Group 2 Members:
- Dayna Karls
- Ryan Herrington
- Cody Nelson
- Nick Mitchinson
- Ethan Yu
Descriptions of Questions:
Given a player with a known height and weight, what offensive position (QB, RB, WR, TE, or OL) are they most likely to play in the NFL?
This data and models will be helpful for deciding a new position for players that need a position change in the NFL. Position changes are common in the NFL in order to best use players’ skills effectively. With these models, coaches should be able to plug in a player’s height and weight and get a position recommendation for the player.
import pandas as pd
data = pd.read_csv('Basic_Stats.csv')
data.head()
| Age | Birth Place | Birthday | College | Current Status | Current Team | Experience | Height (inches) | High School | High School Location | Name | Number | Player Id | Position | Weight (lbs) | Years Played | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | NaN | Grand Rapids , MI | 5/23/1921 | Notre Dame | Retired | NaN | 3 Seasons | 71.0 | NaN | NaN | Evans, Fred | NaN | fredevans/2513736 | NaN | 185.0 | 1946 - 1948 |
| 1 | NaN | Dayton , OH | 12/21/1930 | Dayton | Retired | NaN | 1 Season | 70.0 | NaN | NaN | Raiff, Jim | NaN | jimraiff/2523700 | NaN | 235.0 | 1954 - 1954 |
| 2 | 56.0 | Temple , TX | 9/11/1960 | Louisiana Tech | Retired | NaN | 1 Season | 74.0 | NaN | NaN | Fowler, Bobby | NaN | bobbyfowler/2514295 | NaN | 230.0 | 1985 - 1985 |
| 3 | 30.0 | New Orleans , LA | 9/30/1986 | LSU | Retired | NaN | 5 Seasons | 73.0 | NaN | NaN | Johnson, Quinn | NaN | quinnjohnson/79593 | NaN | 255.0 | 2009 - 2013 |
| 4 | 25.0 | Detroit , MI | 3/31/1992 | Central Michigan | Active | Pittsburgh Steelers | 3rd season | 77.0 | Clintondale HS | Clinton Twp.,Macomb Co., MI | Walton, L.T. | 96.0 | l.t.walton/2552444 | DE | 305.0 | NaN |
Variables:
The columns we are interested in from this dataset are: Name, Player ID, Position, Height (inches), and Weight (lbs).
We will be finding the predicted position for each players, to test the effectiveness of models for choosing a position for players with a specific height and weight.
Methods:
We will first have to clean the data, specifically the position column since the dataset only has positions for current players. Since the data set is large, and we want positions to be accurate, it makes the most sense to drop the examples with missing position data.
Next, we will need to use one-hot encoding on the remaining position data to make columns for each individual offensive position.
Our model will need to be a classifier. The two best options for this would be k-NN and Decision Tree algorithms, so we will create a model for both and determine which performs better while tuning the size of k and tree depth.