Volume 8, Issue 1

From an almost standing start at the beginning of the 1990s, the number of statues of U.S. baseball and English soccer heroes has risen inexorably. By 1st September 2011, 33 soccer players and 67 Major League Baseball (MLB) players were, or were soon to be, depicted by existing or commissioned subject specific statues inside or adjacent to the stadia they once performed in. Yet even amongst the very finest exponents of their sport, relatively few players are honored in this way.This paper investigates and compares the defining characteristics of stadium statue subjects in these two national sports. We first developed a shortlist of potential causal factors likely to influence subject selection by considering the motivations behind statue building. The MLB Hall of Fame and the English Football League “100 Legends” list were then used as samples of the best performers from each sport. Logistic regression models were built to test the effects of potential predictors for the selection of statue subjects; these included loyalty, locality, longevity, performance of the player and their team, national recognition, sympathy and the effect of nostalgia or memory (i.e., the era a player performed in).The optimal models for soccer and baseball correctly identified depiction or non-depiction for 87% and 90.6% of the respective samples, and their significant constituent effects indicated the importance of club loyalty and era. Players who played most or all of their careers at one club or franchise and those active in the 1950s and 1960s were most likely to be depicted. This latter finding in particular suggests that the role of a statue as a nostalgia/heritage marketing object impacts upon subject choice, which is thus dependent in part on the “chance” effect of birth era. Distinct characteristics of each sport, such as baseball franchise relocation and international soccer success, were also found to have a significant effect upon the probability of depiction. Predicted probabilities were calculated for players with statues who were not Football League “Legends” or MLB Hall of Famers; these confirm the viability of the model outside of the elite performers it was constructed upon.

This paper uses the difference in seeding ranks to predict the outcome of March Madness games. It updates the Boulier-Stekler method by predicting the outcomes by rounds. We also use the consensus rankings obtained from individuals, systems and poll. We conclude that the consensus rankings were slightly better predictors in the early rounds but had the same limitations as the seedings in the later rounds.

The men’s NCAA basketball tournament is a popular sporting event often referred to as “March Madness.” Each year the NCAA committee not only selects but also seeds the tournament teams. Invariably there is much discussion about which teams were included and excluded as well as discussion about the seeding of the teams. In this paper, we propose an innovative heuristic measure of team success, and we investigate how well the NCAA committee seeding compares to the computer-based placements by Sagarin and the rating percentage index (RPI). For the 2011 tournament, the NCAA committee selection process performed better than those based solely on the computer methods in determining tournament success.

Significant work has gone into the development of team and individual statistics in the NBA; for example, the team measures of the “Four Factors.” Less work has been conducted using multivariate analyses of these metrics, including identifying possible new statistical techniques to analyze these data. In particular, this research examines the feasibility of using structural equation modeling (SEM) for multivariate analyses of NBA Four Factors data. SEM consists of both confirmatory factor analysis (CFA) and path modeling. Before SEM is employed, this research first examines the effects of offensive and defensive Four Factors in a linear regression model, expanding previous research and providing a baseline for the SEM. In doing so, the data show the importance of effective field goal percentage. Next, structural equation modeling is employed. The CFA finds that offensive Four Factors are indicators of a single latent factor, labeled “offensive quality.” The “defensive quality” latent factor is estimable when replacing opposing teams’ free throw rate with steals per possession. The SEM is extended to regress winning percentage on latent offensive and defensive quality as well as salary. Salary is an important and often overlooked part of multivariate models examining team statistics, but it is easily incorporated in SEM. The explained variance for the regression in the SEM is similar to that of the linear regression model and indicates the importance of both offensive and defensive quality, with offensive quality having a larger effect. Team salaries are related to offensive quality, but not defensive quality or winning. As such, a second structural equation model is proposed where the effect of salary on winning is mediated by its relationship with offensive and defensive quality. Since salary is related to offensive quality but not defensive quality, and offensive quality is more important to winning percentage, this suggests that money spent is done so for offensive performance and affects winning through the performance paid for. These results suggest potential team strategies, as well as the applicability of SEM to sports analytics, not only to NBA data, but to other sports data as well.

We present evidence, based on play-by-play data from all 6087 games from the 2006/07– 2009/10 seasons of the National Basketball Association (NBA), that basketball scoring is well described by a continuous-time anti-persistent random walk. The time intervals between successive scoring events follow an exponential distribution, with essentially no memory between different scoring intervals. By including the heterogeneity of team strengths, we build a detailed computational random-walk model that accounts for a variety of statistical properties of scoring in basketball games, such as the distribution of the score difference between game opponents, the fraction of game time that one team is in the lead, the number of lead changes in each game, and the season win/loss records of each team.

This study analyzes how the efficiency of basketball players depends on experience, player’s nationality and team type. Using data for players in the Turkish Basketball League (TBL) from 1997 to 2010, we first show that efficiency of players initially increase with experience, but after a certain level decrease with it. Second, foreign players are on average more talented than native players. Third, the experience–efficiency link is not valid for foreign players indicating that more experience does not affect efficiency. Fourth, efficiency of foreign players in top-teams is no different than that of foreign players in regular (non-top) teams; whereas, native players in top-teams are more efficient than native players in other teams. Finally, in the presence of foreign player quota, top-teams hire the maximum number of foreign players allowed, while financially constrained regular teams can afford less foreign players. Consequently, under foreign player quota the quality of native players and the extent to which a team can pay for additional foreign players determine the team success.

We examine the question of whether or not momentum exists in an NFL football game. The concept of momentum is often cited by coaches, players, commentators and fans as a major factor in determining the outcome of the game and, consequently, in-game decision making. To examine the existence of momentum, we analyze particular game situations tied to what we consider to be “momentum-changing plays” (MCPs). These MCPs include fourth down conversions/stops, turnovers and scores allowed. We hypothesize that evidence of positive (negative) momentum would be characterized by increases (decreases) in yards gained, higher (lower) probability of converting a first down and greater (lesser) likelihood of scoring after a positive (negative) MCP. Our data set includes all plays from the 2002 to 2007 NFL seasons. We limit our analysis to game situations where the outcome of the game is still in doubt by removing plays that occur when a team is facing an insurmountable score differential. We use a pairwise matching comparison where we control for the game situations of home/away team, field position, time of game and score differential. We find little evidence for the existence of momentum in these events. Our results are in line with previous papers that find little empirical evidence of momentum in sports. While our findings cannot conclusively disprove the existence of momentum in the NFL, they further support the argument that momentum should not be a guiding factor for in-game decision making.

The San Francisco Giants were crowned champions of Major League Baseball in 2010 after defeating the Texas Rangers in the World Series. The World Series matchup may have come as a surprise to many baseball fanatics; the Rangers ended the regular season with the worst record of any of the eight playoff teams, and the Giants ended with the fourth worst. Did these two teams simply catch fire at the right time? Or were they better than their regular season records showed? To answer these questions, the regular season statistics of individual players on each team were used to simulate the postseason. These simulations determined the probability with which each playoff team could have been expected to win the 2010 World Series.

Managers in soccer face a critical in-game decision concerning player substitutions. While past research has investigated the overall effect of various strategy changes during the course of a match, no studies have focused solely on the crucial element of timing. This paper uses the data mining technique of decision trees to develop a decision rule to guide managers on when to make each of their three substitutes in a match. The proposed decision rule demonstrates between 38% and 47% effectiveness when followed versus 17% and 24% effectiveness when not followed based on over 1200 observations collected from some of the world’s top professional leagues and competitions.

Despite the massive popularity of probabilistic (association) football forecasting models, and the relative simplicity of the outcome of such forecasts (they require only three probability values corresponding to home win, draw, and away win) there is no agreed scoring rule to determine their forecast accuracy. Moreover, the various scoring rules used for validation in previous studies are inadequate since they fail to recognise that football outcomes represent a ranked (ordinal) scale. This raises severe concerns about the validity of conclusions from previous studies. There is a well-established generic scoring rule, the Rank Probability Score (RPS), which has been missed by previous researchers, but which properly assesses football forecasting models.

This study sought to establish the most important factors affecting the service in high-level women’s volleyball and the relative weighting of such factors on this technical part of the sport. A total of 1300 services from eight matches played in two Final Fours of the Indesit European Champions League were analysed. The services were delivered by 58 players of 25 nationalities. Observation sheets and two video cameras located at both ends of the court were used. Service speed was measured by radar. The twelve variables studied enabled the service to be divided into four components. The most influential component (19.02% of total variance) comprised variables related to technical service characteristics (type of service and service speed). The second most influential component (15.16% of variance) was related to the opponents’ technique and tactics, and to their position on court at the time of the service. The service was also affected by the technical and tactical movements that the servicer needed to perform in the subsequent play (12.20%). The stage of the match and the score (10.67%) also presented players with different levels of risk and helped to determine the type of service chosen and the power with which it was executed.

We present formulas for calculating wins over replacement player (WORP) for NHL goaltenders. If a team's primary objective is victory over its opponent, then a player's value may best be determined by his ability to help his team win games. Furthermore, a player's contribution to his team's wins would be a universal unit of measure and would allow for direct comparisons of goalie and skater production. We found inspiration for our work in Keith Woolner’s calculations of “value over replacement player“ (VORP) for MLB players, and we will make use of Bill James’ “Pythagorean Method” as it was applied to the NHL by Cochran and Blackstock in 2009.

A team is backed into a 4th-and-26 from their own 25, down 3 points. What are the odds that drive ends in a field goal? In the 2003 playoffs, Donovan McNabb and the Eagles scoffed at such a probability as they converted and ultimately kicked a field goal to send the game into overtime. This study creates a mathematical model of a football drive that can calculate such probabilities, labeling down, distance, and yard line into states in an absorbing Markov chain. The Markov model provides a basic framework for evaluating play in football. With all the details of the model—absorption probabilities, expected time until absorption, expected points—we gain a much greater situational understanding for in-game analysis.

The importance of sports statistics to professional sports teams is clear and there are exciting new opportunities for statisticians to learn from the new types of data that are collected. The Journal of Quantitative Analysis of Sports has shown steady growth since its inception in 2005. The new editorial system and new directions and initiatives of the journal are described.