Whether you’re an amateur Counter-Strike: Global Offensive player looking to climb the rankings, or looking to improve individual and team performance in a professional setting, coming across digestible, deep-level statistical analysis can be tricky.
Coaching and VOD analysis, while useful, can often be a time-consuming affair that’s more scenario-based and qualitative, or based on very top-level data that is readily available.
Driven out of a desire for self-improvement through measurable data – GGPredict is an AI-driven tool that uses quantitative analysis and machine learning to give users precise, easily viewable feedback on their CS:GO skills.
Esports Insider sat down with Co-founder Przemysław Siemaszko to discuss GGPredict’s journey to the launch of its live beta on September 24th, what people have loved about it so far, and what’s next on the cards for the Warsaw-based startup.
RELATED: Digital Creativity Labs develops real-time prediction model for esports
Esports Insider: Would you be able to tell us a bit about yourself and how GGPredict got off the ground?
Przemysław Siemaszko: I’m a data scientist and together with a group of programmers and people specialising in data in my previous company, we were playing Counter-Strike a lot. Eventually, we realised that given how much we play, we are really, really bad and we need to do something to improve.
Being data people, our first idea was we need to get statistically sound information about what makes us so bad. We had some ideas, but there were a lot of arguments about who was the worst and who should improve the quickest. We started looking for solutions that would make us better and tell us what’s wrong with our game.
ESI: So then you’ve got the stats to prove it.
PS: And to be perfectly frank we wanted to prove to one of our friends that rushing with P90 is not the optimal decision in every round. When we noticed there’s no such tool available online that would quickly tell us what’s wrong or what’s right, we started looking at game files and how the game processes everything that happens.
Quite soon we realised its possible to take data from the game files and start working on it. With a couple of weeks of work, we realised that basically everything in the game is in our hands. We can take demo files or live games and then present them in data and numeric format to quantify everything there is in the game.
When we started getting the data out of game recordings a lot of the stuff we’re collecting getting is not possible to see with the naked eye. Obviously, if one player stood in spawn and another was running around their distance differed. But apart from that, you cannot measure reaction times, distance covered, accuracy, very tiny differences. That’s how it got started.
ESI: We see a lot of virtual coaching offered on a one-to-one basis, usually from a pro player or someone considered an ‘expert’. How does GGPredict’s AI-driven approach set you apart from those competitors?
PS: Our coaching idea came from the fact that we can predict the outcomes of the game and see the details in the data that nobody else can. Compared to a traditional coach, we have way more information about anything that happens in the game.
With a traditional coach or pro player that teaches you, they’d have to rewatch your game with you, which already takes an hour and they have only a limited capacity along with their other commitments. For us, it takes a couple of seconds to have a holistic view of everything that happened in the game. We can analyse way more games and deeper insights.
When you watch some video tutorials they advise you “this flash is good, throw it here at this time.” We’re able to give you specific, situational advice on how you can improve, what about your game makes you a good player and where you can improve.
[primis_video widget=”5183″]
ESI: So the feedback is more geared towards a particular team set up or a specific round, and throwing the flash in this spot in this context would be the better option?
PS: Yeah so on one hand regarding flashes, when we do our analysis of utilities generally one thing we would tell you is “ok your flashes are suboptimal, here’s how you throw them” and there would be a map with everything you threw. Then it’ll be compared to the positioning of your opponents, compared to the hundreds of thousands of matches we’ve analysed to know where people generally position themselves so they can clearly see.
That’s just one example, but we have extremely personalised information. When giving generic advice you know, “you need to defend this passage or this area.” But when you play you can either stand in obvious areas or you can play off certain angles to surprise your opponents.
We can measure what the optimal idea is, and if you position yourself in the wrong positions and you end up losing we can look at where you are, your problems with positioning and so on. We can suggest training maps, look at heat maps and then learn from that.
ESI: Do you get the level of detail down to peeking or jiggle peeking, is it that granular? Or is that a bit harder to nail down?
PS: We can have this data but we’ve decided against it for now as we feel it would be a tad overwhelming for players. If your problem is the exact second of peeking then your positioning, tactical responsiveness isn’t really the issue for you.
You would be better offer looking at information regarding your utilities and how you interact with your teammates in the game. We decided going too deep into detail wouldn’t provide much value for users, but it’s something we’re measuring as well.
RELATED: How Athlane intends to simplify esports sponsorships
ESI: How has your partnership with TDJ Pitango allowed you to upscale through to your beta launch this month?
PS: They’re our investor and a part of Pitango Venture Capital which is one of the biggest Israeli funds. Being technical people, CS:GO players, programmers and so on we didn’t have a lot of business know-how. Having someone like them behind us and giving us business advice, doing workshops with us on how to run a startup has been really valuable.
Without them, we wouldn’t have been able to get to where we are right now. At this stage, we are already looking for another round of investment and hopefully, there will be even more partners like them joining up with us.
ESI: I know it’s only been a few days, but have you had the uptake you anticipated since the beta launch?
PS: It’s been good so far, we’re quite surprised by how many people visited the website and how many registered, it’s going very well. Obviously, there were a lot of bugs and programming hiccups that we figured out immediately after people poured into the website. But that’s the idea behind beta testing, you find those immediately.
The beta test that we’re doing right now is quite early stage, we’ve only launched a couple of features. We want to check if everything is working well and we can serve as many people as we want.
Now every few days there’ll be new parts and functionalities of the website and platform being turned on. Hopefully, those that have already registered will continue enjoying what we’re bringing in and more people will be happy to join and get value out of our product.
Has there been any commonality in feedback so far, some features people are responding to well?
PS: I’d say the most positive feedback we’ve seen is our ‘similarities view’, which is actually not the main feature of what we’re doing, but given that we have so many metrics describing pro players playstyle, we did this small algorithm that compares your play to pros. Then you’ll appear on a table compared to the top 10 players most similar to you.
People are having a lot of fun with that, even at the office. After every matchmaking or FACEIT game we’re playing we’re checking, “I’m similar to Kirby right now.” So that’s something people are really enjoying and more generally, showing your strong and weak sides in a visual form is something people are enjoying, even now with only some of our functionalities switched on.
After seeing how you go through the launch of this live beta, what are the next big steps for GGPredict? Is it just rolling out more features as you go, is it expanding to Dota 2 as has been mentioned in the past, or is it something else?
Right now we’re going to focus on CS:GO, we have so many features either ready or about to be ready and there’s so much to do and we’ve got so many ideas that we want to focus all our efforts on providing the best solution for CS:GO.
The moment we feel that we can offer the best you can get in Counter-Strike we’ll move onto our next game which will most likely be Dota 2, but we don’t want to box ourselves in yet.
And is there anything in particular that you’re looking for when it comes to new investment partners?
We already have some offers and very advanced talks with multiple investors, but we’re not rushing it. So we’re slowly going through the mid-meetings, talking with investors from Poland but mostly from overseas. Pitango is a great partner still and we’d like to have someone that not only brings money and investment but knowledge and contacts and so on.
Disclaimer: This is a sponsored piece by GGPredict