The NBA 2K series comprises basketball simulation games released annually by Visual Concepts and published by 2K Sports. Its games are available on multiple platforms, including PlayStation, Xbox, Nintendo Switch, and PC. The game allows players to simulate NBA basketball games, featuring realistic graphics, commentary, and player movements. (SGO Thermal, 2022).
In the game, every NBA player is assigned a set of ratings. These ratings reflect the player's abilities in various areas of the game, such as scoring, defense, rebounding, and playmaking. The scoring range can go as low as 0 to a high of 99 in a particular category. These ratings are determined by Visual Concepts based on analysis of player statistics and performance. Ratings are crucial to the gameplay; players with higher ratings will perform virtually better on court
If players feel unsatisfied with their ranking (like former NBA star John Wall who scored an average of 90 in the 2k18 edition), this brings a negative outlook to the game and the player.
Outside the game, player ratings generate a lot of discussion and debate among NBA fans and players. As NBA 2K is a popular video game that is played by millions of fans around the world, players recognize that their virtual presence in the game can have an impact on their real-life popularity and brand image. They want their virtual representation in the game to be as accurate and flattering as possible, which can translate to more endorsement opportunities and higher marketability.
As player rankings are subjectively decided by the team at Visual Concepts, such rankings can be an issue as players can be scored too high or too low, which the latter, could affect public perception of the player. If a player is scored high, it causes unwarranted hate towards the player.
Many factors like personal bias can contribute to a player’s ranking. Data Science and predictive modeling would be useful to see if the team at Visual Concepts are consistent and unbiased with their player scoring. With this in mind, if the Visual Concepts team is consistent with their ranking, then predictive modeling/machine learning could be used to effectively predict the new 2K game player ranking.
Efficient player ranking data is not only helpful in improving public perception, but can also serve as valuable data to players. This ranking is a way for them to gauge their virtual skills and compare them to their peers, which can be a motivating factor for them to improve their real-life performance.
We will use a converted CSV NBA 2K + player Stats CSV file from Github by benrallet.
The data has been derived from HoopsHype and Basket-ball Reference.
The following categories will be used:
import pandas as pd
pd_2k = pd.read_csv("ranking_2k.csv")
pd_2k.head()
| Rk | Player | Pos | Age | Tm | G | GS | MP | FG | FGA | ... | USG% | OWS | DWS | WS | WS/48 | OBPM | DBPM | BPM | VORP | Year | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | Quincy Acy | PF | 22 | TOR | 29 | 0 | 342 | 42 | 75 | ... | 14.7 | 0.7 | 0.4 | 1.1 | .157 | -1.0 | 1.0 | 0.1 | 0.2 | 2014 |
| 1 | 2 | Jeff Adrien | PF | 26 | CHA | 52 | 5 | 713 | 72 | 168 | ... | 15.6 | 0.5 | 0.4 | 1.0 | .064 | -2.1 | -0.9 | -3.0 | -0.2 | 2014 |
| 2 | 3 | Arron Afflalo | SF | 27 | ORL | 64 | 64 | 2307 | 397 | 905 | ... | 22.5 | 1.5 | 0.5 | 2.0 | .042 | -1.1 | -1.8 | -2.9 | -0.5 | 2014 |
| 3 | 4 | Josh Akognon | PG | 26 | DAL | 3 | 0 | 9 | 2 | 4 | ... | 20.3 | 0.0 | 0.0 | 0.0 | .196 | -3.1 | 0.1 | -3.0 | 0.0 | 2014 |
| 4 | 5 | Cole Aldrich | C | 24 | TOT | 45 | 0 | 388 | 44 | 80 | ... | 12.7 | 0.1 | 0.4 | 0.6 | .070 | -3.5 | 0.6 | -2.9 | -0.1 | 2014 |
5 rows × 51 columns
We'll cluster the data based on similar features (i.e. group 2/3 pt field goals together; positions together) to evaluate and predict ranking. Doing so allows us to discover if there are certain factors evaluators hold more importance to when rating a player.