📚 Upside Studies: (1) Soccer Study: Perfect Shot Reveal (2) NBA Study: The Effect of Basketball Analytics Investment on NBA Team Performance (3) NCAA Study: Get Your Head in the Game

⚽ Upside Study: “Perfect Shot Reveal – Machine Learning Analysis of Goal-Scoring Strategies in Soccer”

Authors: Maha Yazbeck, Mohammad Abdullah, Sura Alhanouti, Emaly Vatne, Joshua Hagen, Theodore T. Allen, Samantha Krening
Published in: International Journal of Sports Science & Coaching (2025)

---

🔍 Introduction

As machine learning (ML) and data science continue to reshape sports analysis, soccer remains a prime arena for innovation—particularly in how we understand shot selection and scoring probability. However, most prior research has been heavily male-dominated, leaving a critical gap in the analysis of women's game dynamics.

This study takes a pioneering approach by analyzing shot data from the 2022 FIFA Men’s World Cup and the 2019 FIFA Women’s World Cup using ML algorithms. The goal was twofold:

1. Build accurate models to predict whether a shot results in a goal.

2. Determine the most influential features for shot success, comparing male and female players to identify gender-specific goal-scoring strategies.

By doing so, the study not only enhances tactical understanding but also offers data-driven training recommendations tailored by gender, directly benefiting coaching, scouting, and performance optimization.

You can download the full PDF study by clicking on the link below:

---

📊 Key Stats

• Total Events Analyzed:

  • 461,845 (Men’s WC 2022)

  • 443,304 (Women’s WC 2019)

• Total Shot Events Used in Models:

  • 3,200 (Men)

  • 3,248 (Women)

• Number of Final Input Features for ML Models: 16

• Model Accuracy (Best Model – Random Forest):

  • Men: 92%

  • Women: 86%

• Shot Outcome Distribution:

  • Goals made up ~12% of all shots, necessitating oversampling techniques to balance model training data.

---

⚙️ Methodology Overview

The authors used a comprehensive machine learning pipeline:

• Data Source: StatsBomb optical tracking and event data, with high-resolution spatial and temporal features.

• Shot Outcome Classification: Converted into binary form — “Goal” vs. “No Goal.”

• Modeling Techniques:

  • Random Forest (best performer)

  • Decision Tree

  • Neural Network

  • Logistic Regression

• Validation:

  • 5-fold cross-validation

  • External test dataset from the FA Women’s Super League (2020–21)

• Feature Engineering:

  • Field divided into 100 zones for location analysis

  • Game time split into 15-minute intervals

  • Applied Recursive Feature Elimination (RFE) to rank the importance of features

• Hyperparameter Tuning:

  • Used Design of Experiments (DOE) via JMP software to systematically tune model parameters such as tree depth, learning rate, number of estimators, and activation functions.

---

🎯 Top Predictive Features: Male vs. Female Players

🔑 Interpretation:

• For men, body mechanics and technical execution—where the player is and what body part they use—are critical to shot success.

• For women, the tactical context and spatial positioning—especially structured play and shot proximity—play a larger role.

• The shared top feature across both groups is “duration of pressure”, i.e., how long a player is pressured before shooting.

---

🧠 Insights for Coaches & Training Design

For Men’s Soccer:

• Training should simulate high-pressure scenarios and include restrictions based on the body part used (e.g., requiring headers or weak-foot shots).

• Emphasize decision-making under pressure, as “duration” was the top predictor.

For Women’s Soccer:

• Emphasize structured offensive setups like corner and free kicks.

• Target zones near the 6-yard box—shots from this area had the highest success rate.

• Limit shots to specific tactical patterns, encouraging play progression that replicates successful shot setups.

Universal Application:

• Coaches can use this data to visualize goal heat maps, guide players into high-probability zones, and minimize inefficient shot selections.

• The study reinforces the idea that gender-specific strategies and training are not only valid but vital.

---

🧪 Scientific Contributions

• Demonstrated how Recursive Feature Elimination and Design of Experiments can boost model accuracy while maintaining interpretability.

• Validated ML models across multiple datasets, showing robustness and transferability.

• Bridged a significant research gap in women’s sports analytics, advocating for equity in data-driven insights and tactical support.

---

📌 Conclusion

This study presents a landmark in applying machine learning to elite soccer performance, providing gender-specific insights on shot strategy and success. While male shot outcomes hinge more on position and mechanics, female outcomes are influenced by play pattern and spatial location.

By showing how pressure, play design, and player attributes affect shot success, the study offers a powerful tool for for analysts, coaches, and sports scientists to design more informed, objective training environments.

➕ Future directions include:

• Incorporating digital twin simulations to model spin, velocity, and aerodynamics.

• Exploring player-specific traits like dominant foot and fatigue.

• Extending the models to other leagues and tournaments.

---

🏀 Upside Study: The Effect of Basketball Analytics Investment on NBA Team Performance

Authors: Henry Wang, Arnab Sarker, and Anette Hosoi (MIT)
Published in: Journal of Sports Economics, 2025

---

🔎 Introduction

Over the past two decades, data analytics has become a strategic cornerstone in professional sports. The National Basketball Association (NBA) is no exception. Teams now collect vast volumes of data on player movement, in-game actions, biometric feedback, and more. Despite the hype and anecdotal success stories—such as the NBA’s embrace of tracking technologies or the enduring cultural impact of Moneyball—there remains a critical gap in the academic literature: Does investment in analytics actually help NBA teams win more games?

While sports executives, media, and fans often assume that analytics drive smarter decisions and better outcomes, there is surprisingly limited quantitative research proving a clear link between analytics investment and team success in professional basketball. Recognizing this gap, the authors set out to determine whether there is a statistically significant and causal relationship between analytics staffing levels and regular season wins in the NBA.

This research is especially timely as franchises continue to ramp up analytics spending—often hiring teams of data scientists, analysts, and engineers—despite little public evidence that such investments directly correlate with better on-court results. The authors argue that a more rigorous econometric analysis, controlling for numerous confounding variables, is needed to evaluate the true return on investment (ROI) of basketball analytics.

You can download the full PDF study by clicking on the link below:

---

🎯 Research Objective and Scope

The main research question is simple but powerful:
Does a larger basketball analytics staff lead to better performance outcomes in the NBA?

The performance outcome under study is regular season win total, widely considered a key performance indicator (KPI) for teams. The authors use a panel dataset spanning 12 seasons (from 2009–2010 to 2023–2024, excluding shortened or incomplete seasons) for all 30 NBA teams, producing a dataset of 360 team-season observations.

Rather than using indirect or subjective measures of analytics adoption (e.g., ESPN rankings), the authors employ a quantitative and novel proxy: the number of analytics staff per team per season, as recorded by NBAStuffer.com using public directories, LinkedIn, and media guides. This headcount serves as a surrogate for analytics investment, under the assumption that more analysts reflect greater adoption of data-driven practices.

---

⚙️ Methodology

To estimate the effect of analytics staff size on team wins, the authors employ a two-way fixed effects regression model, which accounts for:

• Team-specific effects (e.g., franchise culture, market size, long-term strategy)

• Time-specific effects (e.g., league-wide rule changes, COVID-19 disruptions, CBA renegotiations)

They also control for a rich set of time-varying covariates, such as:

• Roster Salary (inflation-adjusted)

• Roster Experience (average years in the league)

• Coach Experience

• Roster Continuity (returning players’ minutes played)

• New Coach indicator (binary)

• Player-Games Injured

• Number of Road Back-to-Back games (a proxy for travel fatigue and schedule difficulty)

This comprehensive control structure helps isolate the unique effect of analytics staff size on wins, removing biases from confounding factors that also influence performance.

The study also experiments with a nonlinear model using a logit transformation of win percentage to assess diminishing returns and runs multiple robustness checks, including simulations to test assumptions about the interdependence of team results in a zero-sum league.

---

📊 Key Findings

1. Each additional analyst adds approximately 1.25 wins per season

The two-way fixed effects model reveals a positive and statistically significant effect of analytics staff size on regular season wins. Specifically, hiring one additional analyst is associated with a gain of 1.25 wins, even after controlling for injuries, salary, coaching experience, and other key variables.

2. Analytics is a cost-effective performance lever

When compared to roster salary as a means to increase wins, analytics emerges as a more efficient investment. The study estimates that one additional win costs approximately $9.3 million in roster salary. By contrast, hiring an analyst (with significantly lower cost) can yield similar returns, making analytics an attractive complement or alternative to roster upgrades.

3. Diminishing marginal returns

The logit(pwin) model suggests that while the first few analytics hires have a substantial impact, the marginal benefit of each additional analyst decreases as team headcount increases. This is consistent with the idea that analytics output depends not only on staffing levels but also on organizational culture, decision-making integration, and diminishing capacity to act on new data insights.

4. Robustness and validity of results

The authors conduct a series of diagnostic tests and simulations to validate their assumptions. Even accounting for the zero-sum nature of win totals across the league, their models hold up, with residuals displaying low inter-team correlation and normality. This adds confidence to the study’s causal claims.

---

⚠️ Limitations

The authors acknowledge several limitations:

• Proxy limitations: Headcount is not a perfect measure of analytics investment. It doesn’t account for the quality, budget, software tools, or actual influence of analytics staff.

• Public data constraints: The staff numbers are based on publicly available information and may miss behind-the-scenes contributors or misclassify roles.

• Unobserved intangibles: Team morale, day-to-day strategy, locker room dynamics, and other "soft" factors cannot be measured or controlled for but undoubtedly affect performance.

• Linearity constraints: The model doesn’t account for more complex team interactions, such as how analytics influences injury prevention, talent identification, or in-game strategy—mechanisms that could mediate the effect on wins.

---

🧠 Implications for Practice

This study offers multiple practical takeaways for stakeholders in elite basketball and beyond:

For NBA Franchises:

• Investing in a robust analytics department can yield tangible competitive advantages, especially for teams with budget constraints.

• Analytics should not be seen merely as a support function—it is a strategic asset that contributes directly to win probability.

For League Policymakers:

• Given the demonstrated ROI, making advanced analytics tools more equitably accessible (e.g., through league partnerships or shared infrastructure) could enhance competitive balance.

For Other Sports and Business Sectors:

• The NBA case offers evidence that data-driven decision-making boosts performance, echoing findings in corporate settings where analytics investments improve productivity and profitability.

---

📚 Future Research Directions

The study opens the door for numerous follow-ups:

• Mechanism studies: Where does the impact occur? Draft strategy? In-game tactics? Injury prevention?

• Granular performance metrics: Beyond win totals—how do analytics influence offensive/defensive ratings, player efficiency, or lineup optimization?

• Cross-sport replication: Does the effect hold in the NFL, MLB, Premier League, or NCAA?

• Downstream impact: What about financial outcomes like revenue growth, fan engagement, or valuation increases?

• Organizational maturity models: How do internal analytics culture, leadership, and decision-making structures mediate outcomes?

---

✅ Conclusion

This study delivers the clearest empirical evidence to date that analytics investment positively impacts NBA team performance. In an era of increasing parity and financial constraints, the finding that hiring one more analyst equates to 1.25 more wins per season is significant—especially when a single win can be the difference between a playoff berth or elimination.

The authors convincingly argue that analytics is not just a back-office function, but a high-leverage performance driver, similar in strategic value to coaching, scouting, or player development. By validating analytics as a legitimate source of ROI, this study provides justification for teams to expand their analytics departments and for leagues to foster a more data-friendly competitive ecosystem.

In a broader sense, the findings contribute to the growing body of research demonstrating the power of evidence-based management, and underscore the importance of measuring what matters—not just collecting data, but translating it into better outcomes.

---

Upside Study: “Get Your Head in the Game” – A Review of Factors that Impact Collegiate Student-Athlete Mental Health Using a Biopsychosocial-Structural Framework

Authors: Catherine E. Roberts 1 · Dolapo A. Oseni 2 · Bettina Bohle-Frankel 1 · Claudia L. Reardon 3
Published in Current Psychiatry Reports (2025)

---

Introduction

College athletics can foster a sense of purpose, identity, and connection. Yet for over 500,000 student-athletes across the NCAA, these benefits often coexist with mental health risks that are shaped by the intense and layered demands of college sport. This comprehensive review, authored by Catherine Roberts, Dolapo Oseni, Bettina Bohle-Frankel, and Claudia Reardon, explores the unique mental health challenges student-athletes face, using a biopsychosocial-structural framework to examine the interplay of personal, environmental, and systemic influences.

The goal is not only to spotlight individual vulnerabilities but also to explore how institutional, cultural, and policy-driven factors can reinforce or alleviate mental health stressors.

You can download the full PDF study by clicking on the link below:

---

Framework Overview: Biopsychosocial-Structural (BPSS)

The study expands the classic biopsychosocial model—which examines health through biological, psychological, and social dimensions—by adding “structural” factors. This is especially crucial in student-athletes, who exist within highly organized and rule-bound systems that shape everything from their schedules to their access to care.

---

Key Themes and Insights

🧠 Biological Factors

• Injuries and Recovery: Injuries are one of the clearest intersections between physical and mental health. They increase risk of depression, anxiety, disordered eating, and identity crises. Athletes who undergo multiple surgeries or chronic pain are particularly vulnerable.

• Concussions and Anxiety: Athletes often underreport concussions due to fear of missing playtime or losing scholarships. Meanwhile, media attention on CTE (chronic traumatic encephalopathy) has heightened anxiety even in mild head injuries.

• Disorders and Risk Patterns:

  • Anxiety Disorders: Athletes in aesthetic or judged sports (e.g., gymnastics, diving) are more prone to anxiety due to lack of perceived control.

  • Depression: Seen at similar rates as non-athletes, but higher in individual sport athletes, especially track and field.

  • Eating Disorders: More common in athletes in weight-sensitive or aesthetic sports. Transition periods (injury or retirement) increase risk.

  • Substance Use: Student-athletes report less frequent but more binge-focused drinking. Nicotine pouches (e.g., Zyn) are increasingly popular, especially in football.

  • ADHD: May be more prevalent in athletes, particularly those who thrive in high-movement, reactive positions but struggle with structured playbooks or learning routines.

  • Sleep Deprivation: NCAA athletes average <6 hours of sleep/night during the season. Early practice times, travel, social media, and performance anxiety all contribute.

---

🧠 Psychological Factors

• Athletic Identity and Role Conflict: Many athletes have identified as athletes since childhood. College often deepens this identity while adding a conflicting “student” identity, creating internal tension and stress.

• Performance Pressure from Family: Families often invest emotionally and financially in an athlete’s success, leading to unspoken or explicit pressure to perform.

• Retirement and Loss: Only 2% of NCAA athletes go pro. The psychological fallout of “what’s next?” can be destabilizing, particularly for those forced into early retirement due to injury. These athletes show higher rates of depression, anxiety, and substance misuse.

• Stigma and “Toughness Culture”:

  • Cultural narratives of toughness, resilience, and playing through pain discourage many athletes—especially men—from seeking help.

  • Athletes often fear that disclosing mental health concerns could result in loss of playing time or coach disapproval.

---

🧠 Social Factors

• Campus and Team Dynamics:

  • Athletes often experience social siloing from the general student body due to demanding travel and training schedules, exclusive housing, and “athletic clusters” in certain majors.

  • Within teams, competition for playing time, cliques, and internal rivalries can create a tense social climate.

• Stereotypes and Discrimination:

  • Student-athletes, especially Black male athletes, are often reduced to the “dumb jock” stereotype, limiting faculty expectations and peer integration.

• Social Media and NIL:

  • NIL rights and social media have created “student-athlete influencers,” but also opened athletes up to online harassment, particularly during high-visibility events like March Madness.

  • The NCAA’s 2024 pilot study found 16 types of online abuse; female athletes were disproportionately targeted with sexist and sexually abusive comments.

---

🧠 Structural Factors

• NIL and Direct Pay:

  • The monetization of college sports has benefits (reducing financial stress, empowering athletes) but can increase comparison, identity strain, and new social stressors, especially when some teammates are paid and others are not.

• Sexism and Gender Inequity:

  • Despite Title IX, women still receive fewer resources, worse practice times, inferior facilities, and less media coverage.

  • Only 27% of D1 head coaches and 16% of ADs are women, despite women making up 47% of student-athletes.

• Racism and Representation Gaps:

  • In D1 football and basketball, Black athletes generate much of the revenue, yet coaching and administrative leadership remains predominantly white.

  • Black male student-athletes also have lower graduation success rates than white peers (e.g., 80% vs. 93% in football).

• LGBTQ+ Discrimination:

  • LGBTQ+ student-athletes often feel pressure to hide their identities to protect reputations.

  • Transgender athletes face medical barriers and policy ambiguity, as few healthcare professionals are trained to meet their needs.

• Logistical Demands and Time Constraints:

  • The NCAA’s 20-hour rule excludes mandatory travel and "optional" practices. In reality, many athletes exceed 40+ hours/week, which limits sleep, class attendance, and access to care.

  • The #1 mental health concern for D2/D3 athletes in NCAA’s wellness survey was feeling overwhelmed by everything they had to do.

---

Key Statistics

• Mental health diagnoses among NCAA athletes: Comparable to non-athlete peers.

• Top stressor (per NCAA): “Feeling overwhelmed” ranked highest across divisions.

• Suicide: Second leading cause of death among NCAA athletes; highest risk in male cross-country athletes.

• Graduation gaps: D1 white student-athletes (94%) vs. Black student-athletes (81%).

• Female student-athletes: Underrepresented in leadership and more frequently targeted in online abuse.

---

Conclusions and Recommendations

The mental health challenges faced by collegiate athletes are not just psychological or personal—they are deeply embedded in social, structural, and institutional systems. A singular focus on resilience or individual grit risks ignoring these broader forces. This study urges practitioners, athletic departments, and healthcare teams to:

1. Implement systemic reforms in policy, scheduling, and support services.

2. Reduce stigma and increase help-seeking through visibility, peer support, and culturally sensitive programming.

3. Expand mental health care beyond crisis management to ongoing, preventative support.

4. Train providers in addressing the unique needs of marginalized athletes, including those who are Black, female, LGBTQ+, or economically disadvantaged.

5. Ensure equity in resources, leadership, and representation across all levels of collegiate athletics.

---

You may also like: