How the pace and spectator planner works
Grade adjustment
The tool adjusts pace for hills using the energy-cost model of Minetti et al. (2002), published in the Journal of Applied Physiology. Their research measured the metabolic cost of running at different gradients and produced a curve relating grade to energy cost. This tool uses the widely applied quadratic approximation of that curve, which is accurate across the gradients found on road marathon courses. Grade is clamped to plus or minus 8%: steeper readings in course data are treated as measurement artefacts rather than real terrain, since marathon courses do not sustain grades beyond this.
Elevation data and its limits
Course elevation comes from Open-Elevation, a service built on SRTM (Shuttle Radar Topography Mission) terrain data, rather than from device-recorded GPX files. Device altimeters are noisy and can overstate a course's total climb by several times its real value. SRTM data has its own limits: it reads ground or rooftop height, not road-deck height, so bridges, underpasses and dense urban canyons can produce false readings. The preparation pipeline applies a rolling-median smooth to reduce this noise before the data reaches the pacing engine, and known problem segments can be corrected by hand where a bridge or tunnel is identified.
Because of this, the elevation figures behind this tool are a chart-quality reference, not a survey-grade measurement, and are never used as a published elevation-gain statistic elsewhere on the site.
Why a window, not a single time
This tool never shows a single point-in-time prediction. A runner's actual pace on the day varies with congestion, aid-station stops, toilet stops and pacing decisions no model can see in advance. Presenting a single time invites false confidence. A window is the honest form of the answer.
Why the window is asymmetric
The window is not centred evenly on the predicted time. Most of the things that throw a plan off course, corral congestion, an underestimated fade, an aid-station or toilet stop, push a runner's arrival later than the model, rarely earlier. The window is built wider on the late side than the early side to reflect that, and it widens with distance, since small early deviations compound over the remaining course.
The split percentage reference data
Around 87% of marathon finishers run a positive split, meaning a slower second half, with the average slowdown around 8% and the fade worsening for slower finish times. This is drawn from large-scale finisher-time analyses of major-city marathons, including Boston field data of over 300,000 finishers, and from cross-race survey work on pacing patterns. Negative splits, where a runner speeds up in the second half, are typically 0 to 3%, with elite negative splits clustering at 0.5 to 2%. This tool defaults to an even split (0%) rather than the population average, since the fade a given runner should plan for is a personal estimate, not a population statistic. The reference figures are shown alongside the input to inform that estimate, not to set it.
What this tool is not
This is a pre-race planning tool, not live tracking. It has no GPS, no bib lookup and no connection to timing mats. It produces a plan built from a runner's own goal and pacing strategy, to be used alongside the race's official live-tracking app on the day, not in place of it.