|
|
|
Iraq16955 Posts
The data and the analyses you put out are all very interesting.
A few questions:
Have you tried collecting data from replays (when available)? There's some information you can't really collect that way of course (like the aggression information), but it seems like an easier (and more scalable) way of collecting build orders for instance.
Also is there a particular reason WCS America was chosen as the dataset? And have you considered collecting data from a tournament in another region and comparing the results?
|
The reason behind which tournament is chosen to record data is down to whichever Firnafth wanted to watch and record. As far as improving ease and scalability, atm that hasnt been the focus and we didnt want to be dependent on the replay packs yet. It is fully possible that at some point in the future we record through replays but for now there isn't an intention to.
|
Ziggurat - To be honest I originally planned to take data from GSL, which is my favorite tournament but does not release replay packs. Yet when I joined the Root content team, one of our other members claimed GSL as their focus, and I wanted to ensure we had a variety of tournaments covered, so I swapped to WCS Americas (for that time being). That's my region and I had hoped to watch it live while taking data, but to be honest I end up just watching VODs since there is no way I could keep up with an intensive broadcast schedule.
I do very much want to compare regions, but since not all of them release replays I think I will stick with my method, and also because think it would take a fairly long time for me, who has basic coding experience but knows nothing about replays, to figure out how to scrape things like whether an engagement is harassment or not from them. But to be honest the real reason is that I've taken data from over 300 games so far and feel deeply attached to my methods; I like watching the games.
I might consider a more careful build-order piece, derived from replay data, as a one-off I guess, but again I've become attached to my current methods.
|
We have released a tool that allows you to search games by units used (and attack type, so you can search for just units used in harassment)! Search results are sorted by player and each game can be clicked on to bring you straight to a casted vod for it!
Search WCS Winter Americas and GSL Super Tournament 1 by Unit Composition
This piece makes use of the recently finished GSL Super Tournament 1 dataset; the raw data are here.
The next dataset will be for GSL Season 2, although it will be a while before it is completed. I wanted to do WCS Spring, but with so much of it not having casted vods (apparently) the sample size and selection would have been very poor. The good news is that adding GSL Season 2 will send our sample size for KR SC2 high enough that we can do some comparisons between it and WCS Americas, which will be super interesting.
|
We generated some visualizations and statistics on game length in GSL Season 2.
Here's the article!
There were some things that were different (shorter PvZ and TvZ especially) from WCS Winter Americas, the last tournament we analyzed this way, and others that were similar (no difference in game duration between maps).
We will have more articles on GSL Season 2 in the near future.
|
Iraq16955 Posts
Interesting that the PvT and TvP charts were so similar given the rhetoric about the match-up. And obviously the PvZ chart being clustered in such a narrow band is suggestive.
Is there any significance to where on the x-axis the dots were placed?
|
No significance for the dots' placement on the x-axis; it's just used to scatter the points about so it's easier to see how they cluster.
|
We have started our series on Proxies in GSL Season 2, this time focusing on Nydus Worms with some investigation of Dark specifically.
Article here!
Next up I think we'll be doing an article on early proxies by Protosses in the same tournament.
|
We did some basic investigation into patterns of tech and production structures by Protosses -
Article here!
They proxied a bit over a quarter of their games. A third of the time, they proxied more than one kind of tech/production building!
|
Mina just completed a monumental summary of ROOT Gaming's activity over 2019!
Visualization here!
|
We finally put out an analysis based on Replay Data! It uses a set of major tournaments from late 2019 including Homestory Cup XX, Blizzcon, and WCS Fall. ZephyrBlu collected the data for this one, using his parser and scripts to scrape the birth and death times for hundreds of thousands of units from replays.
Unit Survival Rates and Life Expectancies
|
|
|
|
|
|
EPT
