2022 Big Data Cup

Congratulations to team members Ethan BaronDaniel HocevarKabir Malik, and Aaron White for winning first place in the undergraduate category of the 2022 Big Data Cup, a hockey datathon hosted by Stathletes and CANSSI!

Their submission, titled “RIPP: Holistic Player Evaluation with Region-Based Isolated Player Performance,” presents a method to evaluate women’s hockey players using play-by-play data.

Check out their talk at https://www.youtube.com/watch?v=QcfcsO8A3k0.

UTSPAN Hour #8: March Madness Bracket: How Odds, Probabilities, and Machine Learning Can Be Used to Predict Game Winners

Thursday, March 10th, 2022, 8pm EST
Colin Conant

Resources:

UTSPAN Hour #1: How Can I Get Data For Sports Analytics Projects?

Monday, January 10th, 2022, 8pm EST
Hassaan Inayatali

Resources:

Julien Guyon – Risk of Collusion in FIFA World Cup and a New Tournament Design

Friday, December 4th, 2020, 6pm EST
Julien Guyon (Bloomberg, NYU, Columbia)
(1) Risk of Collusion: Will Groups of Three Ruin the FIFA World Cup?
(2) “Choose Your Opponent”, a New Tournament Design

Please find a recording of the presentation below.

In 2026, the FIFA World Cup will for the first time gather 48 men’s national teams. It will consist of a group stage made of 16 groups of three, with the best two teams in each group advancing to the knockout stage. Using groups of three raises several fairness issues, including risk of match fixing and schedule imbalance. In this article we examine the risk of collusion. The two teams who play the last game in the group know exactly what results will let them advance to the knockout stage. Suspicion of match fixing occurs when a result qualifies both of them at the expense of the third team of the group, and can seriously tarnish the tournament. We quantify how often this is expected to happen and explain how to build the match schedule so as to minimize the risk of collusion. We also quantify how the risk of collusion depends on competitive balance. Moreover, we show that forbidding draws during the group stage (a rule considered by FIFA) does not eliminate the risk of match fixing, and that surprisingly when draws are forbidden the 3-2-1-0 point system does not do a better job at decreasing the risk of collusion than the 3-0 point system. Finally we describe alternate formats for a 48 team World Cup that would eliminate or strongly decrease the risk of collusion. Then, we present a new knockout format for sports tournaments, that we call “choose your opponent”, where the teams that have performed best during a preliminary group stage can choose their opponents during the subsequent knockout stage.

Ioannis Ntzoufras – A Unified Bayesian Model for Volleyball Data

Tuesday, November 24th, 2020, 11am EST
Ioannis Ntzoufras (Athens University of Economics and Business)
ntzoufras@aueb.gr
Bayesian Quest for Finding a Unified Model for Predicting Volleyball Games

See a recording of the presentation below.

Volleyball is a team sport with unique and specific characteristics. We introduce a new two level-hierarchical Bayesian model which accounts for these volleyball specific characteristics. In the first level, we model the set outcome with a simple logistic regression model. Conditionally on the winner of the set, in the second level, we use a truncated negative binomial distribution for the points earned by the loosing team. An additional Poisson distributed inflation component is introduced to model the extra points played in the case that the two teams have point difference less than two points. The number of points of the winner within each set is deterministically specified by the winner of the set and the points of the inflation component. The team specific abilities and the home effect are used as covariates on all layers of the model (set, point, and extra inflated points). The implementation of the proposed model on the Italian Superlega 2017/2018 data shows an exceptional reproducibility of the final league table and a satisfactory predictive ability.

Queen’s Geometric Sport Analysis Group – Geometrically Modeling Soccer and Basketball Games

Friday, November 20th, 2020, 6pm EST
Dan Forestall, Emily Hunter, Sara Stephens & Maia Gibbon (Queen’s Geometric Sport Analysis Group)
Geometrically Modeling Soccer and Basketball Games

To watch the presentation, see below.

With a focus on understanding the impact of player position, and how it evolves over time, we look at several 3-dimensional models of soccer & basketball games. This work builds off of that of Pleurer, Spearman, and others.

David Perdomo Meza – Stylistic Representation of Team Playing Style Using Latent Dirichlet Allocation

Monday, November 16th, 2020, 4pm EST
David Perdomo Meza (Twenty3 Sport)
Email: dperdomomeza@gmail.com
Twitter: @dperdomomeza1
Title: Tactical Insight through Stylistic Representation of Team Playing Style Using Latent Dirichlet Allocation

Check out the recording below:

We’ll showcase the application of Latent Dirichlet Allocation as a topic modelling technique on football statistics to obtain a mixture model representation of “team playing style” – and go through its application to evaluating tactical choices in team head to heads in the English Championship.

Jeremy Alexander – Impact of Team Numerical Advantage in Australian Rules Football

Monday, November 2nd, 2020
Jeremy Alexander (Victoria University)
The Impact of a Team Numerical Advantage on Match Play in Australian Rules Football

See a recording below.

The advent of player tracking technologies has supported a more detailed approach to the match analysis of invasion sports. Studies to date that have investigated tactical team behaviour by measuring how players occupy different sub-areas on a playing field in football have inferred performance by assessing a team’s capacity to generate a numerical advantage over a specific area. As such, a limited understanding exists between a team’s numerical advantage and the impact on match play in a continuous manner. Therefore, the primary aim of this study was to provide a proof of concept that determines the relationship between a team numerical advantage and match play in a continuous manner. The secondary aim was to determine how ball position and match phase influence how players occupy different sub-areas of play in Australian Rules Football.

Tutorial: Web Scraping

Information
Our first tutorial took place on Friday, October 23rd at 7:00pm EST. If you missed it, check out the notebook we used at https://utoronto-my.sharepoint.com/:u:/g/personal/eth_baron_mail_utoronto_ca/EaRu6yXz9M1MuO5VBZ0ed6EBocTIFdpsiTgoDMmasYtqNA?e=73asnW.

Learn how to web scrape
Web Scraping is a valuable tool, especially for those that enjoy creating independent coding projects. It allows for more creative projects by giving the user the ability to obtain their own niche data. This tutorial will focus on the fundamentals of how to web scrape, with additional focus on how a webpage is broken down and how to clean your obtained data.

Kenneth Brent Smale – Transitioning from Academia to Industry: Analytics in Pro Sports

Friday, October 30th, 2020
Kenneth Brent Smale (Los Angeles Angels, Apex Skating)
Transitioning from Academia to Industry: Analytics in Pro Sports

Check out the recording below.

As a student, the bulk of your training in analytics comes in the classroom and is heavily involved in the theory and simple strong signal-to-noise examples. In reality, and particularly in sports, things get much noisier with true data and different personalities and stakeholders. Kenneth Smale will talk through just how analytics differs from academia to the industry and provide guidance on how to make the transition as easy as possible.

Dani Chu – It’s Fun Getting Into (Foul) Trouble

Wednesday, October 14th, 2020, 6:00pm EST
Dani Chu (Seattle Kraken)
It’s Fun Getting Into (Foul) Trouble

This project investigates the fouling time distribution of players in the National Basketball Association. A Bayesian analysis is presented based on the assumption that fouling times follow a Gamma distribution. Methods are developed that will allow coaches to better manage their players under the threat of fouling out.

See a recording of the presentation below.

Abdullah Zafar – Mathematical Modelling in Professional Sport

Friday, October 9th, 2020, 6:00pm EST
Abdullah Zafar (Sports Performance Analytics Inc.)
Mathematical Modelling in Professional Sport

How to quantify actions in sport in order to build metrics, get insights, and drive performance? In this talk, we will overview, compare and contrast approaches using football (soccer) data from the Danish Superliga; focusing on how we can model the movement of a team using flow fields and dynamical systems, derive metrics to quantify team tempo, and then demonstrate the utility and application to the physical training of players as well as team performance as a whole. We will then break down tempo further using topological time series analysis to better understand the dynamics of a football match and highlight the difference in teams during goal-scoring moments.

For a recording of the presentation see below.

2015 Toronto Blue Jays’ Hitters: A PITCHf/x Preview

Written by: Doug Duffy

With the Blue Jays flying north and spring training squarely in the rear-view mirror, Torontonians can leave behind their masochistic winter ritual, watching the Maples Leafs, in favor of their spring ritual, asking “Is the dome open yet?”. Incidentally, in the year 2015, there’s a twitter account for that. Along with the arrival of Opening Day has come the annual barrage of previews, some more quantitative than others. My personal favorite was the series of previews published on Grantland for each of baseball’s divisions, nicely melding projection systems, depth charts and win projections with more qualitative subjects like strengths/weaknesses and storylines. I, however, chose to take the opportunity to dig into the Pitch f/x database for the first time to see what it can tell us about the 2015 Blue Jays hitters, especially the new arrivals. Hopefully, I’ll get around to performing something similar for the pitching side of things soon.

Read more 2015 Toronto Blue Jays’ Hitters: A PITCHf/x Preview

Using Projection Models for 2015 Fantasy Baseball Drafts

Written by: Doug Duffy

If you’ve ever participated in a fantasy draft of any kind, you’re familiar with the concept of projections. Projections, they’re (almost) as simple as they sound. What do you project a given player to accomplish based on his past accomplishments? Projections are not restricted to the realm of fantasy sports however; teams utilize projections as well, to assist them in player valuation. In this post I’ll explain how you can use projections for player valuation for your own fantasy baseball league, using a model based either on Standing Points Gained above replacement, or Fantasy Points above replacement, depending upon the scoring system of the league [1]. In addition, I’ll be posting the R code used to perform the models, as well as Draft Cheat Sheets containing relevant draft info from many of the sources we searched.

5x5Roto10TeamDraftsheet

5x5Roto12TeamDraftsheet

Points10TeamDraftsheet

Points12TeamDraftsheet

Special Request (AL-only 4×4 10 Team no R or K) : 4x4ALonly10Team

Update (3/21/2015) : The projection database and draftsheets have all been updated, and the R code used to calculate TOTspgAR and FPtsAR has been posted. Enjoy.

Read more Using Projection Models for 2015 Fantasy Baseball Drafts

Sports Industry Conference 2015 & Next Meeting

Earlier today the UTSPAN team had the pleasure of attending the 2015 Sports Industry Conference hosted by the University of Toronto Sports and Business Association (UTSB). UTSB were kind enough to let us set up a small booth with a poster to show off a bit of our work and extend an invitation to all attendees. The conference was a big success and we would like to thank UTSB for hosting an extremely well-organized event with top class panels.

We would like to announce that our next meeting will be held on Monday 16 March 2015 at 7PM in room 3008 of the Bahen Centre for Information Technology (40 St George Street). Join us to learn more about the group and what we hope to achieve in the coming months! We will also be hoping to discuss some possibilities for the upcoming UTSPAN Data Hack-a-thon!

A big thank you to everyone who stopped by the booth to chat with us! If you have any questions feel free to reach out to us on Twitter or by email at: sportsanalytics@utoronto.ca

Introduction to Analytics in… Soccer

Written by: Valentin Stolbunov

Soccer, or football, or footy, or “the beautiful game” is the world’s most popular sport. When trying to prove this to a fan of North American sports, a soccer fan’s best weapon is usually global TV audience numbers. The 2014 Super Bowl had an audience of about 160 million viewers worldwide. The same year, the FIFA World Cup final had a global audience of about 1 billion. So, yeah, soccer is popular.

The recent sports analytics movement, however, didn’t originate from the world’s most popular sport. Most would agree it started with baseball and then spread to other North American sports – hockey, basketball, and football (the one with helmets). Compared to these sports, the use of advanced or “fancy” stats in soccer is still in the early stages.

Read more Introduction to Analytics in… Soccer

Introduction to Analytics in… Baseball

Written by: Kurtis Judd

Whether you’re simply interested in following home run races, or using programming languages to predict next year’s MVP, it’s hard to argue that baseball isn’t a statistics driven sport. Every event in the game is so discrete, that it’s a statistician’s dream of clean, easy to work with data.

Read more Introduction to Analytics in… Baseball

UTSPAN Forum

We have set up a free forum for UTSPAN on ProBoards! A link has also been added to the menu bar in the top right of our webpage.

This will serve as our primary project work platform as it allows us to share ideas, images, code, links, and just about anything else. The forum will be visible publicly and there are no restrictions on membership. Sign up and get started!

utspan.proboards.com

Meeting: Project Discussion

We will be holding our second meeting on Monday 26 January 2015 at 7PM in BA2135.

This meeting will have a number of short presentations from the smaller project exploration teams. We aim to decide on the focus of the handful of projects which will ultimately be our work for the next couple of months.

New members are always welcome! Come out and see what UTSPAN is planning to do!

Meeting: General Information

We will be holding a general information meeting for anyone interested in joining the group or just hearing more about what we plan to do!

  • Monday 12 January 2015
  • 7PM
  • Bahen Centre for Information Technology (40 St George Street)
  • Room #2135

The meeting will likely cover:

  • An introduction from the executives
  • A quick introduction to the field of sports analytics
  • An overview of how the group will operate
  • A discussion of the milestones we are currently looking at

This meeting will be followed by a more technical meeting in the coming weeks.

Introduction to UTSPAN

Written by: Valentin Stolbunov

Our mission statement

At the University of Toronto Sports Analytics Group, we aim to connect members who share an interest in the field. We also aim to support the analytics process and help members explore their own interests. Members can work together to find and manage data, to develop and test analytic models, and to present and publish their findings. Last but not least, we hope to connect members with industry professionals.

Our methodology

It is clear from our earlier introduction to sports analytics that Alamar’s structure of the sports analytics process is about feeding the decision makers. This is an indication of the place sports analytics has in a sports organization. However, because we do not operate within any particular organization, the structure needs to be revised slightly.

Fig3

 

We do not have decision makers to whom we would provide our findings. Instead, our work is motivated by a question or a general area of interest. In many ways, this is similar to a coach asking the analytics team a question along the lines of “who is the best player we can sign for around $3 million?”. So instead of using a methodology that is designed to feed into a larger framework, we have altered the structure to instead begin with an element of motivation.

With no decision makers to facilitate, the information systems are no longer used to support the final element in the process. As the final element themselves, these systems must now focus on presenting the findings of the data and/or models in the same effective and efficient manner. Hence their new name: Presentation of Results. This presentation may be verbal (an essay-style argument that Player A is a better “scorer” than Player B), or visual (a visualization which shows the two players’ scoring habits), or both.

Our methodology will follow the sports analytics process above. A more specific, question-based, step-by-step process would look something like this,

  1. What is the area of interest we are looking to explore?
  2. Given (1), what type of data do we require? How do we best obtain and organize this data?
  3. Given (1) and (2), what type of analytic models, if any, should we use?
  4. Given (2) and (3), what is the best way to present our findings?

A basic example

Let us assume we were looking to determine who is currently the most dominant scorer in the NBA. This problem is ultimately as complicated as you would like to make it, but for the purposes of this demonstration we will keep it simple. The answers to our questions above would look something like this,

  1. Area of interest: offensive performances in the NBA
  2. Data required: points per game and field goal percentage this season
  3. Analytic models: none
  4. Presentation of results: rank players based on offensive data

Fig4

A more advanced example

A more difficult problem would be determining the best “lock down defender” in the NBA. Some sort of model would probably be necessary this time and the answers would look something like this,

  1. Area of interest: 1 on 1 defensive performances in the NBA
  2. Data required: defensive statistics (steals, blocks) as well as change in opponent’s shooting percentage, preferably adjusted for opponent’s time on the court and
  3. Analytic models: obtain distributions for defensive statistics and the defender’s impact on the opposing player’s offensive performance
  4. Presentation of results: use average statistics to compare defending ability and variance of statistics to compare consistency, rank players

Fig5

Introduction to Sports Analytics

Written by: Valentin Stolbunov

Defining: Analytics

Before jumping to Google and adding “wiki” to the end of my search query, I thought I’d try to define analytics myself. When I think of analytics, I usually think of finding patterns in data and using those patterns to answer questions. Wikipedia says I am not too far off:

Analytics is the discovery and communication of meaningful patterns in data.

The important thing to note at this point is that analytics is a process. In fact, it is an interdisciplinary process which usually brings together mathematics, statistics, computer science, predictive methods, data visualization, and other fields of study.

It is also important to note that analytics relies on the presence of data. This ultimately differentiates the term from “analysis” and unfortunately creates confusion when trying to decide if what you are doing is analytics or “data analysis”. For our intents and purposes, the two are essentially the same process. However, the field has been dubbed “sports analytics” and not “sports data analysis”, so we will accept the name and move on.

Before continuing to the sports side of things, it should be noted that the term “analytics” may also be used to describe the results of this process. For example, “the analytics of our last project suggest that…” is a perfectly valid sentence. However, in my experience I find that here it is best to just replace “analytics” with “analytical findings” or “results” and reserve the term “analytics” for the process through which these results are obtained.

Defining: Sports Analytics

This is where Wikipedia does not offer much help – nor does it need to. Sports analytics is essentially the analytics process, as described above, applied to sports.

It is the process of using sports-related data (anything from player statistics to game day weather) to find meaningful patterns (strong correlations, hidden trends, etc.) and communicate those patterns (using graphs, charts, essays, etc.) to help make decisions.

Fig1

 

In his book, Benjamin Alamar presents a helpful graphic to illustrate the overall sports analytics process. In his framework, sports analytics consists of four elements: data management, analytic models, information systems, and the decision maker.

I have done my best to provide both Alamar’s definition of each element and my own thoughts on their uses and values:

  • Data management: This includes any and all processes associated with acquiring, verifying and storing data. The data management element is ultimately about facilitating the modelling and information-extraction elements. As mentioned earlier, you can’t have analytics without data.
  • Analytic models: This element is essentially the process of applying statistical tools to data. The use of models to “forecast” player or team performance is often the most popular goal, but it is by no means a necessity. The models may or may not offer insight into the future. It is most accurate to say that they are concerned with using mathematics and statistics to describe the data.
  • Information systems: Unlike the previous two elements, the information systems are slightly more abstract. The purpose of these systems is to extract and present the data and/or model results as effectively and efficiently as possible. A scouting report is a good basic example.
  • Decision makers: The end goal of analytics is to extract relevant and insightful information from the data and present it to the decision makers. In modern sports these tend to be the coaching staff or management, however, players themselves may also benefit from the whole process.

What is the history of the field?

Although many professionals believe that modern model-heavy sports analytics is at a point of exciting growth, the field of sports analytics is by no means new. Technically speaking, any time anyone has ever used data to make a decision related to sport, they were conducting analytics. However, the general consensus is that sports analytics began sometime in the 19th century with baseball. The data (basic statistics such as hits and pitches) was collected with good old pencil and paper. It was then used create scouting reports which a coach or manager would use to make decisions about their team.

Referring back to Alamar’s graphic of the entire process, this type of analytics would lack a modelling element but still follow a logical flow toward the decision maker. These 19th century baseball decisions to be made were perhaps fewer and less detailed, but not necessarily easier.

Fig2

 

What is the current state of the field?

We now have two things which we didn’t have in the 19th century of baseball analytics. The first of these is more sports nerds. Sports have grown in popularity and fans have become much more demanding of information. More often than not sports arguments include statistics, even if they are about whether or not “number of rings” is a statistic. Everyone and their parents have a fantasy team and compulsively refresh Twitter in hopes of finding out how long Derrick Rose will be out for this season.

The second thing we now have, which in some ways overlaps with the first, is more data. The recent advances in technology have affected just about every aspect of life, and sports is no different. The following things have all contributed to the recent growth in the field:

  • The improvements in computing power and digital memory
  • The increased quantification of our world (aka the ultimate buzzword: “big data”)
  • The advances made in solving complex engineering problems like vision and inference

Modern sports analytics uses database management systems and things like SQL where pen and paper were once the norm. Analytical models from machine learning and data mining are now used to help sort through the data and find patterns. Models are now updated in real time and together with innovative visualization techniques are the new breed of information system.

With more data, and more people interested in sports analytics, organizations are doing their best to gain every possible advantage in every aspect of sports from training routines to player recruitment and valuation.

What does the future hold?

The field is growing.

More and more sports organizations are hiring analytics “teams” and “departments”, usually composed of professionals with STEM (science, technology, engineering, mathematics) degrees. The media appears to be following suit by recruiting data science professionals to find and visualize the unique trends that their viewers want to see. There is no reason to believe that these new opportunities will stop popping up or disappear all together.

In addition, social media has helped connect fans and form communities of the analytically-inclined. Whether out of personal interest or in hopes of being noticed, more fans will create stats-based blogs and continue to explore the numbers of their sport.

The field has also not gone unnoticed in academia. If conferences like MIT’s Sloan, journals like Quantitative Analysis in Sports, and new courses offered by top universities are any indication, institutions have noticed the growth in sports data and are interested in conducting research in the field.

Sports analytics is sometimes discounted as just an invention of weird metrics. But it is much more than that. From engineering solutions in data, like SportsVu, to innovative information systems, like shot charts, the future of the field is in ultimately working to advance each step of the whole process.