Code accompanying the manuscript "Cultural transmission of move choice in chess" by Lappo, Rosenberg, and Feldman (2023), available at doi.org/10.1098/rspb.2023.1634.
First, please ensure that you are working on a machine with an Intel or AMD processor (i.e. x86_64 architecture). (If you are working on a Mac computer with an M1/M2 processor, follow the guide below that does not use Nix. This is because the cmdstan package has issues with selecting the right precompiled headers.)
0. Install nix and download the code repository with data. To do this, first run
curl --proto '=https' --tlsv1.2 -sSf -L https://install.determinate.systems/nix | sh -s -- install
to install nix.
Then, download the repository that includes the input data file from Zenodo: zenodo.org/doi/10.5281/zenodo.10038192.
1. Run nix develop to log into a shell with all scripts and packages necessary for processing data, fitting the model of move choice, and generating all figures. Note: this may take a long time.
2. The data file data/csv/caissa_clean.csv is the exact dataset we used to fit the model and draw figures, containing 3448853 games. If you want to access the original database of games, it is available from Caissabase. You will need to use Scid vs. PC to export all games into a PGN file and put it into data/pgn (use this guide). Then, run scripts parse.py and clean.py from the data_processing folder to produce the .csv table caissa_clean.csv. You do not need to run parse.py and clean.py if you use the .csv file we provide on Zenodo. Note that Caissabase data may be updated in the future by retroactively adding more games, in which case the results can be slightly different.
All the .csv files we use in our analysis are also duplicated at this GDrive link.
3. Run the script to generate additional tables caissa_counts_by_ply_{i}.csv that will be used in making streamplots in Figure 3.
cd data_processing
python generate_tables.py
4. Fit the Bayesian model of move choice in chess with Stan (Section 5(b)(iv) of the paper).
cd model
python prepare_data.py
Rscript --vanilla run_stan.R
5. Run each of the scripts in the make_figures/ folder to generate all figures in the main text and the supplement.
cd make_figures
python figure_2.py
python figure_3.py
python figure_4.py
python figure_5.py
python figure_s2.py
python figure_s3.py
python figure_s4.py
python figure_s5.py
Rscript --vanilla figure_s1.R
6. Email me at [email protected] with any questions or issues.
To reproduce analysis without the help of the Nix system, clone the repository and decompress the source .csv file as described above. Then, make sure that your version of python has all the necessary packages installed by running pip3 install -r requirements.txt. Creating an environment with miniforge can help. Install cmdstanr following this guide. Finally, install required R packages: c("tidyverse", "bayesplot", "posterior", "scales", "ggh4x").
Now you are ready to run all the scripts. Follow the guide above starting with 3.