Partially constrained group variable selection to adjust for complementary unit performance in American college...

Area of Science:

• Sports Analytics
• Statistical Modeling
• American College Football

Background:

• Accurate American college football (CFB) team rankings are crucial for title and playoff implications.
• Existing metrics often evaluate offensive and defensive units independently, neglecting their synergistic potential.
• The concept of 'complementary football,' where units influence each other (e.g., defense creating scoring opportunities), remains largely unquantified.

Purpose of the Study:

• To identify key features of complementary football through a data-driven approach.
• To adjust offensive and defensive performance metrics based on the complementary unit's impact.
• To enhance the accuracy of CFB team performance evaluations.

Main Methods:

• Utilized data from 2009-2019 American college football seasons.
• Incorporated natural splines with group penalty methods for statistical analysis.
• Employed partially constrained optimization to adjust for strength of schedule and home-field advantage.

Main Results:

• Identified consistently influential features of complementary football.
• Developed a method to adjust team offensive and defensive performance scores.
• Provided a more holistic evaluation of team strength beyond independent unit analysis.

Conclusions:

• Complementary football is a significant, quantifiable factor in American college football performance.
• The proposed methodology offers a more accurate and nuanced approach to ranking CFB teams.
• Future evaluations should incorporate these complementary dynamics for improved predictive power and rankings.