Machine learning and econometrics meet virtual economy management
Cambridge, Massachusetts, United States United States
  • Overview
  • Team
Company description
Built by Harvard, MIT, Yale, and Berkeley researchers, SavvyStat is a predictive analytics platform utilizing cutting edge machine learning and econometrics for virtual economy management. We are starting with tackling the game industry. With our simulation and parameter optimization tools, game developers can manage their virtual economy effortlessly. For example, we extract and predict a sign of currency inflation in the game real-time and recommend necessary actions (i.e., item pricing, supply control) to stabilize the economy for sustainable revenue and player retention maximization. 
We aspire to play the ministry of finance role in the future when the fast-growing virtual ecosystem reaches its maturity (where everyone has a significant online persona). We are starting with the game industry for two main reasons (1) games capture the types of data that allow us to realistically model virtual-version of econometrics (i.e., people's purchasing behavior, risk-taking behavior, etc.) with a very high few-seconds temporal resolution + (2) the industry is already big and mature (~$100B). Once we reach the trajectory of conquering the game industry, we strongly believe that our models are going to be applicable to other industries as well, such as eCommerce. 

Business model

SavvyStat is a SaaS product offering subscription based services to game developers. There are three tiers to pricing - categorized in terms of monthly active users (MAU). The three tiers are Basic (free up to 10K MAU), Pro ($0.03 per user between 10K and 100K MAU), and Enterprise ($0.02 per user for 100K+ MAU). Before subscribing to SavvyStat, clients are offered a month-long pilot period, during which we only charge a discounted flat fee. We are entering the market by first pursuing pilot projects with beta clients worldwide. We are targeting established game publishers as our beta clients (those who manage multiple games), in an effort scale quickly (after a successful pilot project with one game, we seamlessly tap into other games under them) and build a brand name that others can trust; game industry is very tight-knit, and the word-of-mouth effect is tremendously strong. 

Competitive advantage
Although we are really the first to offer virtual economy management in gaming industry, conventional data visualization/analytics companies have been seen as our competitors when talking to potential beta clients. However, these companies are mostly limited to visualizing very basic economic indicators (i.e., revenue). Some of them do bring predictive elements to measuring future revenue, but they handle the data without the context of econometrics and only rely on generalized linear models (which inherently lead to inaccurate prediction when dealing with complex data). Compared to them, our analytics engine is superior. We are driven by research-backed econometrics and machine learning expertise, incorporating non-linear, time-series, model stabilization, and hyper parameter optimization to our approach. These collectively result in state-of-the-art economy simulation. We verified our competitive advantage by receiving verbal approvals on pursuing pilot projects with major game publishers (including one of the biggest in Asia) - contract writing in the process.