Context: Valorant, developed by Riot Games, has quickly become one of the most popular tactical first-person shooter games since its release. The game emphasizes strategic team play, individual skills, and tactical execution, making it a fascinating subject for performance analysis. Understanding the various metrics that contribute to player success can offer insights into effective strategies and gameplay techniques. This dataset was created to help players, coaches, and analysts delve into the detailed aspects of player performance and identify key areas for improvement.

Sources: The data for this dataset was collected from various online sources, including:

In-Game Statistics: Aggregated from player profiles and match histories available within the game client. Third-Party Valorant Trackers: Websites and tools that track player statistics and match performance, such as Tracker.gg and Blitz.gg. Community Contributions: Insights and data shared by the Valorant community, including professional players, streamers, and analysts, who often provide detailed breakdowns of their gameplay. Inspiration: The inspiration for compiling this dataset stems from several key areas:

Performance Analysis: In competitive gaming, understanding the granular details of player performance is crucial for improvement. Metrics like win rate, damage per round, and headshot percentage provide actionable insights. Strategic Development: By analyzing this data, players and teams can develop better strategies, identify strengths and weaknesses, and tailor their training regimes accordingly. Predictive Modeling: The dataset serves as a foundation for building predictive models to forecast future performance, which can be useful for coaching, match preparation, and scouting new talent. Community Engagement: Providing this dataset to the wider Valorant community fosters engagement and encourages collaborative analysis. It allows enthusiasts to test hypotheses, share findings, and contribute to a deeper understanding of the game. Educational Purposes: For educators and students in data science, sports analytics, and game design, this dataset offers a real-world application of data analysis techniques and methodologies. Future Directions: The dataset can be expanded by including additional metrics such as agent pick rates, map-specific performance, and team composition analysis. Incorporating more granular data over longer periods can also enhance the depth of analysis and provide a more comprehensive view of player performance trends.

By sharing this dataset, we aim to empower the Valorant community with data-driven insights that can elevate gameplay, inform strategic decisions, and contribute to the overall growth of the esports ecosystem.

Patterns of wins and losses in pairwise contests, such as occur in sports and games, consumer research and paired comparison studies, and human and animal social hierarchies, are commonly analyzed using probabilistic models that allow one to quantify the strength of competitors or predict the outcome of future contests. Here we generalize this approach to incorporate two additional features: an element of randomness or luck that leads to upset wins, and a "depth of competition" variable that measures the complexity of a game or hierarchy. Fitting the resulting model we estimate depth and luck in a range of games, sports, and social situations. In general, we find that social competition tends to be "deep," meaning it has a pronounced hierarchy with many distinct levels, but also that there is often a nonzero chance of an upset victory. Competition in sports and games, by contrast, tends to be shallow and in most cases there is little evidence of upset wins. Methods This data set of pairwise comparisons has been aggregated across a variety of sources given in our paper. These directed interactions are given in adjaency matrix, graph modeling language, and edge list formats.

Online competitive action games are a very popular form of entertainment. While most are respectfully enjoyed by millions of players, a small group of players engages in disruptive behavior, such as cheating and hate speech. Identifying and subsequently moderating these toxic players is a challenging task. Previous research has only studied specific aspects of this problem using curated data and with limited access to real-world moderation practices. In contrast, our work offers a unique and holistic view of the universal challenges of moderating disruptive behavior in online systems. We combine an analysis of a large dataset from a popular online competitive first-person action title (Call of Duty®: Modern Warfare®II) with insights from stakeholders involved in moderation. We identify six universal challenges related to handling disruptive behaviors in such games. We discuss challenges omitted by prior work, such as handling high-volume imbalanced data or ensuring the comfort of human moderators. We also offer a discussion of possible technical, design, and policy approaches to mitigating these challenges.

Here are a few use cases for this project:

1. Real-Time Game Analysis: The NBA-Player-Detector can be used by coaches or analysts to track player movements, interactions between players, and ball possession in real-time. This could provide valuable insights for decision-making during games and fine-tuning of strategies.

2. Enhanced Sports Broadcasting: Broadcast companies can use the model to automatically detect and highlight players on the screen during a live broadcast. It can help viewers follow the game more closely, especially in identifying less known players, and enhance the overall viewing experience.

3. Player Training and Evaluation: The NBA Player Detector can be used to analyze the performance of individual players during training sessions or competitive games. It could help trainers identify areas where a player could use improvement, such as shooting or passing skills.

4. Sports Betting and Predictions: Bettors or prediction companies can use real-time or historical data from the model to predict player or team performance. Such insights may influence betting odds or decision-making in fantasy sports.

5. Fan Engagement and Interaction: Sports apps can integrate the computer vision model for interactive features, such as allowing fans to click on a player during a live game stream to view their statistics or history. This could significantly enhance fan engagement and satisfaction.

Dataset for: Friehs, M. A., Dechant, M., Vedress, S., Frings, C., & Mandryk, R. L. (2021). Shocking advantage! Improving digital game performance using non-invasive brain stimulation. International Journal of Human-Computer Studies, 148. https://doi.org/10.1016/j.ijhcs.2020.102582 Note that this dataset is already filtered; i.e. all participants that provided faulty data are already excluded and only the final sample is in the data. As digital gaming has grown from a leisure activity into a competitive endeavor with college scholarships, celebrity, and large prize pools at stake, players search for ways to enhance their performance, including through coaching, training, and employing tools that yield a performance advantage. Transcranial direct current stimulation (tDCS) is a non-invasive brain stimulation technique that is presently being explored by esports athletes and competitive gamers. Although shown to modulate cognitive processing in standard laboratory tasks, there is little scientific evidence that tDCS improves performance in digital games, which are visually complex and attentionally demanding environments. We applied tDCS between two sessions of the Stop-Signal Game (SSG; Friehs, Dechant, Vedress, Frings, & Mandryk, 2020). The SSG is a custom-built infinite runner that is based on the Stop-Signal Task (SST; Verbruggen et al., 2019). Consequently, the SSG can be used to evaluate response inhibition as measured by Stop-Signal Reaction Time (SSRT), but in an enjoyable 3D game experience. We used anodal, offline tDCS to stimulate the right dorsolateral prefrontal cortex (rDLPFC); a 9 cm² anode was always positioned over the rDLPFC while the 35 cm² cathode was placed over the left deltoid. We hypothesized that anodal tDCS would enhance neural processing (as measured by a decrease in SSRT) and improve performance, while sham stimulation (i.e., the control condition with a faked stimulation) should lead to no significant change. In a sample of N = 45 healthy adults a significant session x tDCS-condition interaction emerged in the expected direction. Subsequent analysis confirmed that the statistically significant decrease in SSRT after anodal tDCS to the rDLPFC was not due to a general change in reaction times. These results provide initial evidence that tDCS can influence performance in digital games.

The purpose of this project is to determine how college students distribute their activities in time (with a particular focus on academic and athletic activities) and to examine the factors that influence such distributions.Each R reported once about each of the seven days of the week and an additional time about either Saturday or Sunday. Rs were told the week before they were to report which day was assigned and were given a report form to complete during that day. They entered the i nformation from that form when they returned the next week.The activity codes included were: 0: Sleeping. 1: Attending classes. 2: Studying or preparing classroom assignments. 3: Working at a jog (including CAPS). 4: Cooking, home chores, laundry, grocery shopping. 5: Errands, non-grocery shopping, gardening, animal care. 6: Eating. 7: Bathing, getting dressed, etc. 8: Sports, exercising, other physical activities. 9: Playing competitive games (cards, darts, videogames, frisbee, chess, Tr ivial Pursuit, etc.). 10: Participating in UNC-sponsored organizations (student government, band, sorority, etc.). 11: Listening to the radio. 12: Watching TV. 13: Reading for pleasure (not studying or reading for class). 14: Going to a movie. 15: Attending a cultural event (such as a play, concert, or museum). 16: Attending a sports event as a spectator. 17: Partying. 18: Religious activities. 19: Conversation. 20: Travel. 21: Resting. 22: Doing other things DIARY1-8: These datasets contain a matrix of activities by times for a particular day. Included is time period, activity code (see above), # of friends present, # of others present. (Rs were allowed to report doing two activities at once. In these cases they were also asked to report the % of time during the time period affected which was allocated to the first of the two activities listed.)THE DIARY DATASETS ARE STORED IN RAW FORM. SUMMARY FILES, CALLED TIMEREP, CONTAIN MOST SUMMA RY INFORMATION WHICH MIGHT BE USED IN ANALYSES. THE DIARY DATASETS CAN BE LISTED TO ALLOW UNIQUE CODING OF THE ORIGINAL DATA. Each R reported once about each of the seven days of the week and an additional time about either Saturday or Sunday.TIMEREP: The TIMEREP dataset is a summary file which gives the amount of time spent on each activity during each of the eight reporting periods and also includes more detailed information about many of the activities from follow-up questions which were asked if the respondent reported having engaged in certain activities. Data from additional questions asked of every respondent after each diary entry are also included: contact with family members, number of alcoholic drinks consumed during the 24 hour period reported on, number of friends and others present while drinking, number of cigarettes smoked on day reported about, and number of classes skipped on day reported about. Follow-up questions include detail about kind of physical activity or sports participation, kind of university organization, kind of radio program listened to and place of listening, kind of TV program watched and place of watching, kind of reading material read and topic, alcohol consumed while partying and place of partying, conversation topics, kind of travel, activities included in 'other' category.Special processing is required to put the dataset into SAS format. See spec for details.

The General Directorate of Sport of the City of Madrid, together with the 21 districts, convenes and organizes the Municipal Sports Games, as part of the Municipal Competitions program. The Municipal Sports Games is a competition that takes place between the months of September and June, with an expert and careful organization and simple rules. The Municipal Sports Games are the most accessible competitive sports offer for the amateur athlete from Madrid, especially aimed at schools and sports clubs with athletes of school age, both for the affordable prices (free in the case of the 16 individual sports convened) and for the simplicity of the procedures for the incorporation of new technologies. The Municipal Sports Games are one of the largest competitions in Europe, with about 120,000 participants, 70,000 of them of school age. The celebration of more than 55,000 meetings of the 9 collective sports and almost 100 days of individual sports, place them as one of the largest sporting events in our city. The City Council of Madrid, through the General Directorate of Sports and Districts, organizes the Municipal Sports Games. In this dataset, you can get the information of participants enrolled in both Collective and Individual Games of previous seasons. Other related datasets are available on this portal: Sports. Registration of participants in collective and individual Municipal Sports Games. Ongoing season Sports. Municipal sports competitions of collective sports. Ongoing season Sports. Municipal sports competitions of collective sports. Previous seasons Municipal Sports Centers (Polideportivos ) Municipal Basic Sports Facilities

Here are a few use cases for this project:

1. Video Game Development: Developers of games similar to "Back4Blood" can use the model to improve their enemy or character AI. The AI in the game can make decisions based on the location and status of characters' heads and bodies.

2. Gaming Tutorials and Strategies: The model could scan gameplay videos to identify where a player is aiming or what character they are attacking. It can then generate advice, feedback, or strategy guides based on this data.

3. Esports Analytics: The "back4blood" model can be implemented in analytics tools for competitive esports teams. It can identify team strategy or competitor habits by analyzing character positions, their heads, and bodies.

4. Game Streaming Enhancement: The model can be used to introduce new overlay features in game streaming platforms (like Twitch or YouTube) such as real-time analysis of player performance and providing gaming tips based on the identification of the head and body of characters.

5. Content Moderation: Incorporate the model into a system to scan and moderate user-generated content in games, identifying inappropriate use, glitches, or potential game exploits based on abnormal head or body identifiers.

Context

Chess has been a game that I've loved since I started when I was 7 (2nd grade). Since then, I've achieved a peak USCF rating of 1778 (albeit I've been quite inactive in competitive play), and a 2000+ Bullet/Blitz rating on Lichess. Now, I've wanted to synthesize my love for chess with my love for coding and sharing into this dataset, which hopefully can help & inspire others who are interested in both chess and coding like me!

Content

games (folder): The primary content in this dataset is the almost million chess games loaded from chessgames.com. Each csv file represents 1000 games, with the ending number being the games' starting ID. For example, if we have the file ../master_games1012000.csv, then this file will have all the games from ID 1012000 to ID 1012000 + 1000 = 1013000.

ratings (folder): This is a folder with Chess ELOs dating back to the late 19th century compiled by Jeff Sonas on his excellent chessmetrics website. Only ELOs up to the year 2005 are included in his dataset though, so any year after must be supplemented by ratings found in the primary data. Additionally, for this folder, the ending of each csv file represents the games' ending ID. For example, if we have the file ../ratings100000.csv, then this file will have ratings from ID ...90000... to ...100000....

openings (csv): Supplementary file with the relation ECO -> Opening Name.

Acknowledgements

All of the data used was scraped using code which can be found in my github repository for this project. There are in-depth descriptions for each of the files in the repo, so feel free to fork the code and play around with it however you would like. If you like my work, feel free to drop a follow as well :)

Side Note

Not every chess game on chessgames.com was able to be scraped (approximately 100-300 games not scraped). Be careful of that when analyzing specific chess games.

Real-world agents, humans as well as animals, observe each other during interactions and choose their own actions taking the partners’ ongoing behaviour into account. Yet, classical game theory assumes that players act either strictly sequentially or strictly simultaneously without knowing each other’s current choices. To account for action visibility and provide a more realistic model of interactions under time constraints, we introduce a new game-theoretic setting called transparent games, where each player has a certain probability of observing the partner’s choice before deciding on its own action. By means of evolutionary simulations, we demonstrate that even a small probability of seeing the partner’s choice before one’s own decision substantially changes the evolutionary successful strategies. Action visibility enhances cooperation in an iterated coordination game, but reduces cooperation in a more competitive iterated Prisoner’s Dilemma. In both games, “Win–stay, lose–shift” and “Tit-for-tat” strategies are predominant for moderate transparency, while a “Leader-Follower” strategy emerges for high transparency. Our results have implications for studies of human and animal social behaviour, especially for the analysis of dyadic and group interactions.

Quantitative response data from a series of experiments on responsibility attributionThis project aims to identify the principles that underlie people's attributions of cause, responsibility and blame in social contexts. This is a complex problem, because people's judgments are responsive to many factors, including the intentions and foreknowledge of the agents involved, and subtle interactions between the different parties. As a guiding framework the project will test and develop a structural model approach to attribution, based upon notions of counterfactual dependence and active causal pathways between events. This approach maintains that assignments of credit and blame are driven by people's causal models, but allows for dissociations between what is judged a cause and what is singled out for blame. The project will use experimental studies to test the applicability of the structural model in everyday human attributions. These studies will explore a variety of contexts, including interactive games with multiple players. The complex interactions between players in competitive games present a novel and rich environment to investigate people's attributions. In addition, the project will investigate how people construct and communicate the causal models that underpin their attributions. These studies will examine how people transform scenarios into causal models, and how these models affect their subsequent judgments of cause and blame.

Esports Analytics AI Market Outlook

According to our latest research, the global Esports Analytics AI market size reached USD 1.37 billion in 2024, demonstrating robust momentum driven by the digital transformation of the esports industry. The market is projected to grow at a CAGR of 23.4% during the forecast period, reaching USD 10.39 billion by 2033. This remarkable growth is primarily fueled by increasing investments in esports infrastructure, surging demand for data-driven decision-making, and the rapid adoption of AI-powered analytics to enhance player performance, fan engagement, and commercial opportunities across the esports ecosystem.

One of the most significant growth factors for the Esports Analytics AI market is the increasing sophistication and competitiveness of the esports industry itself. As esports tournaments draw larger global audiences and substantial sponsorship deals, stakeholders are under mounting pressure to optimize performance and maximize commercial returns. AI-driven analytics tools are enabling teams, coaches, and analysts to dissect in-game data, scrutinize player behavior, and develop winning strategies with unprecedented precision. The ability to process vast datasets in real-time offers a competitive edge, allowing for rapid adaptation to evolving gameplay dynamics and opponent tactics. This analytical depth not only improves individual and team performance but also elevates the overall standard of competition, accelerating the adoption of AI solutions within the market.

Another critical driver is the escalating demand for enhanced fan engagement and immersive viewer experiences. Esports audiences are highly digital-savvy and expect interactive, data-rich content that deepens their connection to the games and players they follow. AI-powered analytics platforms are revolutionizing the way fans interact with esports by providing real-time statistics, predictive insights, and personalized content through various digital channels. These innovations are fostering stronger fan loyalty, increasing viewership, and opening new avenues for monetization through targeted sponsorships and advertising. The integration of AI analytics into broadcasting and streaming platforms is also enabling broadcasters and tournament organizers to deliver more engaging narratives, further propelling market growth.

Additionally, the proliferation of esports tournaments and leagues worldwide is creating a fertile environment for the expansion of the Esports Analytics AI market. With the formalization of esports as a professional sport in many regions, there is a growing emphasis on transparency, fairness, and regulatory compliance. AI-driven analytics solutions are being leveraged to monitor gameplay integrity, detect potential cheating, and ensure adherence to tournament rules. This not only safeguards the credibility of esports competitions but also instills greater confidence among sponsors, investors, and fans. The convergence of technology, regulation, and commercial interests is expected to sustain the momentum of the Esports Analytics AI market throughout the forecast period.

From a regional perspective, Asia Pacific continues to dominate the Esports Analytics AI market, accounting for over 38% of the global revenue in 2024. The region's leadership is underpinned by a vibrant esports culture, substantial investments from both public and private sectors, and a massive base of active gamers and spectators. North America and Europe are also witnessing significant growth, driven by technological advancements, high consumer spending, and the presence of major esports organizations and tournaments. Meanwhile, emerging markets in Latin America and the Middle East & Africa are gradually catching up, presenting lucrative opportunities for market players seeking to tap into new demographics and expand their global footprint.

Component Analysis

The Esports Analytics AI market is segmented by component into software and services, each playing a pivotal role in the ecosystem’s development and value delivery. The software segment constitutes the backbone of this market, encompassing advanced AI-powered platforms designed for data aggregation, real-time analytics, visualization, and predictive modeling. These software solutions are tailored to address the unique demands of esports, such as processing in-game telemetry, player biometrics, and audience engagement metrics. Over the past year, software a

Many species exhibit two discrete male morphs: fighters and sneakers. Fighters are large and possess weapons but may mature slowly. Sneakers are small and have no weapons but can sneak matings and may mature quickly to start mating earlier in life than fighters. However, how differences in competitive ability and life history interact to determine male morph coexistence has not yet been investigated within a single framework. Here we integrate demography and game theory into a two-sex population model to study the evolution of strategies that result in the coexistence of fighters and sneakers. We incorporate differences in maturation time between the morphs and use a mating-probability matrix analogous to the classic hawk-dove game. Using adaptive dynamics, we show that male dimorphism evolves more easily in our model than in classic game theory approaches. Our results also revealed an interaction between life-history differences and sneaker competitiveness, which shows that demography ...

This data set provides data from previous seasons of collective sports competitions organized by the City of Madrid through the General Directorate of Sports and Districts. There are three types of municipal sports competitions: Municipal Sports Games, for all ages Municipal Sports Tournaments for senior categories Spring Trophies, for base categories (data of this type of competition are not included in this data set) These competitions take place between the months of September to June, with an expert and careful organization and simple rules. They are the most accessible competitive sports offer for the amateur athlete from Madrid, especially aimed at schools and sports clubs with athletes of school age, both for the affordable prices (free in the case of the 16 individual sports convened) and for the simplicity of the procedures for the incorporation of new technologies. The Municipal Sports Games are one of the largest competitions in Europe, with about 120,000 participants, 70,000 of them of school age. The celebration of more than 55,000 meetings of the 9 collective sports and almost 100 days of individual sports, place them as one of the largest sporting events in our city. In this same portal there are other complementary data sets such as: Sports. Municipal sports competitions of collective sports. Current season Municipal Sports Centers (Polideportivos ) Municipal Basic Sports Facilities In 'Associated Documentation' the explanatory documents of the data structure of the files are offered.

Here are a few use cases for this project:

1. Automated Chess Games Analysis: This model can be used to input and analyze games from physical chess boards in real-time or from photographs. It can transform these games into digital versions for further analysis, suggestions, or tracking player performance.

2. Developing Intelligent Chess Bots: The "Chess" model can aid in developing AI-based chess software/bots, which can recognize the state of the game by visually analyzing the chess board and decide the next move.

3. Assistive Technology for the Visually Impaired: It can be applied for developing applications assisting visually impaired people to play chess, by recognizing chess board state and using speech to tell the state or the possible moves.

4. Chess Learning and Coaching Apps: The model can be incorporated into educational applications that teach users how to play chess, understand chess movements, or improve their strategies. The app could provide real-time recommendations by recognizing the position of pieces on the board.

5. Cheat Detection in Competitive Chess: In competitive or online chess playing platforms, the model can be used to monitor games and detect anomalous activities or cheating attempts, such as piece position changes in a non-standard manner.

This dataset mainly provides information on the participation and performance of athletes in the Olympic Games over the years.

Games on Stadia: A Comprehensive List

The More You Play, the More You Achieve!

About this dataset

Do you like playing video games? Do you like achieving things? If you answered yes to both of those questions, then this dataset is for you!

The goal of this data is to provide a comprehensive list of all the games currently available on Stadia, as well as some basic information about each game. This dataset includes titles, genres, developers, publishers, Stadia release dates, original release dates, and more.

With this information at your fingertips, you can plan your gaming schedule around which games you want to achieve in and when they'll no longer be available for free on Stadia Pro. So what are you waiting for? Get achievement-hunting!

How to use the dataset

In order to use this dataset, simply download it and open it in your preferred spreadsheet application. From there, you can begin to explore the data and answer any questions you may have about the contents of each column.

Columns: 0: The name of the game. (String) 1: The type of product. (String) 2: The genre or genres that the game belongs to. (String) 3: The developer or developers of the game. (String) 4: The publisher or publishers of the game. (String) 5: The date that the game was released on Stadia. (Date) 6: The date that the game was originally released. (Date) 7:The date that the game was added to Stadia Pro. (Date) 8:The date that the game is no longer claimable on Stadia Pro. (Date)

Research Ideas

• Gamers could use this dataset to find new games to play on Stadia, based on their preferred genres, developers, publishers, etc.
• Game developers and publishers could use this dataset to track the success of their titles on Stadia, and compare release dates and player engagement across different platforms.
• Researchers could use this dataset to study player behavior on Stadia, and how different variables like genre or release date affect player engagement and satisfaction

License

License: CC0 1.0 Universal (CC0 1.0) - Public Domain Dedication No Copyright - You can copy, modify, distribute and perform the work, even for commercial purposes, all without asking permission. See Other Information.

Columns

File: df_16.csv | Column name | Description | |:--------------|:--------------| | 0 | | | 1 | |

File: df_20.csv | Column name | Description | |:--------------|:--------------| | 0 | | | 1 | |

File: df_18.csv | Column name | Description | |:--------------|:--------------| | 0 | | | 1 | |

File: df_11.csv | Column name | Description | |:--------------|:--------------| | 0 | | | 1 | |

File: df_1.csv | Column name | Description | |:----------------------------------------------|:---------------------------------------------------------------| | Title | The name of the game. (String) | | Genre(s) | The genre or genres of the game. (String) | | Developer(s) | The developer or developers of the game. (String) | | Publisher(s) | The publisher or publishers of the game. (String) | | Stadia release date | The date the game was released on Stadia. (Date) | | Original release date[a] | The date the game was originally released. (Date) | | Date added to Stadia Pro[b] | The date the game was added to Stadia Pro. (Date) | | Date no longer claimable on Stadia Pro[c] | The date the game is no longer claimable on Stadia Pro. (Date) | | Ref. | A reference to where the information was found. (String) |

File: df_4.csv | Column name | Description | |:--------------|:--------------| | 0 | | | 1 | |

File: df_21.csv

File: df_17.csv | Column name | Description | |:---------------|:------------------------------------------------| | Hardware | The hardware the game is available on. (String) | | Hardware.1 | The hardware the game is available on. (String) |

File: df_9.csv | Column name | Description | |:--------------|:-------------...

Counter Strike 2 Cheat Detection Context Windows - Length 256

  Overview

The Counter Strike 2 Cheat Detection Context Windows - Length 256 (Context_window_256) is a dataset conprised of "context windows". This dataset was created using extracted data from the CS2CD dataset. For more information regarding the data see AntiCheatPT: A Transformer-Based Approach to Cheat Detection In Competitive Computer Games by Mille Mei Zhen Loo and Gert Lužkov 1/6/2025.… See the full description on the dataset page: https://huggingface.co/datasets/CS2CD/Context_window_256.

Kaggle’s March Machine Learning Mania competition challenged data scientists to predict winners and losers of the men's 2016 NCAA basketball tournament. This dataset contains the 1070 selected predictions of all Kaggle participants. These predictions were collected and locked in prior to the start of the tournament.

How can this data be used? You can pivot it to look at both Kaggle and NCAA teams alike. You can look at who will win games, which games will be close, which games are hardest to forecast, or which Kaggle teams are gambling vs. sticking to the data.

The NCAA tournament is a single-elimination tournament that begins with 68 teams. There are four games, usually called the “play-in round,” before the traditional bracket action starts. Due to competition timing, these games are included in the prediction files but should not be used in analysis, as it’s possible that the prediction was submitted after the play-in round games were over.

Data Description

Each Kaggle team could submit up to two prediction files. The prediction files in the dataset are in the 'predictions' folder and named according to:

TeamName_TeamId_SubmissionId.csv

The file format contains a probability prediction for every possible game between the 68 teams. This is necessary to cover every possible tournament outcome. Each team has a unique numerical Id (given in Teams.csv). Each game has a unique Id column created by concatenating the year and the two team Ids. The format is the following:

Id,Pred
2016_1112_1114,0.6
2016_1112_1122,0
...

The team with the lower numerical Id is always listed first. “Pred” represents the probability that the team with the lower Id beats the team with the higher Id. For example, "2016_1112_1114,0.6" indicates team 1112 has a 0.6 probability of beating team 1114.

For convenience, we have included the data files from the 2016 March Mania competition dataset in the Scripts environment (you may find TourneySlots.csv and TourneySeeds.csv useful for determining matchups, see the documentation). However, the focus of this dataset is on Kagglers' predictions.

Context

The Interactive Fiction Competition is the internet's oldest game programming competition, starting in 1995 and continuing to this day. Interactive Fiction is an unusual genre because it has been centralized in the Interactive Fiction Database for the last decade, and essentially all games of interest are recorded there. It is also a niche genre, and the amount of public interest has varied much less than other genres since 1995.

Content

This database contains the placement (i.e. 1st place, 2nd place, etc.) of all games ever entered into the IFComp, together with their genre, number of ifdb ratings and average ifdb rating. It also contains the system; a few programming languages have tended to dominate the competition. It also contains the forgiveness rating, which indicates how difficult the game can be expected to be. Much of the information is null.

Note that the IFDB was created in 2006, 11 years after the competition began.

Inspiration

IFDB ratings represent lasting interest, as these have been gathered over many years, while IFComp rankings represent a short, 6-week period.

It would be interesting to see the relationship between the two, how that relationship has changed over time, etc.

Also, Inform and TADS have always dominated the competition, but it's recently been changing. The change in systems over the years should be interesting.

Finally, it would be interesting to see time-adjusted averages of reviews for comp games over time, to see if new games are receiving less interest than old games.

Acknowledgements

Mike Roberts has been the curator of IFDB since its inception. The IFComp is currently run by Jason Macintosh under the aegis of the Interactive Fiction Technology Foundation.