scholarly journals Can Elite Australian Football Player’s Game Performance Be Predicted?

2021 ◽  
Vol 20 (1) ◽  
pp. 55-78
Author(s):  
J. Fahey-Gilmour ◽  
J. Heasman ◽  
B. Rogalski ◽  
B. Dawson ◽  
P. Peeling
Keyword(s):  
Model Performance ◽  
Test Set ◽  
Australian Football League ◽  
Team Selection ◽  
Individual Player ◽  
Australian Football ◽  
Team Characteristics ◽  
Using Data ◽  
Player Performance ◽  
Game Performance

Abstract In elite Australian football (AF) many studies have investigated individual player performance using a variety of outcomes (e.g. team selection, game running, game rating etc.), however, none have attempted to predict a player’s performance using combinations of pre-game factors. Therefore, our aim was to investigate the ability of commonly reported individual player and team characteristics to predict individual Australian Football League (AFL) player performance, as measured through the official AFL player rating (AFLPR) (Champion Data). A total of 158 variables were derived for players (n = 64) from one AFL team using data collected during the 2014-2019 AFL seasons. Various machine learning models were trained (cross-validation) on the 2014-2018 seasons, with the 2019 season used as an independent test set. Model performance, assessed using root mean square error (RMSE), varied (4.69-5.03 test set RMSE) but was generally poor when compared to a singular variable prediction (AFLPR pre-game rating: 4.72 test set RMSE). Variation in model performance (range RMSE: 0.14 excusing worst model) was low, indicating different approaches produced similar results, however, glmnet models were marginally superior (4.69 RMSE test set). This research highlights the limited utility of currently collected pre-game variables to predict week-to-week game performance more accurately than simple singular variable baseline models.

scholarly journals Multifactorial analysis of factors influencing elite Australian football match outcomes: a machine learning approach

2019 ◽  
Vol 18 (3) ◽  
pp. 100-124
Author(s):  
J. Fahey-Gilmour ◽  
B. Dawson ◽  
P. Peeling ◽  
J. Heasman ◽  
B. Rogalski
Keyword(s):  
Machine Learning ◽  
Prediction Models ◽  
Model Performance ◽  
Test Set ◽  
Australian Football League ◽  
Linear Relationships ◽  
Machine Learning Approach ◽  
Football Match ◽  
Australian Football ◽  
Relative Form

Abstract In Australian football (AF), few studies have assessed combinations of pre- game factors and their relation to game outcomes (win/loss) in multivariable analyses. Further, previous research has mostly been confined to association-based linear approaches and post-game prediction, with limited assessment of predictive machine learning (ML) models in a pre-game setting. Therefore, our aim was to use ML techniques to predict game outcomes and produce a hierarchy of important (win/loss) variables. A total of 152 variables (79 absolute and 73 differentials) were used from the 2013–2018 Australian Football League (AFL) seasons. Various ML models were trained (cross-validation) on the 2013–2017 seasons with the–2018 season used as an independent test set. Model performance varied (66.5-73.3% test set accuracy), although the best model (glmnet – 73.3%) rivalled bookmaker predictions in the same period (70.9%). The glmnet model revealed measures of team quality (a player-based rating and a team-based) in their relative form as the most important variables for prediction. Models that contained in-built feature selection or could model non-linear relationships generally performed better. These findings show that AFL game outcomes can be predicted using ML methods and provide a hierarchy of predictors that maximize the chance of winning.

Classifying player positions in second-tier Australian football competitions using technical skill indicators

2021 ◽  
pp. 174795412110102
Author(s):  
Adrian J Barake ◽  
Heather Mitchell ◽  
Constantino Stavros ◽  
Mark F Stewart ◽  
Preety Srivastava
Keyword(s):  
Positional Information ◽  
Multinomial Logistic Model ◽  
Sports Science ◽  
Lower Tier ◽  
Australian Football League ◽  
Australian Football ◽  
Second Tier ◽  
Performance Statistics ◽  
Player Performance ◽  
Player Positions

Efficient recruitment to Australia’s most popular professional sporting competition, the Australian Football League (AFL), requires evaluators to assess athlete performances in many lower tier leagues that serve as pathways. These competitions and their games are frequent, widespread, and challenging to track. Therefore, independent, and reliable player performance statistics from these leagues are paramount. This data, however, is only meaningful to recruiters from AFL teams if accurate player positions are known, which was not the case for the competitions from which most players were recruited. This paper explains how this problem was recently solved, demonstrating a process of knowledge translation from academia to industry, that bridged an important gap between sports science, coaching and recruiting. Positional information which is only available from the AFL competition was used to benchmark and develop scientific classification methods using only predictor variables that are also measured in lower tier competitions. Specifically, a Multinomial Logistic model was constructed to allocate players into four primary positions, followed by a Binary Logit model for further refinement. This novel technique of using more complete data from top tier competitions to help fill informational deficiencies in lower leagues could be extended to other sports that face similar issues.

scholarly journals Biomechanical Loads and Their Effects on Player Performance in NCAA D-I Male Basketball Games

2021 ◽  
Vol 3 ◽  
Author(s):  
Sigrid B. H. Olthof ◽  
Tahmeed Tureen ◽  
Lam Tran ◽  
Benjamin Brennan ◽  
Blair Winograd ◽  
...  
Keyword(s):  
Movement Analysis ◽  
Individual Performance ◽  
Loss Probabilities ◽  
Point Field ◽  
Individual Player ◽  
Measurement Units ◽  
Player Performance ◽  
Game Performance ◽  
Level Training ◽  
And Training

Basketball games and training sessions are characterized by quick actions and many scoring attempts, which pose biomechanical loads on the bodies of the players. Inertial Measurement Units (IMUs) capture these biomechanical loads as PlayerLoad and Inertial Movement Analysis (IMA) and teams collect those data to monitor adaptations to training schedules. However, the association of biomechanical loads with game performance is a relatively unexplored area. The aims of the current study were to determine the statistical relations between biomechanical loads in games and training with game performance. Biomechanical training and game load measures and player-level and team-level game stats from one college basketball team of two seasons were included in the dataset. The training loads were obtained on the days before gameday. A three-step analysis pipeline modeled: (i) relations between team-level game stats and the win/loss probabilities of the team, (ii) associations between the player-level training and game loads and their game stats, and (iii) associations between player-level training loads and game loads. The results showed that offensive and defensive game stats increased the odds of winning, but several stats were subject to positional and individual performance variability. Further analyses, therefore, included total points [PTS], two-point field goals, and defensive rebounds (DEF REB) that were less subject to those influences. Increases in game loads were significantly associated with game stats. In addition, training loads significantly affected the game loads in the following game. In particular, increased loads 2 days before the game resulted in increased expected game loads. Those findings suggested that biomechanical loads were good predictors for game performance. Specifically, the game loads were good predictors for game stats, and training loads 2 days before gameday were good predictors for the expected game load. The current analyses accounted for the variation in loads of players and stats that enabled modeling the expected game performance for each individual. Coaches, trainers, and sports scientists can use these findings to further optimize training plans and possibly make in-game decisions for individual player performance.

scholarly journals To Include, or Not Include, that is the Question: An Empirical Analysis of Dealing with Patients who are Lost to Follow-up when Developing Prognostic Models Using a Cohort Design

2020 ◽  
Author(s):  
Jenna M Reps ◽  
Peter Rijnbeek ◽  
Alana Cuthbert ◽  
Patrick B Ryan ◽  
Nicole Pratt ◽  
...  
Keyword(s):  
At Risk ◽  
Synthetic Data ◽  
Model Performance ◽  
Test Set ◽  
Loss To Follow Up ◽  
Using Data ◽  
Lost To Follow Up ◽  
The Impact ◽  
Prediction Problems

Abstract Background: Researchers developing prediction models are faced with numerous design choices that may impact model performance. One of the main decisions is how to include patients who are lost to follow-up. In this paper we perform a large-scale empirical evaluation investigating the impact of this decision. In addition, we aim to provide guidelines for how to deal with loss to follow-up. Methods: We generate a synthetic dataset with complete follow-up and simulate loss to follow-up based either on random selection or on selection based on comorbidity. We investigate four simple strategies for developing models using data containing some patients with loss to follow-up. Three strategies employ a binary classifier with data that: i) include all patients (including those lost to follow-up), ii) exclude all patients lost to follow-up or iii) only exclude patients lost to follow-up who do not have the outcome before being lost to follow-up. The fourth strategy uses a survival model with data that include all patients. In addition to our synthetic data study, we empirically evaluate the discrimination and calibration performance of these strategies across 21 prediction problems using real-world data. Results: The synthetic data study results show that excluding patients lost to follow-up can introduce bias when loss to follow-up is common and does not occur at random. However, when loss to follow-up was completely at random, the choice of addressing it had negligible impact on the model performance. Our empirical results showed that the four design choices investigated to deal with loss to follow-up resulted in comparable performance when the time-at-risk was 1-year, but demonstrated differential bias when we looking into 3-year time-at-risk. Removing patients who are lost to follow-up before the outcome but keeping patients who are loss to follow-up after the outcome can bias a model and should be avoided. Conclusion: Based on this study we therefore recommend i) developing models using data that includes patients that are lost to follow-up and ii) evaluate the discrimination and calibration of models twice: on a test set including patients lost to follow-up and a test set excluding patients lost to follow-up.

Game movements and player performance in the Australian Football League

2012 ◽  
Vol 12 (3) ◽  
pp. 531-545 ◽  
Author(s):  
Daniel Hiscock ◽  
Brian Dawson ◽  
Jarryd Heasman ◽  
Peter Peeling
Keyword(s):  
Australian Football League ◽  
Australian Football ◽  
Player Performance

Predictors of individual player performance in junior Australian football

2014 ◽  
Vol 18 ◽  
pp. e82
Author(s):  
C. Tangalos ◽  
S. Robertson ◽  
M. Spittle ◽  
P. Gastin
Keyword(s):  
Individual Player ◽  
Australian Football ◽  
Player Performance

scholarly journals Comparing bottom-up and top-down ratings for individual soccer players

2021 ◽  
Vol 20 (1) ◽  
pp. 23-42
Author(s):  
Lars Magnus Hvattum ◽  
Garry A. Gelade
Keyword(s):  
Reliability And Validity ◽  
Data Sets ◽  
Top Down ◽  
Bottom Up ◽  
Team Selection ◽  
Selection Decisions ◽  
Individual Player ◽  
Using Data ◽  
Or Team ◽  
Main Ideas

Abstract Correctly assessing the contributions of an individual player in a team sport is challenging. However, an ability to better evaluate each player can translate into improved team performance, through better recruitment or team selection decisions. Two main ideas have emerged for using data to evaluate players: Top-down ratings observe the performance of the team as a whole and then distribute credit for this performance onto the players involved. Bottom-up ratings assign a value to each action performed, and then evaluate a player based on the sum of values for actions performed by that player. This paper compares a variant of plus-minus ratings, which is a top-down rating, and a bottom-up rating based on valuing actions by estimating probabilities. The reliability of ratings is measured by whether similar ratings are produced when using different data sets, while the validity of ratings is evaluated through the quality of match outcome forecasts generated when the ratings are used as predictor variables. The results indicate that the plus-minus ratings perform better than the bottom-up ratings with respect to the reliability and validity measures chosen and that plus-minus ratings have certain advantages that may be difficult to replicate in bottom-up ratings.

Is injury associated with team performance in elite Australian football? 20 years of player injury and team performance data that include measures of individual player value

2019 ◽  
Vol 54 (8) ◽  
pp. 475-479 ◽  
Author(s):  
Daniel Tyler Hoffman ◽  
Dan Brian Dwyer ◽  
Steven J Bowe ◽  
Patrick Clifton ◽  
Paul B Gastin
Keyword(s):  
Team Performance ◽  
Injury Severity ◽  
Mean Difference ◽  
Injury Incidence ◽  
Australian Football League ◽  
Individual Player ◽  
Australian Football ◽  
The One ◽  
Injury Burden ◽  
The Impact

ObjectivesTo determine whether specific injury measures were associated with team performance in the Australian Football League (AFL).Methods15 289 injuries caused players from 18 teams to miss 51 331 matches between 1997 and 2016. Data were aggregated to the team level. We analysed the associations among injury measures and team performance (reaching finals/playoffs and specific ladder/table position). Injury measures per team included: injury incidence, injury severity, injury burden, player match availability and percentage of the full player roster injured. We also weighted injury measures by five measures of player value.ResultsAFL teams’ injury burden and player match availability were associated with final table position (r2=0.03, p<0.05). Player value weighted injury burden was different between finalists and non-finalists (mean difference=−8, p<0.001) and explained 12% of the variation in the table position of teams (p<0.001). For a team, nine missed matches due to injury (burden weighted by a best and fairest player rating system) was associated with one lower table position. Player match availability weighted by player value was higher for finalists than non-finalists (mean difference=1.7, p<0.01) and explained 7% of the variation in the table position of teams (p<0.001).Discussion and potential implicationsThe impact of injury (burden weighted by best and fairest) explained up to 12% of the variation in final table position—this is particularly relevant to making/not making playoffs as well as home ground/travel advantages for those teams that make the one-game format of AFL playoffs (not home-away or best of seven format).

Sign in / Sign up

Export Citation Format

Share Document