Illusion of persistence in NBA 1995-2018 regular season data

In this repository we provide the code and the data behind the paper "Illusion of persistence in NBA 1995-2018 regular season data" [1].

The order in which files should be run in order to reproduce (or obtain similar) results:

1. Scrape the data with data-get.py
2. Transform the full regular season record into individual team record data with data-transform.py
3. Explore Hurst exponents of the original data with data-analyze-original.ipynb
4. Explore the first passage times (streak lengths) of the original data in comparison with some random models with data-analyze-passage-times.ipynb.
5. Run shuffle the original data to obtain 95% CIs for H with data-shuffle-\*.py
6. Explore the autocorrelation functions of the original data in comparison to the autocorrelation fucntion of the shuffled data with data-analyze-correlation.ipyb (total shuffle) and data-analyze-correlation-inseason.ipynb (in-season shuffle).

Note that we have also shared the .csv files we have obtained. These are stored in the data folder.

The stats folder contains couple of custom functions taken from another repository.

Licensing: The scripts, scraped and generated data are made available under CC0.

References

[1] A. Kononovicius. Illusion of persistence in NBA 1995-2018 regular season data. Physica A 520: 250-256 (2019). doi: 10.1016/j.physa.2019.01.039. arXiv: 1810.03383 [physics.soc-ph].