Starcraft Statistics: Room to Grow
Published on 10/31/2016 12:00 PDT by ROOT Gaming
I come from a traditional sports media where statistics have been growing in importance for the past few decades and are finally seen as main stream when it comes to discussion. Now traditional sports benefit from the fact every play or punch is public information and all this meta data is generally easy to access, even for the most casual fan. Are you an football fan and want to know how often a team scores or a basketball fan who wants to know the rate at which a player turns the ball over, easy stuff. But due to the younger nature of eSports this isn't the case. While in many cases it should be easier to extract meta data from a video game through replay analysis tools that do it automatically we still haven't seen anyone find out that "when a Terran gets their first Medivac after 5 minutes their chance of winning is _____ compared to getting their first Medivac after 7 minutes." Now part of this is due to traditional sports giving their fans and researchers 100% of their data but in eSports you only have sample sizes, usually from tournaments. Now in statistics it's immeasurably more beneficial to have the entire population of data over a sample but in video games population (again, population meaning all data, not the size of the population) data is rarely, if ever, released. That leaves eSports researchers to use what we have and pigeonhole our data. We can say "at the Summer Championship Terran won ___% of games against Zerg when they got Ghosts out" but we can't speak for all players, regardless of skill. Obviously Blizzard, and most developers for their games, have this information internally and is key to balancing and designing their game.
Having said all this, I do think statistical analysis of video games through meta data is still extremely valuable but something that hasn't been used much, though recently we did see a great article about the topic. As a statistical researcher this makes me a bit sad but as has been said by many wiser persons, if you want something but it hasn't been done, do it yourself. That is part of my intention with joining ROOT to write, to give this research a platform. I am not the first to attempt this and I pray I am not the last either but I do want to present data to questions many may have and to stop misinformation within the community. Often we see statements or questions like this:
- "Everyone hates this map, no pro plays on it."
- "There is no way (insert race) can win against (insert race) after/before (insert unit) comes out."
- "(insert unit) is never used."
These questions are actually measurable and factual, we can look them up, but we rarely see that happen in discussion. I hope that this research can help answer these questions or clear up misconceptions people have. Is this map really always vetoed in pro games? Let's find out. So let's take this journey, let's answer these questions, let's not rely on our gut for how often Protoss wins when they build only Adepts and Warp Prisms. We can answer them, so let's do it.
My goal is each week to add a new table or piece of data that we can use. I can't personally go through and recreate Blizzard's meta data tables but if you have a question you want answered (For example: What race sees their win rate grow and shrink the most in Archon mode?) and I'll get to work on it.
For next week I will be looking at the age breakdown of Premier tournament winners and see which players had success at an early age and which managed to keep producing wins into the later part of their careers.
About the author:
Topher is a sports and eSports writer with a focus on statistical metadata research.
You can follow Topher on Twitter