tools to collect data from the spreadsheet and load it
Spreadsheet with the ranked data JSON files: https://docs.google.com/spreadsheets/d/1g0jW7k-GJiHueQ0ZVYe4WilupnUkBYLVlbB9GEdqQ98/
The zips directory contains monthly/seasonally zip files with all the JSON
files for that month/season.
The 2 scripts are meant to download all the ranked data listed in the spreadsheet and conveniently load it for analysis/usage. The scraping part is mostly handled well, but some data cleaning/reformatting is necessary.
To download the JSON files, go on the google sheet and use File -> Download ->
CSV. Then use the saved CSV files in the amq_scraper.py script to get the JSON
files. The usage is: amq_scraper.py <sheet csv> <output dir>
Some older files may contain JSON errors that have to be corrected manually. However, if you download the JSON files from this repo rather than scraping them again, I should have those issues taken care of. Another issue is that some links to older JSON files are now dead, but you can still get them here.
The amq_loader.py script has 2 functions: 1 for loading all ranked files from
a directory and 1 for reformatting the data to make it consistent. Documentation
for these is in amq_loader.py. The clean_ranked_data function will work the
way I intended only if you follow the details below about which files I decided
to exclude. As of now, I cannot be 100% sure the data is completely ready.
Upon further inspection, I decided to drop some of the early data that is missing total player count. The percent of players guessing correctly is important to analysis so correct player count does not mean a lot without also having the total number of players. Also, many of these older files are missing video links and only provide 1 anime name (appears to be English). These were probably collected with older versions of the userscripts.
The list of files I am dropping are:
- everything in
amq_2019s03(2019 season 3) - everything in
amq_2020s01(2020 season 1) - the following in
amq_2020s02(2020 season 2)amq_2020s02_01_2020-01-27_central.jsonamq_2020s02_01_2020-01-27_west.jsonamq_2020s02_02_2020-01-28_west.jsonamq_2020s02_03_2020-01-29_west.jsonamq_2020s02_04_2020-01-30_west.json
What remains should be consistent enough for the desirable percent correct analysis I am planning on. The only issue I see remaining is some of the files not having all 75 songs (not surprising because AMQ is often unreliable and you may get disconnected even with a good internet connection). Most of the time only a few are missing so it should not be a big problem for analysis. The bigger issue is several ranked matches missing in older seasons which I cannot do anything about.