I would love to integrate this spread sheet and the site in general with auto updating features using my R package tf2statr, http://www.teamfortress.tv/28872/tf2statr-compile-and-aggregate-tf2-stats-in-r.

Maybe do something like this tennis scoreboard
http://www.fair-play.com/wp-content/photos-drawings/TN-1736-2.jpg

with the courts are replaced by map names and the clocks in the middle. You could combine the bottom cut outs into 1 long grid to put on the left of it or something.

That's what I meant by "JSON API" so I am looking at the raw output. There is so much stuff in there I had to get rid of some of it (with a option to keep) just to make sense of everything.

I am not entirely sure about the accuracy though. I wonder if some configs or plugins need adjusting after the fact that is done on the backend part of logs.tf. In any case, the best solution for this project would still have been to output the raw stats and maybe add in another function to adjust stuff. If there are some adjustments it would be easy to do for leagues and tournaments since they almost always have the same configs throughout.

There have been some corrupted logs but I haven't checked for accuracy. That is why in the i55 data set Froyo players have 17 or 16 games played even though they all played in the same matches.

I was like 2 inches away from throwing it all into a database, however I thought people would like a personalized library instead ... especially with the player name archives which still looks kinda awful. The problem with R databases is the problem with R as a whole, there are a million packages to do the same thing and only like 5 are supported.

I may just make a shiny app so you don't have to go into R if there is enough support. If you don't like R there is a lot of python - R bridges like this one.

Managing logs and figuring out tournament, season or even 'last 3 games' stats is a largely manual process. logs.tf shouldn't include this since its whole purpose is to be lightweight and fast. This utility makes it easy to grab these logs and interrogate them using R. I chose R mostly because its fastest for me to code in but also because it innately has the statistical framework for conversion to other packages like machine learning and forecasting.

You can find the package on github.

In terms of functionality you can do things like "logs <- queryLogstf(tournament = "Insomnia52")" and grab all the logs from the tournament then get a giant table of mean per match stats using something like "tab <- aggregateStats(logs)" A more complete use case can be found as a vignette or in a compiled HTML form.

Note about web scraping worries

I use the logs.tf API to get the raw stats and then parse them instead of going to the page and scraping that. So, yes you can track the "number of donks" per game. I do use comp.tf to get the log IDs for events, however, I cache the results in an archive so it isn't called again. If we share our archives by sending me the file or by submitting a pull requests, the archive could encompass all pro TF2 events.
In a related note, UGC and ESEA leagues are not supported due to logs not being up on comp.tf. Creative queries to the logs.tf search API can be done using the tools provided but top level querying will result in an error. I encourage everyone to help populate these fields and use the website cause its a standardized and easy way to look back at the pro TF2 landscape.

Future

- ss support
- Team queries
- Make a "Media notes" automatic document for upcoming games.