% ssscoring(3) Version 3.1.0 | Speed Skydiving Scoring API documentation

Name

SSScoring - Speed Skydiving Scoring high level library in Python

---

Synopsis

pip install -U ssscoring

Have one or more FlySight speed run track files available (can be v1 or v2), set the source directory to the data lake containing them.

# synopsys.py
from ssscoring.calc import aggregateResults
from ssscoring.calc import processAllJumpFiles
from ssscoring.calc import roundedAggregateResults
from ssscoring.flysight import getAllSpeedJumpFilesFrom

DATA_LAKE = './resources' # can be anywhere
jumpResults = processAllJumpFiles(getAllSpeedJumpFilesFrom(DATA_LAKE))
print(roundedAggregateResults(aggregateResults(jumpResults)))

Output:

python synopsys.py
                           score  5.0  10.0  15.0  20.0  25.0  finalTime  maxSpeed
01-00-00:v2                  472  181   329   420   472   451       24.7       475
resources test-data-00:v1    443  175   299   374   427   449       25.0       449
resources test-data-01:v1    441  176   305   388   432   442       25.0       442
resources test-data-02:v1    451  164   295   387   441   452       25.0       453

Speed run summary example: https://raw.githubusercontent.com/pr3d4t0r/SSScoring/refs/heads/master/resources/SSScoring-speed-run-summary.png

SSScoring processes all FlySight files (tagged as v1 or v2, depending on the device) and SkyTrax files. It aggregates and summarizes the results. Full API documentation is available at:

https://pr3d4t0r.github.io/SSScoring/ssscoring.html

The SSScore apps are available from:

• web app: https://ssscore.streamlit.app
• SSScore for Mac (universal .dmg)
• SSScore for Windows (installer)

---

Installation and Requirements

• Python 3.12 or later
• pandas and NumPy
• The requirements.txt file lists all the packages required for running SSScoring or using the API
• The devrequirements.txt file lists all the pacckages required at build time

---

Quickstart

• The SSScoring interactive quickstart notebook for Jupyter/Lucyfer is the fastest way to learn how to use the library
• The ssscore command line tool implements the same functionality as the interactive quickstart, can be used for scoring speed skydives from the command line with minimum installation
• Read the SSScoring API documentation

SSScore end-user apps

Analyze single tracks or a group of tracks using the SSScoring API in a full-featured web application.

• Mac: download link coming soon
• SSScore web app - requires Internet connectivity
• Windows: download link coming soon

ssscore command line tool

ssscore is a comnand line tool that scores one or more speed skydiving files with as little user participation as possible. It supports options for specifying the DZ altitude MSL in feet and for "simple training output" that shows rounded speed values, useful for physical log book updates.

ssscore -e 616 -t ./TRACKS

Produces this outout:

elevation = 187.76 m (616.00')
Processing speed tracks in quickstart-example/...

                                   score  5.0  10.0  ...  25.0  finalTime  maxSpeed
R3_13-32-20:v2                       490  187   333  ...   490       24.2       493
quickstart-example R1_09-20-26:v1    325  135   211  ...   319       25.0       328
quickstart-example R2_11-00-34:v1    476  185   333  ...   315       24.9       481

[3 rows x 8 columns]

Total score = 1291.00, mean speed = 430.33

See the ssscore man page for details on this quickstart tool.

---

Running the development stand-alone apps

While the web-based app shows the single and multiple jumps scoring functions as part of a single app, they are two distinct executables. During development and for local execution, it's easier to run them from the command line.

These commands assume that the code is installed in a Python virtual environment and that the streamlit package is installed.

Prepare the local run-time environment

Installs all the required packages via pip -e . in the local target, and it only needs to run once per session, and only after make test or make clean.

make local

Scoring a multiple jumps set

# installs all the required packages via pip -e .
# it only needs to run once per session, and only after make test or make clean
make local
streamlit run ssscoring/ssscoremultiple.py

These commands will start a new SSScore instance, current branch version, in the system's default web browser.

Running a SSScore container

The docker-compose.yaml files included in the master SSScoring repository are ready to run on any Docker-enabled system with access to Docker Hub. They'll pull the latest SSScore web app image from the cloud for local use. Users don't need to build their own images in order to use this feature.

• Intel:

   docker compose -f dockerize/docker-compose.yaml up

• ARM:

   docker compose -f dockerize/ARM/docker-compose.yaml up

Once it's running click on SSScore web app link to score your jumps or go to http://localhost:8501 in your favorite web browser.

---

Description

SSScoring provides analsysis tools for individual or bulk processing of FlySight GPS competition data gathered during speed skydiving training and competition. Scoring methodology adheres to International Skydiving Commission (ISC), International Speed Skydiving Association (ISSA), and United States Parachute Association (USPA) published competition and scoring rules. Though FlySight is the only Speed Measuring Device (SMD) accepted by all these organizations, SSScoring libraries and tools also operate with track data files produced by these devices:

• FlySight 1
• FlySight 2
• SkyTrax GPS and barometric device

SSScoring leverages data manipulation tools in the pandas and NumPy data analysis libraries. All the SSScoring code is written in pure Python, but the implementation leverages libraries that may require native code for GPU and AI chipset support like Nvidia and M-chipsets.

Features

• Pure Python
• Supports output from FlySight versions v1 and v2, and SkyTrax devices
• Automatic file version detection
• Bulk file processing via data lake scanning
• Automatic selection of FlySight-like files mixed among files of multiple types and from different applications and operating systems
• Individual file processing
• Automatic jump file validation according to competition rules
• Automatic skydiver exit detection
• Automatic jump scoring with robust error detection based on exit altitude, break off altitude, scoring window, and validation window
• Produces time series dataframes for the speed run, summary data in 5-second intervals, scoring window, speed skydiver track angle with respect to the ground, horizontal distance from exit, etc.
• Reports max speed, exit altitude, scoring window end, distance traveled from exit, and other data relevant to competitors during training
• Internal data representation includes SI and Imperial units; implementers may choose either one when working with the API

The latest SSScoring API is available on GitHub: https://pr3d4t0r.github.io/SSScoring/ssscoring.html

The SSScoring package can be installed into any Python environment version 3.9 or later. https://pypi.org/project/ssscoring

SSScoring also includes Lucyfer/Jupyter notebooks for dataset exploratory analysis and for code troubleshooting. Unit test coverage is greater than 92%, limited only by Jupyter-specific components that can't be tested in a standalone environment.

What is a data lake?

A data lake is a files repository that stores data in its raw, unprocessed form. A speed skydiving data lake often has one or more of these types of files:

• FlySight versions 1 or 2 files
• SkyTrax files
• Video files (MP4 or MOV of whatever)
• PDFs of meet bulletins and related event information
• Miscellaneous other junk

SSScoring identifies FlySight and SkyTrax files regardless of what other file types are available in the data lake. SSScoring also identifies speed files from other types of tracks (e.g. wingsuit) based on the performance profile and scoring windows. Tell the SSScoring tools where to get all the track files, even if they are several levels deep in the directory structure, and SSScoring will find, validate, and score only the speed skydiving files regardless of what else is available in the data lake. The only limitation is available memory. SSScoring has been tested with as many as 467 speed files during a single run, representing all the training files for a competitive skydiver over 10 months.

Additional tools

• nospot shell script for disabling Spotlight scanning of FlySight file systems
• umountFlySight Mac app and shell script for safe unmounting of a FlySight device from a Macintosh computer
• DumbDriver Mac app to disable the SMART SSD / SSHD / HDD kernel driver, used when the FlySight 2 isn't detected in Mac systems that have SMART drivers installed.

---

Building the code

General

1. Procure a Python 3.12.0 or later virtual environment for building (venv or pyenv)
   • Install both Apple Silicon and Intel Python run-times if you plan to make universal binaries
2. Ensure that make is available
3. All shell commands retain backward compatibility with bash except for those specific to macOS
4. All artifacts are generated to ./dist
   • make clean wipes the whole directory out
5. Versioning: Release version follow standard conventions. Beta and test releases use .99.99 and decrement with every release. Artifact versions in the 99-80 range are considered "throw away development code."

Building all the Python, system-independent artifacts

make clean && make all

Generates all the Python weel, images, and documentation artifacts.

-rw-r--r--  1 ciurana  staff  55699 May 23 08:46 ssscoring-2.98.97-py3-none-any.whl

Mac build

1. The builds are biased to Apple Silicon-first
2. Ensure that Xcode command line tools are installed: xcode-select --install

macOS build

make clean && make all && make app

drwxr-xr-x  3 ciurana  staff     96 May 23 08:51 SSScore.app
-rw-r--r--  1 ciurana  staff  55699 May 23 08:46 ssscoring-2.98.97-py3-none-any.whl

macOS Intel build

Requires an Intel-only, x86_64 Python virtual environment configured somewhere reachable in the file system. The user specifies this venv's path to the activate script in .env. Example:

PYTHON_INTEL_VENV="~/Python-3_14_4-x86_64/bin/activate"

The build command:

make clean && make all && make app-intel

drwxr-xr-x  3 ciurana  staff     96 May 23 08:56 SSScore-Intel.app
drwxr-xr-x  3 ciurana  staff     96 May 23 08:51 SSScore.app
-rw-r--r--  1 ciurana  staff  55699 May 23 08:46 ssscoring-2.98.97-py3-none-any.whl

Universal binary build

The universal binary build requires an Apple Developer Connection account and a signing certificate. See the Apple documentation for details. Building a universal binary isn't necessary for everyday, personal use. The build process is biased toward universal binaries for third-party distribution.

make clean && make all && make app && make app-intel && make universal

drwxr-xr-x  3 ciurana  staff     96 May 23 08:51 SSScore.app
-rw-r--r--  1 ciurana  staff  55699 May 23 08:46 ssscoring-2.98.97-py3-none-any.whl
./dist/SSScore.app: satisfies its Designated Requirement
✓ signed: ./dist/SSScore.app
Architectures in the fat file: ./dist/SSScore.app/Contents/MacOS/SSScore are: x86_64 arm64

Productized macOS build

Generates a disk image with universal binaries of all the SSScoring tools, signed, notarized, and ready for distribution:

make clean && make all && make mac

The disk image is generated to ./dist as SSScoring-3.0.0.dmg, alongside the Python wheel and all other project artifacts. The full product build takes between 3 and 30 minutes to complete because it relies on compulsory Apple services for artifact notarization.

drwxr-xr-x  3 ciurana  staff         96 May 23 09:01 DumbDriver.app
-rw-r--r--@ 1 ciurana  staff  190369691 May 23 09:06 SSScore-3.1.0.dmg
drwxr-xr-x  3 ciurana  staff         96 May 23 08:51 SSScore.app
-rw-r--r--  1 ciurana  staff      55699 May 23 08:46 ssscoring-2.98.97-py3-none-any.whl
drwxr-xr-x  3 ciurana  staff         96 May 23 09:01 umountFlySight.app

Publishing a new release

Requirements:

• A GitHub account with 2FA and your keys already define
• The gh package (installed from Homebrew or pacman, as applicable)

Sequence:

1. On the macOS building machine, after successful notarized .dmg creation:

   make release

2. On the Windows building machine, after succcessful installer creation:

   make release

The artifacts are listed, along with the current build version, in the SSScore releases page.

Windows build

1. The builds are based on the latest versions of Windows 10, but Windows 11 works best
2. Only Windows Intel is supported.
3. Ensure that MSYS 2 is installed.

There is a concerted, ground up, no-compromises, no bullshit directive tokeep this codebase free of Windows tooling unless there is no alternative. That's non-negotiable project and repository policy.

At the time of writing, the optimal MSYS2 session is UCRT64.

Windows building steps

Requires an Intel-only Python virtual environment configured somewhere in the MSYS + user file system (something like C:/msys/home/joeuser/Python-3_13_3-x86_64). Whenever possible, avoid /c/whatever paths and keep everything in Unix-land.

Important: PyInstaller support for Windows lags about a minor release version vs the macOS and Linux versions.

The build command, in the virtual environment:

alias winmake='make -f Makefile.win'
winmake clean && winmake all

-rw-r--r-- 1 crystal None 56233 May 23 12:36 ssscoring-2.98.97-py3-none-any.whl
drwxr-xr-x 1 crystal None     0 May 23 12:37 SSScore

SSScore is packaged as a Windows onedir application. Ensure to deploy all the contents of ./dist/SSScore together, or the application won't start. The SSScoring API, Python and data science components are all in the SSScore/_internal sugdirectory.

Productized Windows build

Windows builds aren't signed or notarized, but they do come with a productized installer. Contributors to the project are welcome to help notarize the SSScore installer for Windows. As far as building a distributable version, it's super easy, barely an inconvenience:

winmake clean && winmake all && winmake installer

This produces a standar Windows installer wizard. SSScore is installed as a first class citizen in C:/Program Files like any other software with all the expected Windows app behavior.

drwxr-xr-x 1 crystal None        0 May 23 12:43 SSScore
-rwxr-xr-x 1 crystal None 79116394 May 23 12:44 SSScore-3.1.0-Setup.exe
-rw-r--r-- 1 crystal None    56233 May 23 12:42 ssscoring-2.98.97-py3-none-any.whl

./dist ends up with three distributable packages, ready to go:

• SSScore - onedir executable
• SSScore-3.1.0-Setup.exe standard installer
• Python wheel with the latest code (the same one you'll find in PyPI)

Docker build

Dockerized SSScore is a good alternative to the native installations for users who prefer to work in a web browser environment.

These instructions are system-agnostic, as long as a somewhat recent version of Docker is available in the local system.

You may want to modify these two files to match your Docker environment and your Docker Hub account:

• dockerimagename.txt
• dockerimageversion.txt
  • The build process automagically updates the version number during the build, to make it consistent with the current branch. That's the reason why Docker images for human consumption are always built from the master branch. This version number is consistent across master, PyPI, and Docker Hub.

Dockerized SSScore for Intel

git fetch && git pull && git checkout master

make clean && make dockerize

• Gets the latest stable version
• The Dockerfile will pull the same wheel from PyPI
• The Intel Docker image is ready for deployment

Dockerized SSScore for ARM

git fetch && git pull && git checkout master

make clean && make dockerize.arm64

• Guaranteed to work on Apple Silicon and RasPi 5
• Uses the latest stable version from PyPI

IMAGE                                        ID             DISK USAGE   CONTENT SIZE   EXTRA
pr3d4t0r/ssscore-p:2.10.13                   1216fd1debf0       2.25GB             0B
pr3d4t0r/ssscore-p:latest                    1216fd1debf0       2.25GB             0B
pr3d4t0r/ssscore:2.10.13                     c9771dea3985       1.91GB             0B
pr3d4t0r/ssscore:latest                      c9771dea3985       1.91GB             0B

Yes, the images are big. Apache dependencies in pandas are the reason. A lot of surgery, love, and care went into slimming down the macOS and Windows standalone executables.

---

Contributors

Name	GitHub
Jochen Althoff	@Quadriga14193
Eugene Ciurana	@pr3d4t0r
Michael Cooper	@FlySight
Nik Daniel	n/a
Alexey Galda	@alexgalda
Marco Hepp	n/a
Stepan Sgibnev	@kotek14

See Also

SSScoring API documentation - github.io SSScore app on-line - Streamlit Cloud ssscore(1) https://github.com/pr3d4t0r/SSScoring/blob/master/ssscore.md

License

The SSScoring package, documentation and examples are licensed under the BSD-3 open source license.