KartRider Rush+ is a mobile game developed by Nexon where players can compete and race each other. There are two main mode in the game: Speed race and Item race. Also there are solo mode and team mode. In this project, we will only talk about team mode, which is a team consists of 4 players.

Speed race vs Item race

In speed race, skill is all you need to win. There's no luck involved, so you can win even with a cheap kart if you're skilled enough. Whether in item race, there's some luck involved, and your kart plays important role. There's a huge difference when you use one kart and another, because each kart has unique abilities.

Average players vs Top players

In the game, there is a noticeable difference between average players and top players: win rate (%). Top players are those people who have >75% win rate. They only lose to another top players and extremely lucky opponent.

Knowing the game mechanism, I raise a question: If both use the same karts and builds, what's the difference between average players and top players?

Both use same kart, both have same luck. But how can top players win so easily? There maybe several aspect that's important in the game, such as: reflex speed, item management, communication, etc.

But I realize something. When I play against top players, they tend to move faster than my team. They always stay at the top position, and they catch up quickly once they fall behind. Therefore, I have one hypothesis about what's the most important thing that differs average players and top players: Position.

In this project, I will analyze whether there's a significant difference in position between average players and top players.

The data comes from videos format. It's a full screen gameplay, which includes position list at the top left corner.

For the pro gameplay, I downloaded the video directly from their facebook stream. For the average players, I acquire it by asking my friends in instagram.

To extract the data, there are two methods that come into my mind.

I use digit recognizer package, recognize digit and record them every ~0.8 second.