there a ramp leading into it, or does it appear without preparation?
3. How large is the delta?
Five to ten percent faster than surrounding laps is within the normal noise band of a long ultra: terrain variation, a cooler section, a stop that did not eat into moving time the way you expected. Fifteen to thirty percent faster is outside that band. A number that size means something changed beyond normal variability. Dowdle's 26% at lap 87 is the range where you stop looking for a terrain explanation and start looking for a different one.
4. What came after?
A fast lap followed by a return to roughly average pace suggests controlled effort: the runner spent energy deliberately and then resumed a sustainable range. A fast lap followed by a dramatic collapse or a DNF suggests the effort was an overspend. Seven more completed loops after lap 87 places Dowdle's split in the first category, not the second.
Running this analysis does not require specialized software. Pull up your last race, sort the laps by time, and look at the distribution across the race timeline. Identify the laps that sit more than two standard deviations from your mean. Where are they in the race? What surrounds them? That pattern usually tells you more about what you were actually doing at that moment than your post-race memory will.
If your last race is on Strava, NavRun pulls the lap data directly and shows you the full distribution across the race timeline. The pattern is clearer when you can see all 94 loops at once instead of a column of numbers scrolled off a watch screen.
The split analysis framework in this post goes deeper in Ultra Race Post-Mortem: Read Your Data, which covers cardiac drift, the positive split cliff, and aid station bleed as a structured debrief process. If you want to apply the same lens to your own last race, that is the next read.