Volume 6, Issue 4

The aim of this paper is to develop a game-theoretic framework to study the impact of player personnel changes on offensive productivity in American football. We develop a new model, the improvement in passing model, which is used to determine the optimal proportion of run and pass plays that a team should call. The model determines the optimal run/pass balance in terms of parameters that reflect a recent change (generally, in our case, an increase) in efficacy of a team's passing offense. The model assumes a residual positive effect on the team's running game occurs as a result of the improved passing attack. Several conclusions are drawn, most surprisingly that improvements in a team's expected gains via the passing game imply that the team should, in fact, run more frequently to optimize their overall offensive productivity. We conclude with an example studying the 2009 acquisition of Jay Cutler by the Chicago Bears.

This study aimed to evaluate individual and team judo strengths using an analytic hierarchy process (AHP). The source of the decision was based only upon the international regulation values assigned to each of the decisive technique: 10 points for an ippon, 7 for a waza-ari and 5 for yuko(s). The data were obtained from Japan’s interscholastic athletic competition for men’s judo (high school teams) in 2005. Our AHP technique demonstrated an ability to predict judo strengths of individual players, and to detect potential players regardless of unsuccessful team results. There was a significant but small correlation between the actual team rankings and predicted team strengths. The predicted team strengths, however, may be more informative and appropriate to evaluate judo strengths than the tournament results, as they were based on the content of each individual match, rather than just win/lose results. Coaches or talent scouts should consider the estimated individual or team strengths as well as competition results to make more accurate decisions when selecting players or teams.

The purpose of this study was to examine the relationship between the opponent block, i.e. the blockers starting points and the number of blockers, compared to the hitter chosen by the setter to attack in elite male volleyball competition. 4531 sequences were analyzed, corresponding to 97 sets, of 12 national male teams participating in the 2007 World Cup. In order to test the association between variables, descriptive and inferential statistics the chi-square test with analysis of the adjusted residual were used. Intra- and inter-observer reliability was assessed via the percentage error method. Results showed that the pinched block “starting point” and the double block “final blocking action” were the most recurrent ones and the opposite player (right side) was the most requested hitter. This study showed a significant association between the blockers starting points and the hitter chosen by the setter to finalize the play set. There was a downward trend in the attack from the middle hitter with pinched block and an upward trend in the attack of the opposite (right side) with mixed block (pinched zone 2). A significant association between hitter and number of blockers was also verified: middle hitters show a high tendency to attack facing individual block, opposite players (right side) usually attack vs. double block and left-side hitters usually face double and triple block. These results suggest that blocker’s starting points are taken in consideration by the opponent setter to create the best conditions for the hitters. Hence, the setter takes advantages of the specific features (skills and ability to score) of its hitters and the space where they perform the attack in order to create difficulties to the opponent block (i.e. middle hitters has a high trend to attack with individual block, opposite player with double block and left-side hitters with double and triple block). The present study shows that the volleyball game at the elite male level is characterized by a constant adaptation between setter’s options and the opponent block tactics and strategies. These trends must be considered in the training process to prepare hitters and blockers for the most common situations occurring in the competition and, therefore, to create better and more appropriated and effective solutions to the game problems.

Football (soccer) is certainly the most popular sport in Belgium. In 2008-2009, the Belgian premier league (the so-called Jupiler Pro League) consisted of 18 teams, each having played others twice, home and away. Wins and draws earned 3 points and 1 point, respectively, and teams were first ranked by total points and then by total wins, regardless of the goal difference (or the number of goals scored), unlike in most European leagues. Obviously, the Belgian Champion should have been the team which had ended the season on top of the ranking, but a very unlikely event happened: the first two teams, Sporting Anderlecht and Standard Liege, completed the season with exactly the same number of points (77) and the same number of victories (24). Although the goal difference was in favor of Anderlecht, an extraordinary playoff match was hastily organized to decide the championship, and Standard ended up winning. That dramatic denouement sustained the discussions among the large football-lover part of the Belgian population and opens the interrogation: which team was really the best? The aim of this work is to objectively investigate this question through statistical modelling. Given the results of the whole season, a semiparametric model for the conditional probabilities of home win, tie and away win for each match, given the involved teams and other explanatory variables, is fit. The semiparametric nature of the model grants a great flexibility and allows us to identify interesting and up to now ignored patterns in the above probabilities. Then, a large number of Monte-Carlo simulations are run as if the season was replayed a large number of times, which permits the estimation of absolute probabilities of many events of possible interest. In particular, clear evidence about which team really deserved the 2008-2009 Belgian football champion title appears.

Although there is a modest amount of literature analyzing the career prospects of athletes in United States (U.S.) based football, basketball, and baseball leagues, little is known about the career prospects of professional soccer players in the U.S. In our analysis of Major League Soccer players, using a sample of approximately 1,100 players and 3,435 player-year observations, we identify the time-varying and time-constant variables that affect player career duration. Initial career expectancy was 2.4 years, more than a quarter of all players exit the league after one year, and the risks of career exit never fall below 12%. In addition, relative to U.S.-born players, foreign-born players face a disproportionately high exit rate.

We treat the number of years spent in major league baseball as a random variable and estimate probability distributions for this random variable through the use of recursive formulae. Distributional characteristics, including major league baseball worklife expectancies, are estimated for players by age and current activity status in the major leagues. Data from a recent time period (1977-2007) are used to calculate current characteristics of time spent in major league baseball. However, the contemporaneous nature of our data leads to censoring because many players in our data set had not completed their major league careers by the end of 2007. We deal with censoring through a Markov process model that captures transitions between activity and inactivity in major league baseball.

There is considerable interest in two consecutive game streak records in baseball, namely, the celebrated 56-game hitting streak of Joe DiMaggio and the less famous 84-games-reaching-base streak of Ted Williams, and how likely these records would be predicted to occur if the history of Major League Baseball were repeated. I strive to answer this question through simulated replication using a series of Bernoulli-type models. I assume that the number of games played by each player in each season is held constant while the batting and on-base averages are estimated from and shrunk towards the career trends of each player to smooth over outlying seasons. These simulation models are then verified against streaks that might be expected to occur, such as all-time streaks ranked 6 through 30, and are allowed to vary over time to reflect the changing distribution of opposing pitching. I find that a validated model for predicting hitting streaks contains no “hot hand” effect and suggests that the variability of opposing pitching has decreased markedly in the past 140 years. I also find that under this model, the DiMaggio streak can be considered exceptional while validated models for on-base streaks require considerably more complexity, including but not limited to a term that dampens on-base streaks.

An important aspect of pitching is the progression through the count of balls and strikes. Using data from the Pitch F/X database, we examine pitch-by-pitch data for twenty starting pitchers of “elite” and “non-elite” types. We see that the current count has a significant effect on the quality of the batting event. Elite pitchers are more likely to move from two-strike counts to strikeouts, and these pitchers are more likely to take advantage of eventual value of the plate appearance for particular counts. We examine how the pitch value of fastballs depends on the location of the pitch in the strike zone. A better understanding of count tendencies should lead to better measures of pitching performance.

At young ages, a few extra months of development can make a big difference in size, strength, and athletic ability. A child who turns 5 years old in January will be nearly 20% older by the time a child born in December has their 5th birthday. In many sports, including hockey, children born in the early months of the calendar year get noticed by their coaches because of the superiority they demonstrate due to their age advantage. As a result, boys born early in the year are more likely to reach the professional ranks of the National Hockey League (NHL). The phenomenon just described has been labeled the relative age effect (RAE). Previous work studying the RAE in the NHL has focused on individual NHL seasons, often encompassing many of the same players across multiple seasons. We investigate the RAE using complete data on every player who has ever played in the NHL. We focus the majority of our analysis on Canadian born players and examine the RAE across hockey position and hall-of-fame status. For the first time, we provide strong evidence of when the RAE began to manifest itself in Canada. Our change point analysis indicates that the RAE began for players born since 1951. Finally, we make a case for what initiated this change in the way young hockey players develop, particularly in Canada, which produced over 90% of NHL players at that time.

Judo is an intermittent sport. As such, it is important to characterize the actions involved in combat so that training may be structured in such a way as to simulate competitive demands. However, to do so, an objective notational analysis system is necessary. The aim of the present study was to design a computer program that would aid in the analysis of time structure of specific actions during judo combat and test its inter-evaluator and intra-evaluator reliability. Ten male judokas, divided by class and category, were evaluated during three combats each. The matches were filmed and the evaluators used the computer program Saats™ (Structural Analysis of Action and Time in Sports) to analyze the following actions: break, grip, technique, fall and groundwork. The sequences were characterized by the sum of actions between each break. A total of 276 action sequences were evaluated, with a mean of 11 action sequences per combat, with four on the ground. Two evaluations were carried out by three evaluators for each judo match (inter-evaluator agreement), with only one being an expert in the software used. There was a lack of similarities in the results of only two of the variables (p<0.05). Evaluations by the same evaluator (intra-evaluator agreement) demonstrated a high reliability on all six variables. It was concluded that the use of this computer software for notational analysis in judo greatly assists in the detailing of actions performed by the athletes. The use of this software by professionals unfamiliar with it likely requires a short learning period. Knowledge of judo actions will very likely allow practitioners of the sport to be trained more specifically.

A conceptual basis for examining team defense in the National Football League is presented. Based on these concepts, a simple method is proposed to analyze the weekly defensive performance of NFL teams. Using overall game data that is easily obtained, the procedure computes six indices which characterize the functioning of each defensive team for every game and over the entire season.The results presented here can be used to determine a rank ordering of all team defenses for the entire season. The resultant ranking compares quite favorably with a similar ranking determined by a process that employs play-by-play data.