Based on your in-game stats (eliminations, deaths, healing for example) tries to guess your current competitive Skill Rating (SR). After that, it includes a report of what it thinks each of your hero's SR's would change if you played that character MUCH more often, but still flexed to the other heroes on your list. To generate advice, in short, without giving away secret sauce, the neural network not only compared how you are relative to people your rank, it also is able to figure out whether or not a stat is important SPECIFICALLY for you. For instance, suppose you have high eliminations but low hero damage. In the advice section, anything based on "eliminations" would have a very small effect.

I'm using deep learning! I am a PhD student working on these sorts of things, so some of the techniques here are entirely unique to this project. I'd like to publish a paper on it some day!

David, aka theC0der, is an avid Overwatch player (it most certainly contributed to his tendonitis) and PhD student with a concentration in artificial intelligence. His love for programming emerged in high school (thanks Mr. Williams) and hasn’t stopped growing since. Seriously. His girlfriend is constantly trying to get him to go outside away from his computer for like 10 minutes. She’s lucky if she can get him to eat something.

David is very serious about providing the best information possible to people and if you think for one second he’s just in it for the money, you’re an idiot. BTW his gf is writing all of this. She loves him a lot and wants you to know that you’re getting all his figurative blood, literal sweat and happy tears when you use his website/app. And he wants nothing back from you except for you to be able to use the tools to improve your game! Ok, well you could be a Patreon supporter, if you want, but no pressure! ; )