Race predictions report

Current running race predictors from Garmin, Strava, and other platforms may be underestimating or overestimating your fitness. See how your predictions compare to real race results from real runners.

Average marathon times
Garmin Predicted 3:09
Strava Predicted 3:18
Actual Race Time 3:13

*Data from races that had both a Garmin and Strava prediction.

167
Runners Analyzed
258
Race Predictions
5.3%
Garmin's Prediction Gap
4.1%
Strava's Prediction Gap

How far off are predictions?

Bar heights show the typical gap between predicted and actual times. Lower percentage means more accuracy.

Average
Median

Does Garmin or Strava know your pace better?

Each curve shows how their predictions matched up against your actual races. Narrow curves near 0% mean they're dialed in to how you run.

-20%-10%0%+10%+20%← Ran slower than predictedPerfectRan faster than predicted →
Garmin
Strava

Do runners go faster or slower than their prediction?

The shaded area shows the density of error, and the white dot shows the median value of the distribution. When the dot is on the center line, runners race faster and slower than their prediction an equal amount of time. When the dot is to the right of the center line, runners race faster than their prediction most of the time.

-20%-10%0%+10%+20%-1.3%Garmin+2.3%Strava← Ran slower than predictedCenterRan faster than predicted →
Garmin
Strava

Are predictions better for certain races?

Predictions tend to nail your 5Ks and 10Ks, but miss wider on marathons. Why? Longer races have more unknowns — your pacing strategy, fueling, how your body holds up over hours.

See Distance Breakdown
Sign in and share your race predictions to unlock detailed analysis.
Share Your Data

How often do predictions work?

Each line shows the percentage of runners that had predictions under a given error. The faster a line reaches the top, the better.

Nailed it!0%25%50%75%100%0%5%10%15%20%Absolute error →% of runners
Garmin
Strava

Which predictor is more conservative?

For races with both Strava and Garmin predictions, this shows which one predicted a slower (more conservative) time.

Strava (76.1%)Garmin (23.9%)

The data on your predictions

Each platform's performance across your races. "Typical" is the median error. Standard deviation shows consistency — lower means less variation.

Platform Typical Absolute Error Avg Absolute Error Typical Error Avg Error Consistency
Garmin 3.9% 5.3% -1.6% -1.7% 6.9%
Strava 3.3% 4.1% +2.0% +2.6% 5.0%

Real predictions from real runners

Every row represents an actual runner's prediction and race result. See how your peers performed compared to what their watch told them.

Platform Predicted Actual Distance Difference
Garmin 3:20:02 2:57:57 Marathon 12% faster
Garmin 0:18:27 0:17:30 5K 5% faster
Access Full Dataset
Sign in and share your race predictions to access the full dataset.
Share Your Data

How to Read These Numbers

Positive error (+%): Your watch predicted a slower time than you actually ran. For example, +5% means you were 5% faster than predicted—good news!

Negative error (-%): Your watch predicted a faster time than you actually ran. For example, -5% means you were 5% slower than predicted.

Why this matters: Understanding prediction accuracy helps you set realistic race goals and know whether to trust your watch's estimates or adjust them based on real-world data.

Note: All data is self-reported and anonymized. Results may vary based on training, conditions, and individual factors.

Definitions

Error (Signed)
Error=Predicted TimeActual TimeActual Time×100%\text{Error} = \frac{\text{Predicted Time} - \text{Actual Time}}{\text{Actual Time}} \times 100\%
Measures how far off the prediction was from actual performance. Positive values mean you ran faster than predicted; negative values mean you ran slower.
Absolute Error
Absolute Error=Predicted TimeActual TimeActual Time×100%\text{Absolute Error} = \frac{|\text{Predicted Time} - \text{Actual Time}|}{\text{Actual Time}} \times 100\%
Measures prediction accuracy without regard to direction. Used to compare overall accuracy since positive and negative errors don't cancel out.