I Compared Garmin and Strava’s Race Day Predictions and Both Were Wrong in Different Ways.
As the average running watch becomes increasingly high-tech, a new norm is emerging in pre-race rituals: a week before a big event, you check your watch, hoping the algorithm will tell you good news.
Both Garmin and Strava offer race time predictions, but how accurate are they really? I tested each platform during the Brooklyn Half Marathon on May 16, 2026. Here’s what I experienced, and what I think every runner should know about what these tools actually measure.
How Garmin’s Race Predictor Works
Garmin’s race prediction feature has been a staple of its mid- and high-end devices for over a decade. It provides estimated finish times for 5K, 10K, half marathon, and marathon races, and works primarily by converting estimated VO2 max values into race pace equivalents. Garmin claims it also uses personal data (age, gender) and recent training history to smooth out short-term fluctuations.
Of course, this model assumes you’ll run a perfect race. This means optimal pace, ideal weather, full race preparation, proper nutrition, mental resilience, and so on. And while Garmin displays a heat or altitude indicator on the VO2 max widget when conditions affect this estimate, this indicator isn’t carried over into the race prediction itself. Therefore, it more accurately predicts your aerobic ceiling rather than your true expected finish time.
On more expensive devices, such as the Forerunner 965 and 970, Garmin offers a more advanced “Race Forecast with Course and Weather,” which can be applied when entering race information into the Garmin Connect calendar. It can take into account elevation changes along the course and environmental factors, such as heat on race day (which can be considered a sign of good fortune).
For context, I used my Garmin Forerunner 970 as my primary watch for the race. (I’m currently working on comparing the 970 and the Amazfit Cheetah 2 Pro on race day.)
How Strava’s performance predictions work
Strava’s performance prediction feature is relatively new, having launched in April 2025. Instead of relying on theoretical VO2 max estimates, Strava’s system uses artificial intelligence and “real-world activity data”—your own and that of other runners. This activity data is said to include over 100 attributes, including a runner’s historical activity, recent training load, best results, and the results of other Strava users with similar training histories.
Because each race’s distance is calculated independently, Strava claims its system provides greater accuracy for each distance, rather than extrapolating a single metric across all distances. The model generates a new prediction after each run data upload and requires at least 20 runs over a 24-week period.
In my circle of runners, most consider Strava’s predictions to be somewhat unstable and “random” compared to Garmin’s. This is understandable, as I’ve personally seen my predictions change dramatically after one bad (or exceptional) run. On the other hand, Strava acknowledges that its model heavily weights historical data, which can lead to delays for runners recovering from injury or a long break. One significant limitation: performance predictions don’t take terrain or elevation into account. They assume a flat course, similar to a treadmill.
What Garmin and Strava predicted before my race
Garmin’s prediction: 2:00:51. This would have been a personal best. In retrospect, this provides a useful insight into how Garmin’s model works. The prediction almost certainly reflected high VO2 max values achieved during training runs, which were converted into an idealized result on race day.
Strava’s prediction: 2:10:34. This is a much more conservative estimate, even slower than my last official half marathon last September (2:05). Given that Strava relies heavily on historical performance data, including all-time bests, this prediction could have been based on the pace of my easy training runs rather than race data, or it could have reflected a training block that didn’t include much high-intensity running specific to a half marathon.
The difference between the two predictions—almost ten minutes!—is a story in itself. For comparison, for a 10K run in early May, Garmin predicted 54:04, while Strava predicted 58:14—a difference of more than four minutes. (That run was ultimately run with extreme caution due to a knee injury, so I don’t have any interesting results for you.) But the pattern is telling: Garmin tends toward optimistic predictions, while Strava tends toward conservative ones .
My results: right in the middle
Since this was a real-world test, I want to highlight the real-world conditions that influenced my time. The Brooklyn Half Marathon had a fantastic course advantage, with the entire second half being a downhill course. Many runners strive for personal bests precisely because of this.
Unfortunately, the weather at races in May is unpredictable, and on race day it was unbearably hot compared to training. The temperature was at least ten degrees higher than any of my pre-race runs—a significant factor for a runner who quickly gives up in the heat. Furthermore, there were no clouds on the descent. I made special stops at water stations to monitor my heart rate, even at the cost of slowing my pace. Any runner who has previously pushed themselves too hard in the heat knows how the mental state changes: finishing healthy is more important than finishing fast.
My final time was 2:04:49. This result matches the Garmin and Strava results suspiciously closely. Garmin’s prediction was 3 minutes and 58 seconds faster , while Strava’s was 5 minutes and 45 seconds slower . So, Garmin was more accurate, but neither prediction was so wrong as to cause the runner to make a catastrophically poor pace decision.
Remember that Garmin’s race prediction feature is designed to show you what your aerobic system is theoretically capable of under ideal conditions. For shorter distances like 5K and 10K, this ceiling and reality can be quite close. For half marathons and marathons, the gap widens—and it widens dramatically when race-day conditions differ from the cool, easy training runs that formed your VO2 max estimate. Runners who use Garmin’s prediction as a target pace without considering the heat, course difficulty, or their own race readiness risk starting too fast and paying the price in the second half of the race.
Strava’s reliance on historical data and data from comparable athletes can lead it to underestimate runners who are in excellent shape but haven’t posted recent workouts for the Strava algorithm to consider. If you primarily train at an easy pace and rarely compete, Strava may not have enough information to determine your current maximum. Furthermore, the Strava community notes that predictions can vary significantly from one workout to another, making it difficult to build confidence on race day with a constantly changing target load.
Result
Garmin estimates your aerobic potential under ideal conditions; Strava estimates the actual performance of a runner with your training history. Both approaches have their drawbacks, and both will mislead you if you take their results as definitive.
Personally, the Garmin readings were closer to the actual finish time, but at the same time, following this forecast on a hot day would have been more dangerous if I hadn’t been careful. Whatever your forecast on race morning, remember to take the weather forecast into account, know the course, and allow a little extra time.