Over at Shadow, we just announced the release of our new automated scrim data collection tools for LoL, which use computer vision technology to generate scrim stats for pro teams using only uploaded VODs.
I love everything we build at Shadow, but I find this feature especially exciting for a few reasons.
First, this tech helps to solve one of the most crucial challenges of LoL stats: sample size. So many things about LoL analytics are limited by the tiny sample size of games we get during regular play. An LCS team will typically play only 4 official games on a given game patch, which is nearly useless for any kind of trend analysis or predictive work. In other words, sample size issues restrict LoL stats to being almost purely descriptive, unless you really open up your parameters and accept the uncertainty and instability that comes along with that (as I’ve done for things like my useful-but-imperfect Early-Game Rating models).
LCS teams play 10 to 15 times as many scrims as tournament games in an average week, so the potential value of statistical analysis from those games is huge, relatively speaking. You gain the ability to slice your data a lot more different ways. For example, if you want to know how your team performs in the early game when playing a certain champion in the mid lane, you can now actually get a useful sample of games and evaluate your stats while running that champion. The difficult part has always been accessing the data in the first place, and now we’re solving that problem.
Second, this tech allows teams to go beyond win rate when assessing their scrims. The standard approach, for most teams, is to base their scrim takeaways on game outcomes: we won a lot of scrims so things are going well; we lost a lot of scrims so things are going poorly. I’ve always been a proponent of using stats to contextualize results. I created stats like GSPD and GPR to identify how convincingly teams are winning or losing, and access to detailed gold data from scrims gives teams that same ability to judge their performances with more nuance. With good data access, it’s possible to realize that your 8-2 record with a certain team comp was actually inferior to your 6-4 record with a different comp, based on how the games played out.
Third, the tech is just so cool. We’ve only just begun to scratch the surface of the on-screen data available, and as we add more of those data points to our system, it’s opening up brand new lines of analysis for our clients. I love getting to innovate myself, and help others to innovate as well, so I really enjoy the conversations I’ve been having with coaches and analysts around the world about the possibilities we’re creating.
I’m happy to answer questions or give demos to anyone who wants to learn more about what we’re doing at Shadow. Just hit me up at firstname.lastname@example.org.