Horsepower-The Bot That Calculates The Best Tech For eSports

Inspiration

Physical sports is dominated by statistics and statisticians. Nate Silver, the statistician who famously predicted the outcome of electoral districts in the 2008 U.S. presidential election, has written many pieces on baseball in a field known as sabermetrics. Football, also known as "soccer" to our American friends, is also dominated by statistics with many organisations attempting to bet everything from who will win what match to what player will score in a particular match; indeed, a simple search on Google during the current World Cup 2022 shows live probabilities based on who is likely to win. Other sports such as basketball, horse-riding and boxing are similarly dominated by statistics.

Despite the popularity of eSports, which is more popular than some physical sports even (cf. eSports for Dummies by Professor. Phill Alexander), there hasn't really been a widespread use of statistics as seen in traditional sports. There are many reasons for this. I speculate three: 1) eSports is relatively new compared to the traditional sports 2) Many eSports players do not come from an academic background or are too young to have formally completed tertiary education and the same likely applies to those who manage them and 3) eSports, whilst popular, hasn't yet gained the attention of mainstream media, meaning that the audience is largely confined to those within eSports and/or gaming, most of whom are young (by contrast, in football/soccer, everyone knows who Cristiano Ronaldo and David Beckham are, even if they aren't active watchers of football/soccer).

I come from an economics background although now work as a programmer. I have a casual interest in sports economics, especially after learning about it and how it played a part in the 2010 World Cup in South Africa (for soccer). I decided to use the knowledge I have of sports economics to see if it can be used within the context of eSports.

This bot hopes to make inroads in eSports by integrating statistics (no pun intended) into this form of sports. By including statistics, it is hoped that this can help eSports mature more as a field as well as assist eSports players become better at what they do using concrete quantifiable measures.

It is called "Horsepower" after the measurement used to gauge engine performance for cars. This bot similarly hopes to fine-tune an eSports player's performance. In keeping with the theme of the hackathon, "Wave Of One", the bot also has the capability to recommend other individuals to form teams; this feature is particularly helpful for those in team/multiplayer eSports.

What it does

This is a bot that eSports players can use to inquire about what equipment they can/should buy (e.g. whether to use mechanical vs membrane keyboard) for an eSports activity in order to increase their chances of success. The bot is powered by the Kore.ai platform, which allows people to create bots that can respond to human queries easily without a need to learn fields such as natural language processing (NLP). The bot uses a field of statistics known as hedonic regression which is a form of regression whereby an dependent variable (y) is deemed as being influenced by constituent characteristics represented by independent variables (x). It is commonly used in housing, whereby a model stipulates that the price of a house is determined by a range of variables such as size of windows, location, distance from public transport, etc. Here, eSports performance is seen as being determined by a number of independent variables such as internet speed, hours spent practising on a particular game, as well as whether a keyboard is mechanical or membrane. In some cases, some of the independent variables are a special kind known as "dummy variables", which mean they can only take on two categories (e.g. mechanical vs membrane keyboard).

How we built it

On a statistical level, in one perspective, success in eSports can be modeled as regressing a particular Elo Rating as a continuous dependent variable (y) and a range of variables such as internet speed, type of keyboard as well as practice hours, which are independent variables (x). Some of these independent variables are of a special type known as "dummy variables" because they can only take (or, at least, can be modeled as) taking one of two values - examples include male or female, whether a keyboard is mechanical or membrane, and so forth. One model put forward by myself is that eSports success can be modeled using the following equation:

Challenges we ran into

The short time constraint was sadly an issue so it meant the bot could not be as advanced as envisaged. As this was my first time creating a bot, I also ran into issues in getting the bot to understand different games such as League of Legends (LoL), Fortnite, Grand Theft Auto V, etc.

Accomplishments that we're proud of

I'm proud of learning more about eSports. I'm also proud to have learned of Team Liquid, as I never heard of the organisation until now. Surprisingly, they are based in The Netherlands, a country I almost moved to recently.

What we learned

I learned a lot through this hackathon. I learned about just how big eSports truly is and the thriving economy around it, together with highly committed eSports athletes. This was the first bot I built with the Kore.ai platform.

What's next for Horsepower-The Bot That Calculates The Best Tech For eSports

On a statistical level, this model can be refined further by using a section in linear regression known as simultaneous linear regressions. One problem with the model above is that it can be seen as naive in that it assumes that eSports players operate in a vacuum. In other words, that the independent variables that influence performance are entirely exogenous to the model above. However, in reality, it is possible that the model suffers from the simultaneity problem which is common in economics.

Built With

• kore.ai
• r