Tim Sevenhuysen, also known as the guy who collects massive amounts of LoL data on Oracles Elixir, also spends his time developing Shadow, a set of analysis tools designed to support pro teams. Today, they're announcing a new automatic scrim data collection feature, and Tim is here to explain why it should make pro scrims much more effective.
Pro teams here in NA play their scrims on a special tournament realm, a private client and server Riot designed for running tournaments and special events. As it turns out, the special privacy of those servers means teams are on a scrim data lockdown.
Tim Sevenhuysen: “Riot Games and the professional teams access the Tournament Realm to run their practice games without being spied on. For one, it’s based in LA so it’s better ping for teams, but it also allows them to play in privacy, because nobody can see when you’re logged, figure out who you are scrimming against, how long your games last.
Along with that privacy comes the API lockdown. If you try to access the data for a specific game like you would on the normal client, those links are not active for the Tournament Realm. And Riot does not intend to change that, they value the privacy for the teams.”
So there’s no easy access to scrim data — like gold numbers or objective stats — as you would for a pro game, which leaves teams handling things the manual way.
“The problem is that there’s really nothing available right now — most teams record nothing more than the draft and who won, and they write that into a Google Sheets document, maybe Excel if they feel really advanced. There are a handful of teams I’ve spoken to who will collect a few extra things by hand —they’ll write down who got first tower or, if they have the analyst manpower, the player CS at some points.
But anything to do with gold or kills or Dragons or Barons, I haven’t come across any teams that were willing to have someone sit there and type that into a spreadsheet.”
And if a team is lacking the time, manpower, or interest in handling things manually, it usually means scrim data comes down to 'what champions won'.
"For most teams, they discuss going into scrims what they’re playing and working on that day, then they play the scrim, the coaches are watching and listening to comms, and after the scrim they debrief. The coaches give feedback, and then they move on to the next one.
For a lot of teams, learning from scrims just comes down to play-discuss, play-discuss. Any kind of trend analysis is limited to what champions you won and lost on. The ability to inject data into that and go beyond discussing small details after the game has been really limited, just because it’s so much work to capture that data."
And that’s where Tim comes in. Shadow uses computer vision to track those games and turn the action on-screen into numbers.
“The technology uses video analysis and image analysis to read data off the screen. So the teams will play their scrims, go onto Shadow, upload a video from their spectator view, and the technology will pull numbers and icons and translate that into data. Then you can analyze everything the same way you would a pro match."
According to their announcement, Shadow records "every kill, tower, dragon, Baron, inhibitor, and more, along with constant monitoring of each team’s gold earned." The team is planning to roll out more data points down the line.
And, as Tim is hoping, that could mean better analysis for the teams.
"Take first tower rate — since we’re capturing every tower kill, you can apply specific filters. Say, 'all the scrims I’ve been on red side, what percentage did I get or lose first tower? Let's talk about why.' You can filter by champions, or gold difference at 15 minutes, then you can actually go back to your team and give more advanced feedback and insight."