Artificial Intelligence continues to surprise us, as chess and Go are not the only games where AI-based programs has bested us. After AlphaGo’s huge success, now we have the AI poker bot player Libratus, designed by a pair of researchers at Carnegie Mellon University.
Poker is fundamentally different from chess or Go, requiring reasoning and intelligence which is difficult to imitate. An opponent’s hand remains hidden in this case, making it a taxing job to calculate the ideal strategy. The underlying AI software must use game theory to calculate optimal plays.
Libratus has marked history for card gaming, by playing thousands of games over a span of three weeks, and emerging victorious in all the games. This is just one instance where AI was used to play against a human. However, that’s not all. Companies are using machine learning, big data,and related technologies to enhance user engagement and gaming experience for the online card gaming industry.
Over 100,000 people subscribe to our newsletter.
See stories of Analytics and AI in your inbox.
Rise of online card gaming industry In India
Gaming has moved beyond physical tables to virtual, and the mobile gaming industry has benefitted the most from this shift. In India, reports suggest that the mobile gaming industry will reach $3 billion by 2019. Within this industry, online card games have stood out for some time, with growth rates of 100 percent every year.
However, the gaming industry is still in its early stages, especially Rummy. Rajinder Balaraman, Vice President, Matrix Partners India shared, “Rummy is a very India-centric game, which we have all grown up playing. The industry is poised for growth. It is still in very early stages but it is growing at 50-100% year on year.” Usually, these portals witness a user engagement in the range of 20,000 to two lakh players, besides an average of a lakh app installs, every week.
Most of these portals are available as both websites and mobile apps, running a single game or a host of casino games. For instance, Rummy Passion has three variants – Points rummy, Pool rummy and Deals rummy. The portal has over 22,000 registered users, and each user spends an average of 1.5 hours per day on the table.
Rummy Circle is another such game on the list, run by 1.5 hours per day. The portal has as added vertical, called Ultimate Games. This vertical contains freemium games like Teen Patti, Poker, and Rummy. The portal aggregates almost 15 million users, each spending an hour a day, on an average.
Big Data and Machine Learning ups user engagement
User engagement is improved either by rewarding the users, or ensuring a great user experience.
Play Games leverages big data technologies to provide a brilliant and engaging user experience. Bhavin Pandya, Co-founder, Play Games remarks, “We gather massive amounts of data on player behavior. All this helps to understand our player’s needs and wants, and enables us to build experiences that our players just fall in love with.” In other words, the firm deploys big data technologies, machine learning, and sophisticated predictive analytics algorithms to anticipate players’ needs, besides offering unparalleled and highly personalized gameplay experiences.
Rummy Passion serves as another case in point. The vibrant game tables in the portal ensure users’ return, while the loyalty rewards program provides additional benefits for free. “We go all out to enhance player experience and engagement. We also have the fastest NEFT withdrawals. And of course, customer support in multiple languages is appreciated by our customers,” comments Bobby Garg, Founder, Rummy Passion.
The success of Poker games generally depends on competitiveness of a tournament and the cash prizes involved. The Spartan Poker has a line-up of fantastic tournaments and activities throughout the year, guaranteeing attractive prizes and super value cash games. Amin Rozani, MD and Co-founder, The Spartan Poker says, “We also offer a complete user-friendly interface experience, ensuring a wholesome experience for the players.”
As mentioned earlier, Rummy games enjoy both traction and popularity in India, and these games usually offer several cash prizes. Most of these games deal in real cash, however, users of apps like Ultimate Rummy simply make points. The app earns from in-app purchases, through their freemium model.
At the end of the day, user engagement help drive revenue, for which you have to depend on machine learning and related technologies. Talking about making money, Rummy Passion had witnessed one crore rupees in money wagered on its platform.
How are these technologies being integrated with the online games?
There are several initiatives undertaken globally at integrating online card games with machine learning and the related technologies. CPRG Homepage, a research group is involved with a project surrounding use of computers to play poker against. Another research group, Bayesian Poker Player applied Bayesian approach for computational poker.
Estimation of hand strength: The idea behind this step entails completion of hands by sampling for inaccessible cards, and counting the wins, to estimate the probability of winning. The approach uses Monte Carlo sampling based algorithms to estimate the winning potential of player’s’ hand as well as the opponent’s.
Moreover, sampling is faster technique for computing winning probability, in comparison with par exact computation. Additionally, parametric estimation leveraging historical data can find several machine learning use cases.
Opponent modeling: In this case, players’ historic data is utilized for estimating the probability for available actions (fold, call, raise) for each opponent. One successful approach is using Neural Network, which takes numerous factors like player count, position, game type, etc. into account. This is one of the most efficient ways of performing opponent modeling.
Decision making and Risk management: The third approach entails coming up with utility functions and listing/rating strategies. Machine Learning’s role for this approach is very significant, and strategies can be scored based on historical or current data.
There are few key approaches which have been successfully applied so far.
- Probabilistic approaches (Bayesian networks)
- Rule based (event, action pairs)
- Function based (neural networks)
- Genetic algorithms
Indian startups are playing their cards well
While Indian startups in the e-commerce and service-aggregation segments struggling to ink a path to profitability, a startup from a narrow segment of online card games is setting an example. The firm is showing all other startups how to draw massive returns for its investors. Matrix Partners India successfully exited online rummy portal Ace2Three, making a staggering 20-fold return. This decision surfaced as Clairvest Group picked up a major stake in the portal for about Rs 474 crore.
Recently, investors of online Rummy and Poker portal Adda52 also exited with a return of 22 times. Delta acquired the card gaming portal for Rs 155 crore in September 2016. It’s obvious that these portals are successfully making using of machine learning and the related technologies to attract huge investor interest.