Small, runnable Python scripts for launch cadence, schedule slip, ISS passes, and a Kepler-III check on real exoplanet data. Built as a portfolio / Git practice project for SyncHeart-adjacent space data work.
python3 -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt| Script | What it does | Data source |
|---|---|---|
launch_cadence.py |
Counts launches per quarter for the top rocket families; writes cadence.png |
Launch Library 2 (free, ~15 calls/hour) |
launch_slip.py |
Snapshots upcoming launches to data/upcoming_YYYY-MM-DD.csv (including window_start, window_end, and net_precision on new snapshots); --diff A B compares two snapshots and reports slip days, with placeholder dates listed separately |
Launch Library 2 |
slip_history.py |
Summarizes slip across every consecutive pair of snapshots; writes data/slip_summary.csv and slip_summary.png. Counts only timed comparisons |
Local snapshots |
slip_by_provider.py |
Aggregates slip per rocket (parsed from launch name, since older list-mode snapshots omit provider); writes data/slip_by_provider.csv. Counts only timed comparisons |
Local snapshots |
iss_passes.py |
ISS rise / peak / set over Spring, TX for the next 48 hours (America/Chicago) | CelesTrak TLEs via Skyfield |
kepler_check.py |
Fits log(a / M^(1/3)) vs log(P) for confirmed exoplanets; Kepler predicts slope ≈ 2/3 |
NASA Exoplanet Archive TAP |
python src/launch_cadence.py # a few LL2 pages; stay under the free rate limit
python src/launch_slip.py # writes today's CSV under data/
python src/launch_slip.py --diff data/upcoming_2026-11-02.csv data/upcoming_2026-11-09.csv
python src/slip_history.py # slip summary table + chart across all snapshots
python src/slip_by_provider.py # per-rocket slip table
python -m pytest tests # unit tests for the slip statistics and precision filter
python src/iss_passes.py
python src/kepler_check.pyPass --include-low-precision to launch_slip.py --diff to also print placeholder and precision-change rows in the main listing (they are still labeled, and they are never mixed into the default summary tables).
launch_slip.py is meant to be run on a schedule (e.g. weekly). Each run freezes the public NET (no-earlier-than) date for every upcoming launch. Diffing two files measures how many days each launch slipped:
- positive
slip_days→ NET moved later (delay) - negative → NET moved earlier
- launches that disappear between snapshots usually flew or were removed
- launches already flown (
Success/Failure/Partial Failure) are dropped from the later snapshot
Over months this becomes your own delay dataset — useful for arguing about vehicle scarcity, pad congestion, or provider backlog without proprietary telemetry. Commit the CSVs (or store them as release artifacts) so the history is reproducible.
When Launch Library 2 only knows the day, month, quarter, or year of a launch, it publishes a placeholder NET such as 2026-11-30T00:00:00Z ("sometime in November"). Swapping a placeholder for a real time — or a timed NET for a month-end date — can look like a multi-day "slip" that never happened.
Each NET is classed as timed (hour, minute, or second precision), day, or coarse (month or longer). A comparison only counts in the slip statistics when both NETs are timed. A slip is a timed NET moving later by at least 1 hour; smaller moves are T-0 refinements and are reported as minor adjustments. Precision changes and same-class placeholders are listed by --diff but excluded from slip_summary.csv and slip_by_provider.csv.
Where precision comes from. New snapshots (from this change onward) store LL2's own net_precision field, plus window_start and window_end, and use mode=normal so those fields are present. The three existing weekly snapshots (2026-11-02, 2026-11-09, 2026-11-16) were taken in list mode and do not have those columns, so a heuristic is used: a NET at exactly 00:00:00 UTC with a zero-length launch window (or no window recorded) is a placeholder (month-or-coarser if it falls on the last day of a month, otherwise day). The same heuristic, checked against the live net_precision field on 2026-11-16 in Launch Delay Watch, agreed on all 100 upcoming launches. A real launch scheduled at exactly 00:00:00 UTC with no window would be misread as a placeholder.
Reproduce the tables and chart from the committed snapshots:
python src/slip_history.py
python src/slip_by_provider.pyVerified by running the commands above against the committed CSVs:
data/slip_summary.csv (mean / median / max are over the timed comparisons only):
| from | to | matched | timed | excluded | slipped | mean (d) | median (d) | max (d) |
|---|---|---|---|---|---|---|---|---|
| 2026-11-02 | 2026-11-09 | 39 | 7 | 32 | 2 | 1.142 | 0.000 | 7.000 |
| 2026-11-09 | 2026-11-16 | 41 | 6 | 35 | 1 | 0.990 | 0.000 | 5.942 |
The three timed slips (≥ 1 hour later):
| Launch | Snapshot pair | NET move | slip (d) |
|---|---|---|---|
| Mir | Demo Flight | 2026-11-02 → 2026-11-09 | 2026-11-04 05:00Z → 2026-11-11 05:00Z | 7.000 |
| Falcon 9 Block 5 | USSF-627 | 2026-11-02 → 2026-11-09 | 2026-11-09 03:17Z → 2026-11-10 03:07:15Z | 0.993 |
| Falcon 9 Block 5 | USSF-317 | 2026-11-09 → 2026-11-16 | 2026-11-13 04:23Z → 2026-11-19 03:00Z | 5.942 |
data/slip_by_provider.csv (timed observations only):
| rocket | observations | slipped | share | mean (d) | max (d) |
|---|---|---|---|---|---|
| Mir | 1 | 1 | 1.000 | 7.000 | 7.000 |
| Falcon 9 Block 5 | 4 | 2 | 0.500 | 1.734 | 5.942 |
| Long March 12 | 2 | 0 | 0.000 | 0.000 | 0.000 |
| Long March 8A | 2 | 0 | 0.000 | 0.000 | 0.000 |
| Soyuz 2.1b | 2 | 0 | 0.000 | 0.000 | 0.000 |
| Spectrum | 2 | 0 | 0.000 | 0.000 | 0.000 |
Before this filter, the 2026-11-09 → 2026-11-16 max was 18.792 days and Mir led the per-rocket table at mean 12.896 / max 18.792 (2 of 2 observations slipped). That 18.792-day move was Mir going from a timed NET (2026-11-11T05:00:00Z) to a month-end placeholder (2026-11-30T00:00:00Z) and is now excluded. The earlier timed Mir slip of 7.000 days remains. Other unfiltered artifacts: Vulcan VC6L +0.746 d and Electron +0.229 d (day placeholder gaining a clock time), and Starship −16.052 d (month placeholder replaced by a timed NET).
The sample is still small: 13 timed comparisons across two weekly pairs. Most upcoming rows in these snapshots are month- or year-end placeholders, which is why matched is much larger than timed.
- Launch Library 2 free tier is rate-limited (~15 requests/hour). Cadence is capped at 4 pages; slip uses one call (
mode=normal, limit=50). - Space-Track.org is not used here (needs an account). Swap CelesTrak for Space-Track if you need the full catalog.
- Outputs are regenerated from live APIs; charts and CSVs in this tree are examples from the first run. Slip tables and
slip_summary.pngare regenerated from the committed weekly snapshots with the precision filter applied.
MIT — data remains under each provider's terms.