⚽ Upside Study (1): Evening Smartphone Exposure Impairs Sleep Quality and Next-Day Performance in Elite Soccer Players: A Randomized Controlled Trial
Authors: Nadia Dridi, Mohamed Abdelkader Souissi, Rim Dridi, et al.
Published in: Biology of Sport, 2026
🔎 Introduction
Sleep is vital for athletic recovery and for optimizing both physical and psychological performance. However, elite athletes are often susceptible to poor sleep quality and habitually short sleep duration. One significant external factor contributing to declining sleep is the use of light-emitting electronic devices, such as smartphones, in the hours before bedtime. Exposure to this blue light can disrupt the human circadian clock, acutely suppressing melatonin, the sleep-facilitating hormone, which can lead to a reduction in overall sleep quality and quantity.
While the acute effects of electronic device use on sleep have been studied, the cumulative impact of consecutive nights of use on an athlete’s next-day cognitive and physical performance, particularly with regard to time-of-day variations, has been largely unexplored in soccer players.
You can download the full PDF study by clicking on the link below:
🎯 Research Objective and Scope
The study aimed to examine the effects of using a smartphone for two hours prior to bedtime for five consecutive nights on young elite soccer players:
To examine the impact on sleep quality.
To assess its effect on both cognitive and physical performance.
To investigate potential time-of-day variations in these effects.
The researchers hypothesized that smartphone use would impair sleep quality (increasing sleep latency and reducing total sleep duration), lead to declines in cognitive performance (attention and reaction time), and reduce physical performance (reactive agility and explosive strength), with effects more pronounced in the morning due to cumulative sleep deprivation.
⚙️ Methodology
This was a randomized controlled crossover trial involving 16 male elite-level soccer players (mean age 19.75±1.0 years) from a professional Tunisian Premier League club.
Conditions: Participants were assigned to two trials separated by a one-week washout period:
Experimental (Electronic Device, ED): Participants used a tablet (with quantified blue light exposure) to watch standardized, neutral entertainment videos for two hours (8:00 PM to 10:00 PM) before bedtime for five consecutive nights.
Control (C): Participants read neutral printed magazines or relaxed while avoiding sleep for two hours (8:00 PM to 10:00 PM) before bedtime for five consecutive nights.
Measurements: Sleep quality was assessed following nights 1, 3, and 5. Cognitive and physical performance tests were conducted on the mornings (7:00 AM-8:30 AM) and afternoons (5:00 PM-6:30 PM) after days 1, 3, and 5.
Sleep Measures (Subjective): Sleep Onset Latency (SOL), Total Sleep Time (TST), Sleep Efficiency (SE), subjective sleep quality (SSQ), Spiegel Score, and Epworth Sleepiness Scale (ESS).
Cognitive Tests: Simple Reaction Time (SRT), Choice Reaction Time (CRT), Trail Making Test (TMT), and Digit Cancellation Test (D-CAT, a measure of attention).
Physical Tests: Squat Jump (SJ), Countermovement Jump with Arm Swing (CMJA), and Reactive Agility Test (RAT).
Controls: The study controlled for lighting, room temperature, noise levels, standardized dinner time, and eliminated caffeine/alcohol intake. Training loads were also standardized across all participants.
📊 Key Findings
Sleep Quality Impairment
Five nights of pre-bedtime smartphone use significantly impaired sleep compared to the control group.
Increased Sleepiness: Subjective sleepiness, measured by the ESS, was significantly higher in the electronic device (ED) group after day 3 and day 5 compared to the control group.
Reduced Duration and Efficiency: The ED condition significantly shortened Total Sleep Time (TST) and reduced Sleep Efficiency (SE).
Delayed Onset: Sleep Onset Latency (SOL)—the time it takes to fall asleep—was significantly increased after nights 1, 3, and 5 in the ED group compared to control.
Performance Decline and Time-of-Day Effects
The impairments worsened cumulatively over the five nights and were stronger in afternoon sessions. The typical time-of-day (TOD) advantage, where performance is often better in the afternoon (observed in the control group), disappeared or even reversed following five nights of smartphone use in the ED group.
Cognitive Performance: After five days of smartphone use, there was a significant deterioration in Choice Reaction Time (CRT) and Simple Reaction Time (SRT) in the afternoon sessions compared to both baseline and the control group. This resulted in enhanced morning performance compared to the afternoon after day 5, a reversal of the pattern seen in the control condition.
Physical Performance: After five consecutive nights of smartphone use, physical performance tests showed a decreased jumping ability (SJ, CMJA) and slower Reactive Agility Test (RAT) times. This decline was particularly pronounced in the afternoon sessions.
⚠️ Limitations
The study had several methodological limitations:
Small Sample Size: The sample size (n=16) limits statistical power and the generalizability of the findings.
Subjective Measures: The reliance on subjective sleep assessments (diaries, ESS) lacks the precision of objective measures like polysomnography or actigraphy.
Mechanism Ambiguity: The lack of physiological marker assessment (e.g., melatonin, cortisol) limits the ability to confirm the hypothesized underlying biological mechanisms (melatonin suppression).
Limited Generalizability: The sample consisted exclusively of elite male soccer players, restricting the generalization of the results to female athletes or other sports/age groups.
🧠 Implications for Practice
The findings emphasize the negative impact of prolonged nighttime blue light exposure. The practical implications for athletes and practitioners include:
Implement Device-Free Periods: Athletes should implement device-free periods prior to bedtime (e.g., minimizing screen exposure) or activate blue light filters to minimize circadian disruption.
Adjust Training Schedules: Given the cumulative performance decline and the disappearance of the typical afternoon advantage, training and competition schedules may need to be adjusted. Technical and high-intensity sessions may be more effective during morning hours following evening screen exposure.
Promote Sleep Hygiene: Coaches should educate athletes on the risks of nighttime screen use and promote optimal sleep hygiene practices for recovery and performance.
✅ Conclusion
The study provides clear evidence that the cumulative effect of five consecutive nights of evening smartphone use significantly impairs sleep quality (reduced TST, increased SOL, increased sleepiness) and negatively impacts both cognitive and physical performance in elite soccer players. Crucially, the effects were stronger in the afternoon, suggesting an interaction with the athlete’s natural circadian rhythm. This underscores the need for strategic management of electronic device use to sustain peak athletic performance.
🏀 Upside Study (2): Optimizing Offensive Gameplan in the NBA with Machine Learning
Author: Eamon Mukhopadhyay, Purdue University 1Published in: Unpublished (Conference or journal not specified, paper dated 2024 based on internal references) 2
🔎 Introduction
The National Basketball Association (NBA) has undergone an analytical revolution over the last two decades, leading to the development of numerous specific metrics beyond general box score quantities. Pioneering work, such as Dean Oliver’s Basketball on Paper (2004), guided the shift toward using metrics alongside the “eye-test” for decision-making4. While the main offensive goal is to score points on a possession, the question arises: what truly drives a team’s Offensive Rating (ORTG), a metric developed by Oliver that accounts for possessions?
Traditional metric evaluation uses individual analysis 6, but an alternative is to model existing metrics using machine learning techniques to gauge the combined effectiveness of a unique set of features, particularly those related to a team’s offensive game plan. This study aims to use machine learning to determine the specifics of game plan execution that maximize a highly functioning offense.
You can download the full PDF study by clicking on the link below:
🎯 Research Objective and Scope
The paper had two main goals:
To develop multiple hypotheses for optimizing team game plan by finding the best-fit machine learning models and, if accurate, displaying hypothetical testing data to maximize a team’s ORTG9999999.
To evaluate the effectiveness of using machine learning models to determine what can build an existing metric (like ORTG) using advanced features (offensive play types), rather than simply creating a new one10101010.
The dependent variable modeled was Offensive Rating (ORTG)11111111. The independent features were derived exclusively from non-box-score, playtype-based statistics12.
⚙️ Methodology
The study used cumulative team playtype statistics from every NBA team from the 2015–2016 to the 2022–2023 NBA seasons, yielding 240 total data points13131313. All data was sourced from the official NBA website and Synergy14.
Dependent Variable: Offensive Rating (ORTG), normalized from 0 to 115151515.
Independent Features: The initial set included 48 features, consisting of 8 play types multiplied by 6 internal features16.
Playtypes Used (8): Isolation, Transition, Pick and Roll Ball Handler, Pick and Roll Roll Man, Post Up, Spot Up, Cut, and Off Screen17.
Internal Features (6): FREQ% (Frequency Percentage), FG% (Field Goal Percentage), FT% (Free Throw Frequency Percentage), TOV% (Turnover Percentage), FREQ% (again), and SCORE FREQ%18181818.
Excluded Playtypes: Putbacks, Misc (Miscellaneous), and Handoff, as they do not typically represent a specific offensive scheme19.
Feature Reduction: Principal Component Analysis (PCA) was used to reduce the feature count to 18 to avoid underfitting and improve accuracy.
Machine Learning Models: Two regression models were tested:
Simple Linear Regression
A supervised Multi-Layer Perceptron Regressor (MLPR) Neural Network
Validation: Leave-one-out cross-validation was used to build multiple model iterations, with the overall performance judged by Root Mean Square Error (RMSE) and Coefficient of Determination ($\text{R}^2$)23232323.
📊 Key Findings
Model Performance
Both models showed a strong correlation with Offensive Rating, but the Neural Network performed slightly better2424242424.
Neural Network (MLPR):
$\text{R}^2$ (Coefficient of Determination): 0.69425.
RMSE: $\approx0.107$ (or 2.16 ORTG points)26.
The optimal network had 1 hidden layer with a size of 327.
Simple Linear Regression:
$\text{R}^2$: $\approx0.665$2828.
RMSE: $\approx0.112$ (or 2.26 ORTG points)29.
Game Plan Optimization Hypotheses
Based on the more effective neural network model, the following play types were encouraged to optimize ORTG30303030:
Heavy Isolation Emphasis: One-on-one (isolation) plays should be heavily emphasized, ideally accounting for nearly 20-25% of play frequency31. This forces the defense to remain honest and can generate kick-out opportunities for spot-up shooting32.
High-Efficiency Spot-Up Shooting: Teams should maximize their spot-up plays with both a high-efficiency (Field Goal Percentage of 40–42% or higher) and a high-rate (25–28% frequency is optimal)33. The model noted spot-up shooting as an essential play type to score on34.
Utilize Transition Opportunities: Transition opportunities should be used to generate offensive firepower, ideally at 17–20% of general plays35. Transition was also noted as a best scoring method36.
Efficient, Non-Aggressive Pick-and-Roll: The game plan should use a general rate of around 15% frequency for pick-and-roll possessions finished by the ball-handler and roll-man37. The emphasis should be on ball efficiency (low turnovers) and forcing defensive help to create additional spot-up opportunities. The model suggested a non-aggressive scoring approach to the pick-and-roll may yield higher offensive output.
⚠️ Limitations
The author noted several limitations in the study:
Dataset Size: The size of the dataset (240 cumulative team-season statistics) combined with the number of features may raise questions, although PCA was used to mitigate this40.
Model Scope: The study did not test higher, more advanced machine learning systems (like Bayesian methods or polynomial regression) despite the initial models having notable correlations41.
Exceptions Exist: The hypotheses represent general offensive trends, and exceptions exist, such as the Golden State Warriors’ unique playstyle or the 2022-23 Sacramento Kings’ record ORTG achieved with a limited use of isolation plays42424242.
Context Missing: The model does not measure the true impact of the schemes on other factors like defensive intensity or offensive discipline43. Also, it delegates the in-depth analysis of whether a team’s players are sufficient to meet the game plan’s goals to other papers44.
✅ Conclusion
The study successfully demonstrated that a machine learning model, specifically a Multi-Layer Perceptron Regressor (MLPR) neural network, can effectively model a well-known statistic like NBA Offensive Rating (ORTG) using advanced playtype features. The derived model showed a strong correlation and was used to mathematically visualize an optimal hypothetical game plan. The four game plan hypotheses—emphasizing high-frequency, high-efficiency isolation, spot-up shooting, and transition plays, along with efficient, secondary use of the pick-and-roll—can guide coaches and executives toward a successful offensive scheme. The paper advocates for the widespread use of this machine learning method in sports analytics to determine what builds an existing metric, rather than relying solely on simple metric analytics.
🏀 Upside Study (3): Investigation of Force Plate Jump Testing Metrics Relevant to Return to Play Decision Making in Basketball Athletes After Anterior Cruciate Ligament Reconstruction
Authors: Christopher S Hart, Elizabeth S Chumanov
Published in: International Journal of Sports Physical Therapy (IJSPT), 2025
🔎 Introduction
Basketball is a high-risk sport for lower limb injuries, with knee injuries accounting for 16.3% of total injuries in adolescent competitive players. The ACL injury incidence rate is 0.16 per 1,000 athlete-exposures for collegiate or semi-professional players. Following Anterior Cruciate Ligament Reconstruction (ACLR), roughly one in three young athletes suffer a second ACL injury upon returning to sport, despite following recommended timelines and achieving satisfactory performance benchmarks.
Force plate vertical jump testing is used in the clinical setting to identify persistent lower limb asymmetries that can remain even after an athlete has been cleared on conventional strength and horizontal hop tests. However, there is no consensus on which vertical jump task, or which of the hundreds of available metric combinations, is most relevant for basketball athletes post-ACLR.
You can download the full PDF study by clicking on the link below:
🎯 Research Objective and Scope
The purpose of this retrospective cohort study was to examine the jump performances of male high school and collegiate basketball athletes in the final stages of ACLR rehabilitation to:
Establish referenceable between-limb symmetry scores.
Identify metrics that best detect between-limb asymmetry during common vertical jump tasks.
⚙️ Methodology
Study Design: Retrospective cohort examination.
Participants: 49 male high school and collegiate basketball athletes who underwent primary ACLR. Participants were in the final stages of rehabilitation and intended to return to competitive basketball.
Grouping: Participants were divided into two groups based on their isokinetic quadriceps peak torque Limb Symmetry Index (LSI):
Quad LSI <90% (n=24).
Quad LSI >90% (n=25).
Testing: Athletes completed testing using in-ground, dual force plates (sampling at 1,000Hz). The tests included:
Double-Leg Countermovement Jump (DL-CMJ).
Single-Leg Countermovement Jump (SL-CMJ).
Single-Leg Repeat Hop (SL-RH).
Analysis: Independent samples t-tests were used to compare the LSI of various jump metrics between the two quadriceps strength groups. The LSI was calculated as (Surgical Limb/Nonsurgical Limb) * 100.
📊 Key Findings
Asymmetry Detection by Task
Basketball athletes with quadriceps strength asymmetry (Quad LSI <90% ) demonstrated greater asymmetry in two of the three jump tasks:
Double-Leg Countermovement Jump (DL-CMJ): Significant differences in LSI were found between the groups.
Single-Leg Repeated Hop (SL-RH): Significant differences in LSI were found between the groups.
Single-Leg Countermovement Jump (SL-CMJ): No significant differences in LSI were found between the groups.
Key Metrics for Detecting Asymmetry
Metrics that showed statistically significant differences (p<0.05) between the Quad LSI <90% and Quad LSI >90% groups (with large effect sizes) included:
DL-CMJEccentric Rate of Force Development (RFD)−22.71.28Concentric Impulse−22.41.06Concentric Impulse 100ms−24.01.50Peak Take-Off Force−12.72.61SL-RHJump Height−12.40.81Flight Time:Contact Time13.30.84
🧠 Implications for Practice
Battery of Tests is Crucial: The inconsistency in findings—especially the lack of difference in the SL-CMJ test between groups—suggests that isokinetic testing (quadriceps strength) does not consistently align with jump test findings across all tasks. This highlights the importance of implementing a battery of jump tests to ensure the restoration of multiple physical qualities required for basketball.
DL-CMJ and SL-RH are More Sensitive: DL-CMJ and SL-RH metrics demonstrated the highest sensitivity for detecting between-limb asymmetry in this population when a quadriceps deficit was present.
Limitations of 90% LSI: The referenceable data shows large variability between jump tasks and metrics, which discourages relying on the 90% LSI benchmark as a single proxy for jump symmetry criteria. Clinicians are encouraged to reference the standard deviations and coefficient of variation for jump metrics as supplemental criteria.
📚 Future Research Directions
Future research should address limitations by:
Including allometric scoring (relative to body weight) to establish meaningful cutoffs.
Conducting longitudinal research to compare performance against uninjured control groups and to analyze the relationship between graft type and outcomes.
Including psychological readiness outcome measures.
Utilizing peak eccentric velocity to standardize the effort of the athlete during testing.
✅ Conclusion
The three studies presented offer critical, data-driven insights into optimizing performance and recovery in elite athletes across soccer and basketball, validating the necessity of an evidence-based approach in modern sports science and management.
⚾ Effectiveness of Single Leg Isometric Bridge and Nordic Hamstring Exercise Testing for Prediction of Hamstring Injury Risk in Professional Baseball Players
Authors: Austin R. Driggers, Andrew C. Fry, Kristen C. Chochrane-Snyman, John P. Wagle, and Jeffrey M. McBride
Published in: The American Journal of Sports Medicine, 2025
🔎 Introduction
Hamstring strain injury (HSI) imposes a significant burden on elite athletes, including those in professional baseball, leading to substantial financial costs for organizations1111. While various factors have been studied, muscular strength is the only modifiable risk factor with robust support for injury risk mitigation strategies2222.
Previous research has linked low eccentric hamstring strength, as measured by the Nordic Hamstring Exercise (NHE), to an increased risk of future HSI in other sports3333. However, the efficacy of strength testing remains inconsistent across studies, leading to a need for further investigation4444. Furthermore, no studies have examined the new modified Single Leg Isometric Bridge Test (SLIBT) as a predictor of HSI risk, nor have either the SLIBT or NHE been specifically studied in professional baseball players5555.
You can download the full PDF study by clicking on the link below:
🎯 Research Objective and Scope
The primary aims of this prospective cohort study were to determine:
Whether preseason hamstring strength and interlimb asymmetry measured using the modified SLIBT and the standard NHE are associated with future HSI in professional baseball players6666.
The relationship (correlation) between hamstring strength and interlimb asymmetry measures obtained from the SLIBT versus the NHE7777.
⚙️ Methodology
Study Design: Prospective cohort study; Level of evidence, 38888.
Participants: 465 male professional baseball players from 8 teams affiliated with a single Major League Baseball (MLB) organization999.
Data Collection: Preseason hamstring strength assessments were completed as part of routine athlete monitoring before the 2019, 2021, and 2022 seasons10. A total of 38 new HSI events were recorded in 36 players over 751 player-seasons11111111.
Testing Procedures (Force Measurement):
Nordic Hamstring Exercise (NHE): Participants performed the standard bilateral eccentric lowering motion on a NordBord apparatus, which independently measured the eccentric peak force of each leg12121212.
Modified Single Leg Isometric Bridge Test (SLIBT): Participants performed a single-leg isometric bridge on dual-force platforms (ForceDecks) with two modifications: a $20^\circ$ knee angle (to maximize biceps femoris activation) and a loaded barbell ($100 \text{kg}$) to secure the hips. Isometric peak force was measured for each leg13131313.
Statistical Analysis:
Absolute Strength Comparison: Injured limb strength was compared to the mean strength of the uninjured group14.
Relative Risk (RR): RR was calculated using ROC-determined strength cutoffs to assess HSI risk15151515.
Logistic Regression: Used to determine the odds ratio for HSI incidence with every $10\text{N}$ increase in strength16.
Correlation: Pearson $r$ was calculated for strength and asymmetry measures between the NHE and SLIBT171717171717.
📊 Key Findings
1. Low Hamstring Strength is a Risk Factor (NHE & SLIBT)
Lower hamstring strength levels in subsequently injured limbs were observed compared with the two-limb mean strength of healthy, uninjured players for both tests1818181818181818.
NHE Absolute Strength: Players with an NHE strength of $<377 \text{N}$ had a 2.5-fold higher risk of subsequent HSI (Relative Risk (RR), $2.49$; $P=.027$)191919191919.
Risk Reduction per 10-N Increase (Logistic Regression): An inverse relationship was found between strength and HSI risk2020.
SLIBT: Each $10\text{N}$ increase in peak force corresponded to a $7.4\%$ HSI risk reduction ($P=.019$)21212121.
NHE: Each $10\text{N}$ increase in peak force corresponded to a $6.2\%$ HSI risk reduction ($P=.019$)22222222.
2. Interlimb Asymmetry is NOT Predictive
Interlimb asymmetry measures from both the SLIBT and NHE were not indicative of subsequent HSI23232323232323. Additionally, no significant differences were found when comparing the strength of the subsequently injured limb versus the contralateral uninjured limb within the injured group24.
3. Correlation Between Tests is Moderate
Strength: Moderate significant correlations were observed for both absolute strength ($r=0.39$; $P<.001$) and relative strength ($r=0.33$; $P<.001$) between the SLIBT and the NHE25252525252525252525.
Asymmetry: No significant correlations were found between measures of interlimb asymmetry when assessed by the SLIBT versus the NHE26262626.
🧠 Implications for Practice
Screening and Intervention: Both the SLIBT and NHE can be valuable tools to identify professional baseball players at risk of future HSI due to low strength levels27272727.
Targeted Training: For players exhibiting low hamstring strength on either test, strength and preventative training prescriptions should specifically target improving these scores to decrease individual HSI risk, especially targeting low levels of eccentric strength (NHE)28.
Combined Testing: The moderate correlation and lack of correlation for asymmetry between the two tests reinforce that a multifaceted strength testing process offers the most value, as results from one test cannot be used to inform the results of the other2929292929292929.
⚠️ Limitations
Recurrent Injuries Excluded: Recurrent HSIs were excluded, limiting analysis of reinjury risk30.
Injury Severity: The grade of the HSI was not included in the study, focusing instead on time missed (player availability) rather than radiological severity31.
Specificity of Injury Mechanism: The difference in SLIBT (isometric) and NHE (eccentric) tests measures different actions, and the low agreement between them aligns with the general challenge of lower body strength testing32323232.
Bland-Altman Bias: Bland-Altman analysis indicated a bias toward higher strength values using the NHE compared with SLIBT scores (mean difference of $34.4 \text{N}$)33333333.
✅ Conclusion
The current investigation demonstrates that low levels of hamstring strength, as measured by both the SLIBT and the NHE, are significantly associated with an increased risk of future HSI in professional baseball players34343434. Crucially, an NHE peak force of $<377 \text{N}$ was linked to a 2.5-fold higher risk35353535. The findings support using both the SLIBT and NHE preseason to identify at-risk players and to inform mitigation strategies aimed at increasing strength to limit HSI rates and maximize time at play36363636. While strength levels were predictive, interlimb asymmetry was not, and the moderate correlation between tests indicates that they measure related but distinct aspects of hamstring strength37373737.
You may also like:






























