22 Years of Official PredictionsRe-entry Forecast Accuracy, 2004-2026
Across 2,221 re-entries from July 2004 to August 2026, official re-entry predictions made one to three days ahead missed by a typical 3.9 hours. In the final six hours, the typical miss was 17 minutes.
Version 1. Last updated . Figures are frozen for re-entries up to 22 August 2026; from 23 August 2026 the live panel carries on, updated hourly. Live forecast accuracy
How Close Were the Predictions?
How close official re-entry predictions came to the final time, by how far ahead they were issued. A prediction "held" when the final time fell inside its stated window.
Middle half of misses Typical miss Typical stated window
Graded using Orbital Radar verification methods: 13,018 official predictions for 2,221 re-entries, 25 July 2004 to 22 August 2026. Sources and methods
What 22 Years Show
More than three days ahead, 62% of objects re-entered later than predicted, and the typical prediction was 3.1 hours early. The pattern holds in 22 of the 23 years since 2004; the exception is 2025.
One to three days ahead, predictions for Starlink satellites, nearly all of them re-entering since 2020, missed by a typical 5.8 hours, against 3 hours for all other objects since 2004. At that range, Starlink came down earlier than predicted 72% of the time.
One to three days ahead, the typical miss was 3 hours in 2004 to 2013, 2.8 hours in 2014 to 2019 and 4.5 hours in 2020 to 2026, a period that includes the latest solar maximum and the first mass Starlink re-entries.
Which Objects Are Hardest to Predict?
Typical miss by period, type of object, mass, Starlink and solar activity, at each lead time.
| Group | Within 6 h | 6 to 24 h | 1 to 3 days | Over 3 days |
|---|---|---|---|---|
| By period | ||||
| 2004 to 2013 | 9 min | 34 min | 3 h | 6.4 h |
| 2014 to 2019 | 10 min | 33 min | 2.8 h | 6.3 h |
| 2020 to 2026 | 21 min | 61 min | 4.5 h | 9.2 h |
| Object type | ||||
| Rocket bodies | 11 min | 40 min | 2.9 h | 5.9 h |
| Debris | 13 min | 58 min | 2.9 h | 5.1 h |
| Payloads | 22 min | 62 min | 5.1 h | 11 h |
| Mass | ||||
| 2 t and over | 11 min | 41 min | 3.2 h | 6.5 h |
| 500 kg to 2 t | 17 min | 52 min | 3.4 h | 6.8 h |
| Under 500 kg | 21 min | 59 min | 4.7 h | 9.6 h |
| Starlink | ||||
| Starlink | 26 min | 69 min | 5.8 h | 13.4 h |
| All other objects | 12 min | 43 min | 3 h | 6.1 h |
| Solar activity | ||||
| Low | 10 min | 35 min | 2.8 h | 7 h |
| Moderate | 17 min | 59 min | 4.2 h | 9.6 h |
| High | 24 min | 60 min | 4.9 h | 7.3 h |
Closer than the 22-year typical Further off Italicfewer than 30 predictions or 10 re-entries
Solar activity is the 81-day average solar radio flux when the prediction was made: low under 100, moderate 100 to 150, high 150 and over. Space weather now
Year by Year
Typical miss by year of re-entry and lead time, with the number of re-entries graded each year.
| Year | Re-entries | Within 6 h | 6 to 24 h | 1 to 3 days | Over 3 days |
|---|---|---|---|---|---|
| 2004 | 19 | 8 min | 44 min | 3.6 h | 6.6 h |
| 2005 | 37 | 12 min | 47 min | 2.7 h | 8.4 h |
| 2006 | 28 | 8 min | 24 min | 2.9 h | 6.8 h |
| 2007 | 31 | 9 min | 29 min | 3.1 h | 7.1 h |
| 2008 | 31 | 10 min | 32 min | 3.7 h | 8.9 h |
| 2009 | 33 | 5 min | 40 min | 2.1 h | 6.1 h |
| 2010 | 34 | 8 min | 38 min | 2.4 h | 4.8 h |
| 2011 | 41 | 7 min | 26 min | 2.1 h | 5.5 h |
| 2012 | 39 | 9 min | 37 min | 3.3 h | 8.3 h |
| 2013 | 31 | 12 min | 26 min | 3.4 h | 4.5 h |
| 2014 | 44 | 7 min | 52 min | 3.4 h | 5.1 h |
| 2015 | 31 | 9 min | 25 min | 2.9 h | 7.1 h |
| 2016 | 31 | 8 min | 40 min | 2.4 h | 7.1 h |
| 2017 | 12 | 11 min | 42 min | 2.5 h | 4.5 h |
| 2018 | 44 | 9 min | 20 min | 2.5 h | 6.5 h |
| 2019 | 56 | 11 min | 42 min | 3 h | 6.6 h |
| 2020 | 52 | 12 min | 48 min | 3 h | 8.1 h |
| 2021 | 105 | 15 min | 35 min | 2.9 h | 7.1 h |
| 2022 | 139 | 14 min | 56 min | 3.4 h | 7.8 h |
| 2023 | 172 | 20 min | 52 min | 3.2 h | 5 h |
| 2024 | 365 | 22 min | 65 min | 4.8 h | 6.7 h |
| 2025 | 511 | 29 min | 64 min | 5.5 h | 12.2 h |
| 2026 | 335 | 25 min | 72 min | 5 h | 13.2 h |
Closer than the 22-year typical Further off Italicfewer than 30 predictions or 10 re-entries
Questions
How accurate are satellite re-entry predictions?
It depends on how far ahead they are made. From July 2004 to August 2026, official predictions made within six hours of re-entry missed by a typical 17 minutes; six to 24 hours ahead, by 53 minutes; one to three days ahead, by 3.9 hours; and more than three days ahead, by 8 hours. The final time fell inside the stated window for 81% to 91% of predictions.
Do objects re-enter earlier or later than predicted?
Within three days of re-entry, about 62% of objects came down earlier than predicted, mostly in the years since 2020. More than three days ahead it reverses: 62% re-entered later than predicted and the typical prediction was 3.1 hours early, in 22 of the 23 years since 2004.
Are Starlink re-entries harder to predict?
On this record, yes. One to three days ahead, predictions for Starlink satellites missed by a typical 5.8 hours, against 3 hours for all other objects, and Starlink came down earlier than predicted 72% of the time at that range. Nearly all Starlink re-entries date from 2020 onwards.
Why do re-entry predictions change?
The re-entry time depends on atmospheric drag, which rises and falls with solar activity and is hard to forecast days ahead. Each new prediction uses fresher tracking, so the window narrows as re-entry nears. One to three days ahead, the typical miss was 2.8 hours at low solar activity and 4.9 hours at high.
How to Read These Figures
- Typical miss is the median gap between a prediction and the final official re-entry time. Half of the predictions came closer, half missed by more.
- Held means the final time fell inside the prediction's stated window. Windows before 2014 were about twice as wide, so compare held rates within a period, not across periods.
- The reference is the final official re-entry time, not an independent measurement. Only re-entries with a final time precise to five minutes can be graded: 2,221 of 3,754 objects. Fewer qualify from 2013 to 2020. Every re-entry is also entered in the Orbital Register.
- Figures count predictions, as on the live panel, so objects with many updates count more. Objects with a single prediction cannot be graded.
Cite this page
Pick a style, then copy it or download it for your reference manager.Orbital Radar. (2026, November 9). Re-entry Forecast Accuracy, 2004-2026. Retrieved November 9, 2026, from https://orbitalradar.com/satellite-reentry-tracker/forecast-accuracy
“Re-entry Forecast Accuracy, 2004-2026.” Orbital Radar, 9 Nov. 2026, orbitalradar.com/satellite-reentry-tracker/forecast-accuracy. Accessed 9 Nov. 2026.
Orbital Radar (2026) Re-entry Forecast Accuracy, 2004-2026. Available at: https://orbitalradar.com/satellite-reentry-tracker/forecast-accuracy (Accessed: 9 November 2026).
Orbital Radar. “Re-entry Forecast Accuracy, 2004-2026.” Last modified November 9, 2026. https://orbitalradar.com/satellite-reentry-tracker/forecast-accuracy.
@misc{orbitalradar_satellite_reentry_tracker_forecast_accuracy,
author = {{Orbital Radar}},
title = {Re-entry Forecast Accuracy, 2004-2026},
year = {2026},
month = nov,
howpublished = {\url{https://orbitalradar.com/satellite-reentry-tracker/forecast-accuracy}},
url = {https://orbitalradar.com/satellite-reentry-tracker/forecast-accuracy},
urldate = {2026-11-09},
note = {Accessed 2026-11-09}
}TY - ELEC AU - Orbital Radar, TI - Re-entry Forecast Accuracy, 2004-2026 PY - 2026 DA - 2026/11/09 PB - Orbital Radar UR - https://orbitalradar.com/satellite-reentry-tracker/forecast-accuracy Y2 - 2026/11/09 LA - en ER -
Live figures change, so keep the date you accessed the page. How to cite Orbital Radar