Replication package for “The Buffer–Pad Rule”
Description
This dataset contains the full replication package for the working paper “The Buffer–Pad Rule”. It includes three self-contained archives: (1) buf_finance_repro_v1_fixed.zip – κ(t) construction for US macro-financial data and the 2025Q3 blind forecast (TCMDO, GDP, bank Tier-1 capital, Federal Reserve assets). (2) buf_sre_repro_v1.zip – toy Site Reliability Engineering (SRE) demo with HTTP toy servers, k6 load-test scripts and summary outputs for baseline vs buffer-gated scenarios. (3) NAB_master.zip – selected Numenta Anomaly Benchmark series and scripts used for the industrial / social time-series tests. Each archive contains its own README, Python or k6 scripts, data extracts and SHA-256 checksums so that all figures and tables in the paper can be reproduced.
Files
Steps to reproduce
1.Download and unzip each archive into a separate folder. 2.For macro-finance tests, follow buf_finance_repro_v1/README.txt and run the Python scripts in scripts/ to reconstruct κ(t), reproduce the 2008–2020 blind tests and the 2025Q3 forecast. 3.For NAB tests, follow the NAB README and run the provided Python scripts on the selected NAB series to reproduce the R1–R2–R3 buffer regimes. 4.For SRE demo, follow README_sre_demo.txt: start the toy HTTP servers and run the k6 scripts for baseline vs surge, with and without the buffer gate, then compare the summary outputs. 5.Verify SHA-256 checksums using the SHA256SUMS.txt files to confirm file integrity.