Marathon Time Predictor

How the marathon time predictor uses race results and mileage

The marathon time predictor estimates performance over 42.195 kilometers using your weekly mileage alongside one or two race results, with the target distance kept fixed throughout. It implements the marathon equations from Vickers and Vertosick’s 2016 research and supplementary material, available under the paper’s CC BY 4.0 license. This version adds unit conversion, input checks and an interface around those equations. You can therefore inspect a mileage-based estimate instead of assuming that doubling your half-marathon result will describe the full marathon.

With one race, the model uses its marathon extrapolation plus typical weekly distance. Adding two different race distances activates another equation that uses the relationship between the performances as well.

Those inputs distinguish this page from the general Race Time Predictor. They also explain why changing weekly mileage can change the answer even when the race time stays fixed. The calculator describes an association in the model; changing an input doesn’t demonstrate a training effect for you personally.

Choose one or two recent races

You can start with a completed 5K, five-mile, 10K, ten-mile or half-marathon race, entering its finish time and adding a second distance if you have another comparable result. The calculator sorts those distances for you, but the results should still describe similar fitness. Each race also has average, difficult and fast condition choices, which apply distance-specific adjustments from the source model. Those choices are broad labels, so use them with that meaning in mind rather than expecting a detailed correction for your particular hills, wind or weather.

A second race should be a different distance. Entering two races at the same distance doesn’t provide the distance-to-time relationship needed by the two-race equation. Also, a longer race that took less time than the shorter race after adjustment needs checking.

The tool limits race-input pace to 4:00 through 20:00 per mile, following the source data filtering used in the implementation. It may reject a combination that implies a marathon pace faster than your longer race.

These checks help catch unhelpful extrapolations. If a result is rejected, inspect the original race distances, clock values and condition labels rather than changing genuine results solely to make the calculator accept them.

Understand the mileage-based prediction

Enter typical weekly running distance, rather than the largest single week or the distance you intend to reach later. You can use miles or kilometers; the equations convert weekly distance to miles internally.

For a 1:50:00 half marathon in average conditions and 35 miles per week, the one-race model gives approximately 4:01:16. The race-only 1.07 comparison is about 3:50:56. The difference shows why a mileage-based estimate isn’t simply the same prediction with a new label.

Equations used by this implementation

D is race distance in meters, T is adjusted race time in seconds, and W is typical miles per week.

R = T × (42195 ÷ D)^1.07
V = 0.16018617 + 0.83076202 × (42195 ÷ R) + 0.06423826 × (W ÷ 10)
One-race time = 42195 ÷ V

For two races, subscript 2 identifies the longer distance.

k = ln(T₂ ÷ T₁) ÷ ln(D₂ ÷ D₁)
K = 1.4510756 − 0.23797948 × k − 0.01410023 × (W ÷ 10)
Two-race time = T₂ × (42195 ÷ D₂)^K

The weekly input accepts up to 150 miles. That ceiling is a software guardrail, not evidence that the full range is suitable or equally well supported. Raising W to obtain a faster displayed time doesn’t establish that increasing your training would produce that outcome.

Predict using training data only (no recent race)

If you don’t have a recent competitive race result or time trial, switch to the Training Data Only mode. This mode implements the published regression model from Dr. Giovanni Tanda (2011), published in the Journal of Human Sport and Exercise (Vol. 6, No. 3, pp. 511–520). Rather than extrapolating from a shorter race, the Tanda model predicts marathon pace directly from two training variables measured over the eight-week period preceding the taper:

• Average weekly distance ($K$): Mean kilometers run per week across the 8-week block.
• Average training pace ($P$): Mean pace across all logged runs in the 8-week block (in seconds per kilometer), including easy runs, long runs, and workouts.

The formula reflects two physiological realities: the exponential decay term ($140.0 \times e^{-0.0053 K}$) models the diminishing endurance benefit of increasing weekly mileage, while the linear pace term ($0.55 \times P$) captures the runner’s underlying aerobic speed foundation.

Important methodological limitations

To preserve scientific honesty, keep the following context in mind when reading a training-only prediction:

• Small study cohort: Tanda’s original formula was derived from 46 marathon races run by 22 recreational male runners aged 28 to 54, with finish times between 167 and 216 minutes (~2:47 to ~3:36). The standard error of estimate was approximately 4 minutes within that cohort.
• Broader performance ranges: For runners finishing outside 2:47–3:36 (including runners targeting 4 to 5+ hours), the model has less empirical validation and greater variance. The calculator displays a notice when an input produces a time outside this range.
• No fabricated confidence bands: We do not present arbitrary ±15-minute confidence intervals, proprietary age penalties, or artificial volume caps that have no foundation in the published literature.
• Association, not causation: The model describes observed correlations. Intentionally forcing your easy training runs to be faster solely to lower your average pace input $P$ is counterproductive coaching that increases injury and overtraining risk without guaranteeing a faster race.

If you want a more precise prediction, consider running a 5K or 10K time trial (see our Track Pace Calculator) and using the primary Race + Mileage mode.

Read marathon pace and checkpoints

Your estimated finish time is converted into average mile and kilometer paces, followed by checkpoints showing how a constant-pace run would unfold on the clock. For the 4:01:16 example, you’ll see about 9:12 per mile or 5:43 per kilometer, with halfway at about 2:00:38 and ten kilometers at about 57:11. These readings explain the time represented by the estimate. If you choose another actual goal after reviewing your preparation, its checkpoints will need to be calculated from that goal instead.

You can copy or print the summary, including the original race and mileage inputs. Printing with checkpoints adds the longer table. Keeping those inputs attached makes it easier to understand a saved result when you revisit it after another race.

Notice that the reference half-marathon input was 1:50, while the marathon halfway checkpoint is slower. The model doesn’t assume you should reach halfway at your half-marathon race effort.

If you choose a different actual goal time after reviewing your preparation, build its splits with the Mile Pace Calculator. That lets you plan the goal you’ve selected instead of using the predicted number automatically.

Use the estimate for race planning

Treat the result as one piece of information alongside your training, course and race plan. The inputs don’t include every factor that can affect a marathon. A calculated finish time doesn’t establish readiness to maintain its pace for the whole distance.

The two-race model isn’t an average of two one-race predictions. It uses another equation, so adding a second result can change the answer in a way that initially looks surprising. The one-race comparison remains useful for inspecting that change.

Check any warning about unusually low mileage, inconsistent race relationships or a very long predicted finish. These messages describe limits of the calculation. You can compare another suitable recent result without treating the fastest available estimate as the preferred one.

For workout references from the race you entered, the Training Pace Calculator addresses a different question. It doesn’t convert this marathon prediction into a complete training schedule. Keep race prediction, goal selection and workout planning as separate decisions, each supported by the information relevant to it.