Weightlifting App Analyze Data

This repo provides a variety of scripts to analyze the data from Weightlifting App 💪

Data setup

You can either use your own data or use the example data provided in the data/example-*.wld files.

Using your own data

1. Open Weightliting App on your iPhone and navigate to User -> Settings -> Export All Data.

2. Send the data to yourself (eg. via email)

3. Place your .wld file in the data folder and pass its path to load_training_dataset or the source adapter used by the analysis. Personal files under data/ are ignored by Git.

Setup

Python 3.10 or newer is required.

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -e ".[notebooks]"

Usage

Use the notebooks in src/ to explore and read analyses. Use commands in scripts/ for repeatable or headless runs. Reusable calculations live in importable modules under src/; notebooks and scripts should call the same implementation rather than duplicate it. See docs/analysis-architecture.md for the full convention.

For analyses that would otherwise walk the nested export structure, load the canonical dataset instead. It returns flat workout, exercise, and set DataFrames with documented columns (see docs/training-dataset.md):

from training_dataset import load_training_dataset

data = load_training_dataset("../data/example-chappy.wld")
data.sets[data.sets["display_name"] == "Flat Barbell Bench Press"]

Key analyses:

• src/analyze_training_program.ipynb: live full-history training style, direct and estimated muscle-group volume, frequency, current split inference, and like-for-like press progression. A run writes its recommendation to outputs/training-program-recommendation.md.
• Use src/analyze_big_three.ipynb for lifetime Big Three progression, annual snapshots, historical trends, and one-year projections.
• Use src/analyze_bodyweight_strength_evals.ipynb for bodyweight-aligned strength history, all-attempt Pareto frontiers, and social-card exports for bench, squat, deadlift, and overhead press.
• scripts/analyze_dexa.py analyzes DEXA history, lean-mass trend residuals, dated scan paths, and modeled one-point body-fat contours.

To forecast the bodyweight at which a target DEXA body-fat percentage is reached:

source .venv/bin/activate
python scripts/forecast_bulk_ceiling.py --output-dir outputs

The run reads data/dexa.csv and writes a Markdown report, a probability-curve CSV, and a chart. It is seeded, so repeat runs reproduce byte for byte.

Modeling assumptions are set with --target-body-fat-pct, --simulations, --seed, --measurement-error-pp, --partition-noise-scale, --resample-unit, and --max-weight-lb. The report prints every one of them alongside the result, along with a held-out predictive score.

Planning inputs are optional and have no silent defaults. Without --current-bodyweight-lb (or --weight-log, a weekly CSV with Week of and Average columns) headroom is measured from the DEXA scan weight and labelled as such. Without --weekly-bulk-rate-lb no duration is reported:

python scripts/forecast_bulk_ceiling.py \
  --current-bodyweight-lb 200 --weekly-bulk-rate-lb 0.5

To refresh the bodyweight and DEXA CSV exports from Weight Log.xlsx:

python scripts/convert_weight_xlsx.py
python scripts/convert_dexa_xlsx.py

Development

Install the test dependencies and run the complete suite from the repository root:

python -m pip install -e ".[dev,notebooks]"
python -m pytest

Agent instructions are in AGENTS.md. Shared chart conventions and the baseline-comparison workflow are in docs/chart-style.md.

Contributing

Feel free to contribute to this repo by adding your own scripts or improving existing ones.

License

This project is open-sourced under the MIT License - see the LICENSE file for details.