DOCSLIB.ORG

Barnabas Blake.Pdf (1.468Mb)

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Barnabas Blake.Pdf (1.468Mb) Cardiff School of Sport DISSERTATION ASSESSMENT PROFORMA: Empirical 1 Student name: Ba rnabas Blake Student ID: 20019089 Programme: SPE Dissertation title: Evaluating the validity of a basketball performance metric in correlation with winning and losing games Supervisor: Darrell Cobner Comments Section Title and Abstract (5%) Title to include: A concise indication of the research question/problem. Abstract to include: A concise summary of the empirical study undertaken. Introduction and literature review (25%) To include: outline of context (theoretical/conceptual/applied) for the question; analysis of findings of previous related research including gaps in the literature and relevant contributions; logical flow to, and clear presentation of the research problem/ question; an indication of any research expectations, (i.e., hypotheses if applicable). Methods and Research Design (15%) To include: details of the research design and justification for the methods applied; participant details; comprehensive replicable protocol. Results and Analysis (15%) 2 To include: description and justification of data treatment/ data analysis procedures; appropriate presentation of analysed data within text and in tables or figures; description of critical findings. Discussion and Conclusions (30%) 2 To include: collation of information and ideas and evaluation of those ideas relative to the extant literature/concept/theory and research question/problem; adoption of a personal position on the study by linking and combining different elements of the data reported; discussion of the real-life impact of your research findings for coaches and/or practitioners (i.e. practical implications); discussion of the limitations and a critical reflection of the approach/process adopted; and indication of potential improvements and future developments building on the study; and a conclusion which summarises the relationship between the research question and the major findings. Presentation (10%) To include: academic writing style; depth, scope and accuracy of referencing in the text and final reference list; clarity in organisation, formatting and visual presentation 1 This form should be used for both quantitative and qualitative dissertations. The descriptors associated with both quantitative and qualitative dissertations should be referred to by both students and markers. 2 There is scope within qualitative dissertations for the RESULTS and DISCUSSION sections to be presented as a combined section followed by an appropriate CONCLUSION. The mark distribution and criteria across these two sections should be aggregated in those circumstances. CARDIFF METROPOLITAN UNIVERSITY Prifysgol Fetropolitan Caerdydd CARDIFF SCHOOL OF SPORT DEGREE OF BACHELOR OF SCIENCE (HONOURS) SPORT AND PHYSICAL EDUCATION 2014-5 TITLE: Evaluating the validity of a basketball performance metric in correlation with winning and losing games DISCIPLINE: Dissertation submitted under the Performance Analysis area NAME: Barnabas Blake STUDENT NUMBER: 20019089 EVALUATING THE VALIDITY OF A BASKETBALL PERFORMANCE METRIC IN CORRELATION WITH WINNING AND LOSING GAMES Cardiff Metropolitan University Prifysgol Fetropolitan Caerdydd Certificate of student By submitting this document, I certify that the whole of this work is the result of my individual effort, that all quotations from books and journals have been acknowledged, and that the word count given below is a true and accurate record of the words contained (omitting contents pages, acknowledgements, indices, tables, figures, plates, reference list and appendices). I further certify that the work was either deemed to not need ethical approval or was entirely within the ethical approval granted under the code entered below. Ethical approval code: 14/5/28U Word count: 10,823 Name: Barnabas Blake Date: 19th March 2015 Certificate of Dissertation Supervisor responsible I am satisfied that this work is the result of the student’s own effort and was either deemed to not need ethical approval (as indicated by 'exempt' above) or was entirely within the ethical approval granted under the code entered above. I have received dissertation verification information from this student Name: Date: Notes: The University owns the right to reprint all or part of this document. TABLE OF CONTENTS ACKNOWLEDGEMENTS i ABSTRACT ii CHAPTER ONE: INTRODUCTION 1 1.1. Performance Analysis 2 1.2. Game-related Statistics 3 CHAPTER TWO: REVIEW OF LITERATURE 4 2.1. Performance Analysis in Basketball 5 2.2. Performance Indicators 5 2.3. Advanced Statistics 7 2.4. Game Pace 7 2.5. Performance Metrics 9 2.6. The Defensive Process 12 CHAPTER THREE: METHODS 14 3.1. Sample 15 3.2. Variables 19 3.2.1. Game-related Statistics 19 3.2.2. Advanced Statistics 20 3.2.3. Preventions 21 3.3. Data Collection 21 3.3.1. Hand Notation 21 3.3.2. Reliability 22 3.4. Data Analysis 23 CHAPTER FOUR: RESULTS 24 4.1. Descriptive Statistics 25 4.2. Tests of Significance 28 4.3. Correlations 33 CHAPTER FIVE: DISCUSSION 38 5.1. Discussion of Results 39 5.1.1. All Games 39 5.1.2. Winners and Losers 40 5.1.3. Close and Unbalanced 42 5.1.4. Correlations 42 5.2. Implications for Practice 44 5.2.1. Defending Shots 44 5.2.2. Defending Passes 45 5.2.3. Motivation 46 5.3. Reliability of the Data Source 46 5.4. Game Pace 47 CHAPTER SIX: CONCLUSION 48 6.1. Concluding Thoughts 49 6.2. Future Research 50 REFERENCES 52 APPENDICES 62 LIST OF TABLES Table 1a. Pool of close games (score differential: x < 5 points) 15 Table 1b. Pool of unbalanced games (score differential: 20 points < 16 x < 40 points) Table 2. Game categorisation thresholds 17 Table 3. Excluded outlying games (score differential: x > 40 18 points) Table 4. Game-related statistics recorded by the FIBA Live Stats 19 to produce the Olympic box scores Table 5. Common advanced statistics generated from box score 20 data Table 6. Kappa scores and ratings 22 Table 7. Descriptive statistics and percentage differences of box 26 score performance indicators Table 8. Descriptive statistics of defensive advanced statistics 27 and Preventions metric outputs Table 9. Statistical significance of the differences between 28 Winners (W) vs Losers (L) Table 10. Statistical significance of the differences between Close 29 Winners (CW) and Unbalanced Winners (UW) Table 11. Statistical significance of the differences between Close 30 Losers (CL) and Unbalanced Losers (UL) Table 12. Statistical significance of the differences between Close 31 Winners (CW) and Close Losers (CL) Table 13. Statistical significance of the differences between 32 Unbalanced Winners (UW) and Unbalanced Losers (UL) Table 14. Pearson’s correlation (Winners) 33 Table 15. Pearson’s correlation (Winners) 34 Table 16. Pearson’s correlation (Losers) 35 Table 17. Pearson’s correlation (Close) 36 Table 18. Pearson’s correlation (Unbalanced) 37 LIST OF FIGURES Figure 1. Descriptive statistics of box score performance 25 indicators Figure 2. Descriptive statistics of defensive advanced statistics 27 and Preventions metric outputs ACKNOWLEDGEMENTS Thanks to the International Basketball Federation (FIBA) and the International Olympic Committee (IOC) for their archived statistics for the 2012 Olympic Basketball Tournament for Women. Also, many thanks to the Cardiff School of Sport’s Centre for Performance Analysis for the footage that formed the sample for this study. Thanks to Alfredo Rodriguez of WammyRadio.com who provided the inspiration for this project with the invention of his Preventions metric. Finally, thanks to Darrell Cobner who provided clear guidance and support throughout the completion of this dissertation project. i ABSTRACT The aim of this dissertation project was to evaluate the validity of the novel Preventions metric, developed by Rodriguez (2013b), in correlation with performance indicators previously defined as discriminants of winning and losing games. The sample consisted of a selection of games from the 2012 Olympic Basketball Tournament for Women (n=18), categorised into two pools, close (n=9; x < 5 points) and unbalanced (n=9; 20 points < x < 40 points), based on final score-line differential. Box score data was scraped from the official FIBA Olympic archives to produce game-related statistics and advanced statistics; while Preventions metric data was recorded using hand notation. Descriptive statistics were generated to distinguish the differences in absolute values between the sample groups; Mann-Whitney U Tests were then performed to identify the significance of these differences. Winners and losers were differentiated by points prevented (p < 0.05), shots challenged (p < 0.05) and botched passes (p < 0.05). Pearson’s correlation data was produced to determine the strength of relationship between the new Preventions metric and academically proven advanced statistics. The variables correlating with the advanced statistics were points prevented (p < 0.01), shots challenged (p < 0.05) and botched passes (p < 0.05). The findings clearly identified points prevented, shots challenged and botched passes as significant variables to be considered as performance indictors due to their discriminant powers and correlation with peer-reviewed advanced statistics. This suggests that coaches should emphasise their defensive strategies regarding closing-out on shooters and disrupting passing lanes through on-ball pressure and off-ball denial defense. ii CHAPTER ONE INTRODUCTION 1 1. INTRODUCTION 1.1. Performance Analysis Performance analysis is a discipline that is vastly varied in its methods
Load more

Recommended publications
  • Rebound 18 10 26 #1
    October 26, 2018 EPA Docket Center U.S. EPA, Mail Code 28221T 1200 Pennsylvania Ave., NW Washington, DC 20460 Submitted via Regulations.gov Docket: EPA-HQ-OAR-2018-0283 Docket: NHTSA-2018-0067 Re: Joint Comments of Health, Environmental, and Conservation Groups on EPA’s Proposed Rule: The Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule for Model Years 2021–2026 Passenger Cars and Light Trucks, 83 Fed. Reg. 42,986 (Aug. 24, 2018) The Environmental Defense Fund and Union of Concerned Scientists (“Organizations”) hereby submit these joint comments in opposition to the Administrator’s proposal to roll back existing light-duty vehicle greenhouse gas emissions standards. See 83 Fed. Reg. 42,986 (Aug. 24, 2018) (“Rollback” or “Proposal”). These comments discuss certain of the Organizations’ objections to the analysis and application of the rebound effect in the Proposal. Some of the Organizations are also filing separate comment letters to provide more detail and to address additional issues. As described in detail in our comments (attached as Appendix A), the agencies’ analysis and application of the rebound effect is fatally flawed. The agencies fail to acknowledge or defend their changes in position. Specifically, they mischaracterize their own analysis of the appropriate rebound rate in their 2010 and 2012 final rules, and completely fail to mention the analyses in the 2016 Draft TAR and EPA’s 2016 Final TSD. The agencies also utilize unweighted, average 1 values of the studies they consider, contrary to prior acknowledgement that such a methodology is unreasonable and inadequate. Moreover, the agencies do not acknowledge that they previously considered 13 of the 16 studies listed in PRIA Table 8-8 and concluded that those studies supported the agencies’ previously adopted value for the rebound effect of 10%.
    [Show full text]
  • Predicting College Basketball Match Outcomes Using Machine Learning
    Predicting NCAAB match outcomes using ML techniques – some results and lessons learned Zifan Shi, Sruthi Moorthy, Albrecht Zimmermann⋆ KU Leuven, Belgium Abstract. Most existing work on predicting NCAAB matches has been developed in a statistical context. Trusting the capabilities of ML tech- niques, particularly classification learners, to uncover the importance of features and learn their relationships, we evaluated a number of different paradigms on this task. In this paper, we summarize our work, pointing out that attributes seem to be more important than models, and that there seems to be an upper limit to predictive quality. 1 Introduction Predicting the outcome of contests in organized sports can be attractive for a number of reasons such as betting on those outcomes, whether in organized sports betting or informally with colleagues and friends, or simply to stimulate conversations about who “should have won”. We would assume that this task is easier in professional leagues, such as Major League Baseball (MLB), the National Basketball Association (NBA), or the National Football Association (NFL), since there are only relatively few teams and their quality does not vary too widely. As an effect of this, match statistics should be meaningful early on since the competition is strong, and teams play the same opponents frequently. Additionally, professional leagues typically play more matches per team per season, e.g. 82 in the NBA or 162 in MLB, than in college or amateur leagues in which the sport in question is (supposed to be) only
    [Show full text]
  • Modeling Basketball Games As Alternating Renewal-Reward Processes and Predicting Match Outcomes a Thesis Submitted to the Department of Mathematics for Honors
    Modeling Basketball Games as Alternating Renewal-Reward Processes and Predicting Match Outcomes A thesis submitted to the Department of Mathematics for honors Conrad De Peuter April 22, 2013 Duke University, Durham NC Abstract We fit an Alternating Renewal-Reward Process model to basketball games and use the models to predict the outcomes of 1209 games in the 2012-2013 National Basketball Association (NBA) season. Using data collected from NBC play-by-play pages we fit various models for each team’s renewal process (time of possession) and reward process (points per possession). Using these estimated distributions we simulate the outcome of each of the 1209 games. We introduce four seperate models to predict these games. To evaulate our models we compared their predictions with that of other commonly used methods such as team record, Pythagorean Win Percentage, and Bookmaker odds. The research suggests that an Alternating Renewal-Reward Process is an appropriate fit to a basketball game. 1 Introduction Possession based statistics have been a cornerstone of most advanced statistical analysis of the National Basketball Association (NBA) in the past decade. Although possessions are not an official NBA statistic, their use is considered essential in advanced statistical analysis. A possession starts when a team gains control of the ball, and ends when the team loses control of the ball [9]. It is important to distinguish between possessions and plays. A play is an individual action to score, and a possession can have multiple plays. For example, a period in which a team shoots and misses, gets an o↵ensive rebound, then shoots and scores has two plays; however, it is only one possession since the opposing team never gained control of the ball.
    [Show full text]
  • Game #5 // Illinois (3-1) Vs. the Citadel (0-3*) *Plays Nov. 19 Nov
    2019-20 UNIVERSITY OF ILLINOIS FIGHTING ILLINI BASKETBALL Game #5 // Illinois (3-1) vs. The Citadel (0-3*) *plays Nov. 19 Nov. 20, 2019 // 8 p.m. CT // Champaign, Ill. // State Farm Center (15,544) TV: BTN – Kevin Kugler (Play-By-Play), Stephen Bardo (Analyst) Radio: Illini Sports Network – Brian Barnhart (Play-By-Play), Deon Thomas (Analyst) // Satellite Radio: Sirius/XM – 84, Streaming – 84 ILLINOIS 2019-20 SCHEDULE & RESULTS PROBABLE STARTERS (FROM THE LAST GAME) Pos. No. Name Ht. Wt. Yr. PPG RPG APG Note DATE OPPONENT TIME/RESULT TV/STREAM G 11 Ayo Dosunmu 6-5 185 So. 13.5 3.5 3.8 Wooden & Naismith Watch Lists NOVEMBER (3-1) G 10 Andres Feliz 6-2 195 Sr. 16.0 8.8 3.5 51.3% 2FG & 80.8% FT 5 NICHOLLS W, 78-70 (OT) BTN+ G 1 Trent Frazier 6-2 175 Jr. 9.8 2.5 1.8 Jerry West Watch List 8 at Grand Canyon W, 83-71 WCIX F 15 Giorgi Bezhanishvili 6-9 235 So. 9.0 5.0 1.3 50% FG & 57.1% 3FG 10 at #21 Arizona L, 69-90 PAC12 C 21 Kofi Cockburn 7-0 290 Fr. 14.3 11.5 0.5 3 double-doubles 18 HAWAII W, 66-53 ESPNU 20 THE CITADEL 8 p.m. BTN OFF THE BENCH 23 HAMPTON 7 p.m. BTN+ Pos. No. Name Ht. Wt. Yr. PPG RPG APG Note 26 LINDENWOOD 7 p.m. BTN+ G 0 Alan Griffin 6-5 195 So. 6.8 4.0 0.3 Leads team w/ 5 threes DECEMBER (0-0) F 2 Kipper Nichols 6-6 220 r-Sr.
    [Show full text]
  • 1 Una Revisión De Los Sistemas De Valoración De Jugadores De Baloncesto (I)
    Una revisión de los sistemas de valoración de jugadores de baloncesto (I). Descripción de los métodos existentes. Martínez, Jose A. Departamento de Economía de la Empresa. Universidad Politécnica de Cartagena. Facultad de Ciencias de la Empresa. Calle Real, 3. 30201. Cartagena. España. Tel.: +34 968 32 57 76; Fax: +34 968 32 70 81. E-mail: [email protected] Cómo citar este artículo: Este artículo es una versión del aceptado para publicación en la Revista Internacional de Derecho y Gestión del Deporte. Martínez, J. A. (2010). Una revisión de los sistemas de valoración de jugadores de baloncesto (I). Descripción de los métodos existentes. Revista Internacional de Derecho y Gestión del Deporte, 10, How to cite this paper: This article is a draft of the accepted paper to be published in the Revista Internacional de Derecho y Gestión del Deporte Martínez, J. A. (2010). Una revisión de los sistemas de valoración de jugadores de baloncesto (I). Descripción de los métodos existentes. Revista Internacional de Derecho y Gestión del Deporte, 10, 1 Una revisión de los sistemas de valoración de jugadores de baloncesto (I). Descripción de los métodos existentes. A review of the basketball player evaluation metrics (I) A description of the existing methods Resumen En este trabajo se realiza una completa revisión de los sistemas de valoración de jugadores de baloncesto que actualmente se utilizan por analistas en todo el mundo, con el fin de describir el gran conjunto de sistemas a disposición de los decisores en los equipos. Así, se pretende facilitar la toma de decisiones directivas racionales, en base a la disposición de un ingente volumen de información sobre la productividad de los jugadores.
    [Show full text]
  • Predicting NBA Player Performance Kevin Wheeler
    Predicting NBA Player Performance Kevin Wheeler Introduction Predicting the outcomes of sporting events and the performances of athletes represents a natural application for machine learning. Major, professional sports such as the NBA, NFL, and MLB contain a significant amount of easily accessible data whose outcomes and player performances tend to be randomly distributed and offer attractive data to predict. The NBA is particularly well suited for machine learning applications because the performance of NBA players across different positions can be measured using the same set of statistics. This is unlike the NFL, for instance, where different positions are measured using different metrics (e.g. QB’s are measured based on how many TDs they throw while linebackers are measured based on how many people they sack/tackle). Furthermore, applying machine learning algorithms to predict the performance of individual athletes also has a natural extension in predicting the outcomes of games. By predicting the scoring performances of each player and summing them up, it can be determined if predicting individual performances can be used an accurate win-loss classifier. In this project, I set out to develop a model using linear regression (with Naïve Bayes and SVM implementations to compare) to predict how many points NBA players would score against an opponent. Methods Feature Selection The ability to accurately predict the performance of NBA players depends upon determining and using features that are strongly correlated with the number of points an NBA player will score against any given opponent. A χ2 test was used to analyze a set of 46 team and individual, gave-level statistics to determine the subset of features to be used in the prediction models.
    [Show full text]
  • NBA Lineup Analysis on Clustered Player Tendencies: a New Approach to the Positions of Basketball & Modeling Lineup Efficiency of Soft Lineup Aggregates
    NBA Lineup Analysis on Clustered Player Tendencies: A new approach to the positions of basketball & modeling lineup efficiency of soft lineup aggregates Samuel Kalman, Jonathan Bosch Track: Basketball ID: 1548738 Abstract Basketball has recently been considered more of a “position-less” sport. Traditionally, the five positions of basketball are point guard, shooting guard, small forward, power forward, and center. However, most NBA players have skills, styles, and preferences that cannot be defined by a single traditional position. With ten seasons (2009-2018) of NBA player stats [1] that account for a player’s efficiency, opportunity, and tendencies, we were able to implement unsupervised machine learning techniques to create a framework for how NBA playing styles can be clustered on the court and used for strategic decision making when building rosters and creating lineup rotations. These player clusters can be considered new positions, and they give more accurate and detailed insight to what role a player possesses when on the court and how effective he could be in that role. Our unique methods give players a soft assignment for all clusters, that is, a probabilistic weighting onto each of the clusters indicating their likelihood of specific cluster fit. After analyzing the distribution of the various player stats within each new position, we were able to generate a player role for each cluster. Previous work [2,3] has also clustered NBA players into new positions. However, our methods incorporate different approaches of unsupervised machine learning, and offers an extension to the player clustering by additionally investigating the lineup efficiencies of different combinations of these new positions, playing on the basketball court together in collaboration.
    [Show full text]
  • A Starting Point for Analyzing Basketball Statistics
    Journal of Quantitative Analysis in Sports Volume 3, Issue 3 2007 Article 1 A Starting Point for Analyzing Basketball Statistics Justin Kubatko, The Ohio State University and basketball- reference.com Dean Oliver, Basketball on Paper and Denver Nuggets Kevin Pelton, Seattle Sonics & Storm Dan T. Rosenbaum, University of North Carolina at Greensboro and Cleveland Cavaliers Recommended Citation: Kubatko, Justin; Oliver, Dean; Pelton, Kevin; and Rosenbaum, Dan T. (2007) "A Starting Point for Analyzing Basketball Statistics," Journal of Quantitative Analysis in Sports: Vol. 3: Iss. 3, Article 1. DOI: 10.2202/1559-0410.1070 ©2007 American Statistical Association. All rights reserved. Brought to you by | University of Toronto-Ocul Authenticated Download Date | 3/2/16 11:03 PM A Starting Point for Analyzing Basketball Statistics Justin Kubatko, Dean Oliver, Kevin Pelton, and Dan T. Rosenbaum Abstract The quantitative analysis of sports is a growing branch of science and, in many ways one that has developed through non-academic and non-traditionally peer-reviewed work. The aim of this paper is to bring to a peer-reviewed journal the generally accepted basics of the analysis of basketball, thereby providing a common starting point for future research in basketball. The possession concept, in particular the concept of equal possessions for opponents in a game, is central to basketball analysis. Estimates of possessions have existed for approximately two decades, but the various formulas have sometimes created confusion. We hope that by showing how most previous formulas are special cases of our more general formulation, we shed light on the relationship between possessions and various statistics.
    [Show full text]
  • Basketball Reference Nba Pace by Year
    Basketball Reference Nba Pace By Year brontosauruses.tawdrily.Holotypic Single-entry Hal kids Aubrey shaggily and neoterizing eutrophic and monthly, Albrechther sasines she alwaysunderspending demonstrably, smear petrologically her she robe-de-chambre attack and it soli. jaunts agglomerate his Fep_object be efficient, and tarik black told silver screen and nail against regression techniques, by reference nba pace factor is an emotional return in what exactly constitutes a professional service synergy sports How will players perform this season? Steal rate estimates the percentage of opponent possessions that end with a steal by a player. TODO: we should review the class names and whatnot in use here. Despite this, player salaries have struggled to keep pace. Made some crucial adds using the smart tool and daily player projection tool. What does the community think of players? Offensive rebound percentage is an estimate of the percentage of available offensive rebounds a player grabbed while he was on the floor. Data: Share of population by weight category. Time leading scorers list. Rebound rate is a common stat found on any advanced stats site. NBA pace factor is an advanced basketball stat to help you bet smarter and make more money betting totals, spreads, and more. How do you become a referee for a math journal? The formula adds positive stats and subtracts negative ones through a statistical point value system. Of course, the downside is the loss of Anderson, to a division rival no less. Fique por dentro do calendário de NBA, por data e com toda a informação do início das partidas, cobertura na TV e mais.
    [Show full text]
  • Game 26 // #4Ap/4 Illinois (19-6, 15-4) at #7Ap/10 Ohio State (18-7, 12-7) March 6, 2021 // 3 P.M
    2020-21 UNIVERSITY OF ILLINOIS FIGHTING ILLINI BASKETBALL Game 26 // #4AP/4 Illinois (19-6, 15-4) at #7AP/10 Ohio State (18-7, 12-7) March 6, 2021 // 3 p.m. CT (4 p.m. ET) // Columbus, Ohio // Value City Arena TV: ESPN – Bob Wischusen (Play-by-Play), Dick Vitale (Analyst) Radio: Busey Bank Illini Sports Network – Brian Barnhart (Play-By-Play), Doug Altenberger (Analyst) // Sirius 83, XM 83 National Radio: Westwood One Sports – Chris Carrino (Play-By-Play), P.J. Carlesimo (Analyst) ILLINOIS 2020-21 SCHEDULE & RESULTS PROBABLE STARTERS (FROM THE LAST GAME) Pos. No. Name Ht. Wt. Yr. PPG RPG APG Note DATE OPPONENT TIME/RESULT TV G 1 Trent Frazier 6-2 175 Sr. 10.9 3.0 2.6 22 points Tuesday at Michigan NOVEMBER (3-0) G 44 Adam Miller 6-3 180 Fr. 8.4 2.7 0.7 18 points vs Nebraska (2/25) 25 NORTH CAROLINA A&T W, 122-60 BTN G 20 Da’Monte Williams 6-3 215 Sr. 5.5 5.5 1.9 53.8% 3FG (28-52) 26 CHICAGO STATE W, 97-38 BTN G/F 3 Jacob Grandison 6-6 205 r-Jr. 4.3 3.5 1.4 Illini 9-1 with him as starter 27 OHIO W, 77-75 BTN C 21 Kofi Cockburn 7-0 285 So. 17.5 9.8 0.1 15 double-doubles (2nd in NCAA) DECEMBER (4-3) 2 vs. #2 Baylor [1] L, 69-82 ESPN OFF THE BENCH 5 UT MARTIN canceled – Pos. No. Name Ht.
    [Show full text]
  • Liu MCS 100 Presentation
    Pairwise Comparison Models A Two-Tiered Approach for Predicting Wins and Losses in the NBA Motivation • Test Bradley Terry model as the basis for finding strong predictive models for NBA games • Test the success of an indirect, two-tiered approach to predicting wins • Only using Win/Loss record might not be optimal Hypothesis • Could be more effective to use a two- tiered approach • First identify the broad features that have high correlation with win rate • Model wins based off of those features • Predict for those features first and then predict wins Dean Oliver’s Four Factors • Effective Field Goal Percentage = (Field Goals Made + 0.5*Three Pointers Made)/Field Goals Attempted • Turnover Percentage = Turnovers/(Field Goals Attempted + 0.44*Free Throw Attempts + Tu r n ov e r s ) • Offensive Rebound Rate = Offensive Rebounds/(Offensive Rebounds + Opposition Defensive Rebounds) • Defensive Rebound Rate = Defensive Rebound Rate = Defensive Rebounds/(Opposition Offensive Rebounds + Defensive Rebounds) • Free Throw Factor = Free Throws Made/Field Goals Attempted Bradley Terry Application • The four factors are all rates • To calculate A’s turnover rate against B, we need 1. A’s mean turnover rate 2. The league’s mean turnover rate 3. The mean turnover rate of teams that play against B Why Bradley Terry? • Simple • Very little data required (only at the team level) • Far fewer features to predict Methodology • Data set 2010-11 NBA season • (82*30)/2 = 1230 observations • 861 in training set and 369 in test set (70%/30%) Models • Two predictive
    [Show full text]
  • The Rebound Effect in Transportation: Understanding the Important Implications for Climate Change
    The Rebound Effect in Transportation: Understanding the Important Implications for Climate Change Constantine Costa Samaras email: [email protected] This Briefing Asks Two Questions • What is the rebound effect in transportation? • What are implications for climate change? 2 This Briefing Asks Two Questions • What is the rebound effect in • Higher fuel economy or transportation? cheaper fuels reduces the fuel costs of traveling per • What are implications for mile – leading to an increase climate change? in miles traveled. Expressed as the elasticity of vehicle use with respect to either marginal costs or fuel economy 3 A Rebound Effect is Observed, but.. • “Research on the magnitude of the rebound effect in light-duty vehicle use dates to the early 1980s, and almost unanimously concludes that a statistically significant rebound effect occurs when vehicle fuel efficiency improves” (DOT, 2011) • Short-run to long run transportation rebound has been estimated at 3-15% but as incomes rise price sensitivity falls. It is also important to include response to fuel prices (Small and van Dender, 2007) • DOT review found that estimates have a mean of 23% with a range of 7-75%. Two thirds of the studies reviewed fell within 10-30% 4 Survey of Transportation Rebound Literature Source: DOT, 2011 5 6 A Rebound Effect is Observed, but.. • DOT uses a 10% rebound rate with sensitivity between 5-15% • Can we be certain that the elasticity is valid in both directions? • Data between states, cars and light trucks and on on-road fuel economy not straightforward
    [Show full text]