How Oleg Baranovskiy Created a Sports Prediction Chatbot

Bot Maker Interview by BotList

What’s your background, and what are you working on?

I have a master degree in electrical engineering, but last year at the University, I fell in love with machine learning and NLP. After University, I was working at a game development company, lead an NLP team at an SEO company, and later leading logistics/location team at DeliveryHero (online food ordering).

I’m passionate about leadership, machine learning and sports.

What motivated you to get started with Collbet?

It started in August 2010, when my wife and I (her background is very similar to mine 😁) had this discussion about if it was possible to predict football with crowd wisdom. I was pretty sure she was skeptical if it could work. So, I started building a website to collect data and prove my point.

What went into building the initial product?

In a few weeks, I built a Django site and very naive prediction model. I asked friends and friends of friends to contribute predictions on the site. At one point, I had around 20 random users who were loving football and willing to contribute to the site. After initial adoption, I started paying around 1 cent for contributions which attracted students. The audience was staying around 20–50 users.

How have you attracted users and grown Collbet?

A year and a half ago we decided to turn this home-project project into a business and advertised Collbet on Facebook. Very quickly we realized that we were advertising something that people never saw before. People viewed Collbet as a betting website, but we were really a prediction analysis website. So we came to the conclusion the best way to introduce Collbet to new users and send the right amount of information at a time was through a chatbot.

What have you done to retain customers?

Our main value is in the predictions we are making. So, if users become active with the chatbot on a winning day tend to stay around.

Traction also depends on the user's motivation. For example, the goal for users in Africa is to win money. African’s are very patient and willing to learn a new product while it’s opposite in UK. In the UK, many users are rude and demand more entertainment. We will satisfy both types of customers, but it will just take more experiments and fine tuning of the bot.

How have you thought about monetization?

We’ve implemented an extended version of the “Free to play” model. We encourage users to contribute their opinion/predictions on matches and reward them with our internal currency, “chips”. The more quality of contributions, the more chips. We then exchange chips for our predictions. If users are not willing to contribute or their quality contributions are low, they can buy chips.

We are also building out the subscriptions now.

What are your goals for the future?

Sports betting is very unfair to people who bet. The odds of winning aren’t in their favor, but with our prediction technology, we have the potential to make those odds better. We want to “shift the balance”.

What’s your advice for Bot Makers who are just starting out?

The chatbot market is very new and there is no single product which has been globally adopted. So, your product can be first of its kind.

Also, details matter!

I was surprised at how many users treat my bot as “a person”. So, the way you phrase your sentences will determine the retention. Chat is emotional. It can be very good and bad at the same time. The right phrase at the right time can make or break your bot’s conversational UX.

Where can we go to learn more about you?

We will present at TechCrunch Disrupt in San Fransisco on 5th — 7th of September.

🤖BotList 📺 YouTube 📘 Facebook 🌎 Collbet.com

Want to build your own chatbot business like Collbot?

You should join the BotList community! 🤖

We’re a few thousand founders creating the next wave of conversational and automation technology.