Coach and analyst of the Toronto Defiant brings VALORANT stats to the public.
Overwatch League analyst and coach, and founder of the Overwatch analysis tool "Winston's Lab", Dennis "Barroi" Matz has announced the launch of CypherCam, a statistical analysis tool for VALORANT. Based on image recognition software, the website will provide statistical analysis to the public. From pickrates to winrate stats based on agent and map, CypherCam provides first statistical insights into the strength of agents in VALORANT. We sat down with Barroi to talk about his new project.
For those not familiar with your work in Overwatch, could you explain how you came to esports analytics, what Winston’s Lab was and what you’re doing now with your skillset?
In 2016 I created Winston’s Lab, which was for most of the games lifetime the only bastion of Overwatch esports stats, out of the pure enjoyment of statistics and the lack thereof in the competitive Overwatch scene - most of the statistics now used on the official OWL broadcast were in fact born on Winston’s Lab. Over the years I have been working with multiple professional esports teams as a statistical advisor and at some point about 2 years ago I got hired fulltime as a coach and analyst for Toronto Defiant, a team in the Overwatch League. That’s still my job and I love it a lot, my main personal focus is to develop as a coach and learn from the great people that are surrounding me.
You’ve so far recorded 360 maps to gather VALORANT stats. Can you talk about the selection of that data set?
The data set is a mixture of games from multiple different streamers of all skill levels, preferably playing unrated games. The main goal was to hit diversity, since I don’t want to just have footage of one person one-tricking Jett. Those 360 maps have thousands of different players in them and thus I hit a mark where I’m confident the data is more or less representative of what works for the average player.
Given no public API access to VALORANT stats just yet, how does CypherCam collect game data?
CypherCam uses artificial intelligence to collect data. More specifically the process of collecting data consists of 2 phases. In the first phase, multiple neural networks scan videos at a superhuman rate and spit out gamelogs for the matches they are watching. At this point they are more reliable at humans at figuring out if they are looking at a Brimstone or a Breach - and trust me you do get confused sometimes.
The second phase takes those gamelogs and converts them into stats. A broadcaster for OWL once called this “The magic and science behind Winston’s Lab” or now CypherCam. This second part is the science of it, here I have to spend all my brainpower, figuring out what kind of statistics I can create, which ones shine a new light on the game, etc.
CypherCam is pretty barebones. What kind of analytics are currently available? What are you working on? Where do you think you will be able to take it?
Currently available to the user are super generic statistics. This means pickrates for each hero by map, a grade of how defense/attack sided each map is and winrates by defense/attack for each agent - this one is probably the most interesting for now, for example Cypher is great at defending sites, while Jett is best used to push into them.
I already have access to a lot more in the database, but making it accessible takes a bit of time and this is more of a passion project for my off-time than actual work. That said I will make more stats available in a broader form, meaning it won’t look pretty, but there will be a whole bunch of tables to explore for everyone. Over time I hope to find new, interesting ways to express what is happening inside of the game with the help of numbers and thus give some insights on how you can perfect your game as a player.
For more VALORANT stats, you can follow CypherCam on its Twitter page or on Barroi’s personal Twitter account.