ADS Capstone Chronicles Revised

13

models considered.

Along with the performance metrics, we can utilize the feature coefficients to understand the impact of each feature on our score predictions. The following formulas are used to calculate the score using our linear regression models. score away = 2.926 + 1.454 ℎ − 0.813 − 1.936 (1) score home = 2.887 + 1.623 ℎ − 0.901 − 1.055 (2) From the coefficients, we can see the number of yards gained via the pass has a greater impact on the scoring a team manages versus the impact from rushing yards per attempt. Additionally, each sack costs the offense, on average, just under a point of scoring.

5.2 Evaluation of Regression Results The linear regression models both yield similar results, which is to be expected, as the features chosen do not strongly correlate with location. For the away team model, our training set exhibits a mean squared error (MSE) of 58.04 with a mean absolute error (MAE) of 6.09, as compared to 56.24 and 5.94, respectively, for the testing set. WithanR 2 value near 0.43, our model moderately captures the variance in scoring using the features. For the home team model, our training set exhibits a mean squared error (MSE) of 61.67 with a mean absolute error (MAE) of 6.24, as compared to 63.38 and 6.32, respectively, for the testing set. Using the R 2 value, we can expect our model to capture near 40% of the variance in scoring. essential. Figure 8 Correlations Between Features and Scoring

feature relationships with the binary outcome are turnovers for each team, both positive and negative relationships. These findings are consistent with much of the existing literature, which also suggests logistic regression tends to perform well in similar contexts. It is worth noting our analysis differs from some previous

6 Discussion Our logistic regression model demonstrates robust performance, with high accuracy on both training and testing sets, indicating its effectiveness in predicting the binary outcome without significant overfitting. The top absolute

40