diff --git a/config.yaml b/config.yaml index ffe0e2a659..7debcfe58c 100644 --- a/config.yaml +++ b/config.yaml @@ -45,10 +45,11 @@ params: alttext: Two orbs orbiting each other. They are displacing gravity around them. url: /case-studies/gw-discov - title: Sports Analytics - text: TODO! + text: Cricket Analytics is changing the game by improving player and team performance through statistical modelling and predictive analytics. NumPy enables many of these analyses. img: images/content_images/case_studies/sports.jpg alttext: Cricket ball on green field. - url: / + url: /case-studies/cricket-analytics + section2: title: KEY FEATURES features: diff --git a/content/en/case-studies/blackhole-image.md b/content/en/case-studies/blackhole-image.md index 496293d2c7..a3be74fe69 100644 --- a/content/en/case-studies/blackhole-image.md +++ b/content/en/case-studies/blackhole-image.md @@ -3,7 +3,7 @@ title: "Case Study: The First Image of a Black Hole" sidebar: false --- -{{< figure src="/images/content_images/cs/blackhole.jpg" caption="**Black Hole M87**" alt="black hole image" attr="(**Image Credits:** Event Horizon Telescope Collaboration)" attrlink="https://www.jpl.nasa.gov/images/universe/20190410/blackhole20190410.jpg" >}} +{{< figure src="/images/content_images/cs/blackhole.jpg" caption="**Black Hole M87**" alt="black hole image" attr="*(Image Credits: Event Horizon Telescope Collaboration)*" attrlink="https://www.jpl.nasa.gov/images/universe/20190410/blackhole20190410.jpg" >}}

Imaging the M87 Black Hole is like trying to see something that is by definition impossible to see.

diff --git a/content/en/case-studies/cricket-analytics.md b/content/en/case-studies/cricket-analytics.md new file mode 100644 index 0000000000..eafc72e8da --- /dev/null +++ b/content/en/case-studies/cricket-analytics.md @@ -0,0 +1,158 @@ +--- +title: "Case Study: Cricket Analytics, the game changer!" +sidebar: false +--- + +{{< figure src="/images/content_images/cs/ipl-stadium.png" + caption="**IPLT20, the biggest Cricket Festival in India**" + alt="Indian Premier League Cricket cup and stadium" + attr="*(Image credits: IPLT20 (cup and logo) & Akash Yadav (stadium))*" + attrlink="https://unsplash.com/@aksh1802" >}} + +
+

You don't play for the crowd, you play for the country.

+
—M S Dhoni, International Cricket Player, ex-captain, Indian Team, plays for Chennai Super Kings in IPL
+
+ +## About Cricket + +It would be an understatement to state that Indians love cricket. The game is +played in just about every nook and cranny of India, rural or urban, popular +with the young and the old alike, connecting billions in India unlike any other sport. +Cricket enjoys lots of media attention. There is a significant amount of +[money](https://www.statista.com/topics/4543/indian-premier-league-ipl/) and +fame at stake. Over the last several years, technology has literally been a game +changer. Audiences are spoilt for choice with streaming media, tournaments, +affordable access to mobile based live cricket watching, and more. + +The Indian Premier League (IPL) is a professional Twenty20 cricket +league, founded in 2008. It is one of the most attended cricketing events in +the world, valued at [$6.7 billion](https://en.wikipedia.org/wiki/Indian_Premier_League) +in 2019. + +Cricket is a game of numbers - the runs scored by a batsman, the wickets taken +by a bowler, the matches won by a cricket team, the number of times a batsman +responds in a certain way to a kind of bowling attack, etc. The capability to +dig into cricketing numbers for both improving performance and studying +the business opportunities, overall market and economics of cricket via powerful +analytics tools, powered by numerical computing software such as NumPy, is a big +deal. Cricket analytics provides interesting insights into the game and +predictive intelligence regarding game outcomes. + +Today, there are rich and almost infinite troves of cricket game records and +statistics available, e.g., [ESPN +cricinfo](https://stats.espncricinfo.com/ci/engine/stats/index.html) and +[cricsheet](https://cricsheet.org). These and several such cricket databases +have been used for [cricket +analysis](https://www.researchgate.net/publication/336886516_Data_visualization_and_toss_related_analysis_of_IPL_teams_and_batsmen_performances) +using the latest machine learning and predictive modelling algorithms. +Media and entertainment platforms along with professional sports bodies +associated with the game use technology and analytics for determining key +metrics for improving match winning chances: + +* batting performance moving average, +* score forecasting, +* gaining insights into fitness and performance of a player against different opposition, +* player contribution to wins and losses for making strategic decisions on team composition + +{{< figure src="/images/content_images/cs/cricket-pitch.png" + class="csfigcaption" + caption="**Cricket Pitch, the focal point in the field**" + alt="A cricket pitch with bowler and batsmen" + align="middle" + attr="*(Image Credits: Debarghya Das)*" + attrlink="http://debarghyadas.com/files/IPLpaper.pdf" >}} + +### Key Data Analytics Objectives + +* Sports data analytics are used not only in cricket but many [other + sports](https://adtmag.com/blogs/dev-watch/2017/07/sports-analytics.aspx) for + improving the overall team performance and maximize winning chances. +* Real-time data analytics can help in gaining insights even during the game + for changing tactics by the team and by associated businesses for economic + benefits and growth. +* Besides historical analysis, predictive models are + harnessed to determine the possible match outcomes that require significant + number crunching and data science know-how, visualization tools and capability + to include newer observations in the analysis. + +{{< figure src="/images/content_images/cs/player-pose-estimator.png" + class="fig-center" + alt="pose estimator" + caption="**Cricket Pose Estimator**" + attr="*(Image Credits: connect.vin)*" + attrlink="https://connect.vin/2019/05/ai-for-cricket-batsman-pose-analysis/" >}} + +### The Challenges + +* **Data Cleaning and preprocessing** + + IPL has expanded cricket beyond the classic test match format to a much + larger scale. The number of matches played every season across various + formats has increased and so has the data, the algorithms, newer sports data + analysis technologies and simulation models. Cricket data analysis requires + field mapping, player tracking, ball tracking, player shot analysis and + several other aspects involved in how the ball is delivered, its angle, spin, + velocity and trajectory. All these factors together have increased the + complexity of data cleaning and preprocessing. + +* **Dynamic Modeling** + + In cricket, just like any other sport, + there can be a large number of variables related to tracking various numbers + of players on the field, their attributes, the ball and several possibilities + of potential actions. The complexity of data analytics and modeling is + directly proportional to the kind of predictive questions that are put forth + during analysis and are highly dependent on data representation and the + model. Things get even more challenging in terms of computation, data + comparisons when dynamic cricket play predictions are sought such as what + would have happened if the batsman had hit the ball at a different angle or + velocity. + +* **Predictive Analytics Complexity** + + Much of the decision making in Cricket is based on questions such as "how + often does a batsman play a certain kind of shot if the ball delivery is of a + particular type", or "how does a bowler change his line and length if the + batsman responds to his delivery in a certain way". + This kind of predictive analytics query requires highly granular dataset + availability and the capability to synthesize data and create generative + models that are highly accurate. + +## NumPy’s Role in Cricket Analytics + +Sports Analytics is a thriving field. Many researchers and companies +[use NumPy](https://adtmag.com/blogs/dev-watch/2017/07/sports-analytics.aspx) +and other PyData packages like Scikit-learn, SciPy, Matplotlib, and Jupyter. +in addition to latest machine learning and AI techniques. NumPy has been used +for various kinds of cricket related sporting analytics such as: + +* **Statistical Analysis:** NumPy's numerical capabilities help estimate the + statistical significance of observational data or match events in the context + of various player and game tactics, estimating the game outcome by comparison + with a generative or static model. + [Causal analysis](https://amplitude.com/blog/2017/01/19/causation-correlation) + and [big data approaches](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4996805/) + are used for tactical analysis. + +* **Data Visualization:** Data graphing and [visualization](https://towardsdatascience.com/advanced-sports-visualization-with-pandas-matplotlib-and-seaborn-9c16df80a81b) + provides useful insights into relationship between various datasets. + +## Summary + +Sports Analytics have changed the way professional games are played, especially +regarding decision making which was until recently primarily done based on +“gut feeling" or adherence to past traditions. NumPy forms a +solid foundation for a large set of Python packages which provide higher level +functions related to data analytics, machine learning and AI algorithms. These +packages are widely deployed to gain real-time insights that help in decision +making for game-changing outcomes, both on field as well as to draw inferences +and drive business around the game of cricket. Finding out the hidden +parameters, patterns and attributes that lead to the outcome of a cricket match +helps the stakeholders to take notice of game insights that are otherwise hidden +in numbers and statistics. + +{{< figure src="/images/content_images/cs/numpy_ca_benefits.png" + class="fig-center" + alt="Diagram showing benefits of using NumPy for cricket analytics" + caption="**Key NumPy Capabilities utilized**" >}} diff --git a/content/en/case-studies/gw-discov.md b/content/en/case-studies/gw-discov.md index 6d82dfa091..13a831c768 100644 --- a/content/en/case-studies/gw-discov.md +++ b/content/en/case-studies/gw-discov.md @@ -3,7 +3,7 @@ title: "Case Study: Discovery of Gravitational Waves" sidebar: false --- -{{< figure src="/images/content_images/cs/gw_sxs_image.png" class="fig-center" caption="**Gravitational Waves**" alt="binary coalesce black hole generating gravitational waves" attr="(**Image Credits: The Simulating eXtreme Spacetimes (SXS) Project at LIGO** )" attrlink="https://youtu.be/Zt8Z_uzG71o" >}} +{{< figure src="/images/content_images/cs/gw_sxs_image.png" class="fig-center" caption="**Gravitational Waves**" alt="binary coalesce black hole generating gravitational waves" attr="*(Image Credits: The Simulating eXtreme Spacetimes (SXS) Project at LIGO)*" attrlink="https://youtu.be/Zt8Z_uzG71o" >}}

The scientific Python ecosystem is critical infrastructure for the research done at LIGO.

@@ -45,12 +45,13 @@ made from warped spacetime. astrophysics, cosmology, particle physics, and nuclear physics. * Crunch observed data via numerical relativity computations that involves complex maths in order to discern signal from noise, filter out relevant - signal and statistically estimate significance of observed data + signal and statistically estimate significance of observed data * Data visualization so that the binary / numerical results can be comprehended. - -### The Challenges + + +### The Challenges * **Computation** @@ -61,7 +62,7 @@ made from warped spacetime. complex relativity equations and huge amounts of data which present a computational challenge: [O(10^7) CPU hrs needed for binary merger analyses](https://youtu.be/7mcHknWWzNI) - spread on 6 dedicated LIGO clusters + spread on 6 dedicated LIGO clusters * **Data Deluge** @@ -89,7 +90,7 @@ made from warped spacetime. {{< figure src="/images/content_images/cs/gw_strain_amplitude.png" class="fig-center" alt="gravitational waves strain amplitude" caption="**Estimated gravitational-wave strain amplitude from GW150914**" attr="(**Graph Credits:** Observation of Gravitational Waves from a Binary Black Hole Merger, ResearchGate Publication)" attrlink="https://www.researchgate.net/publication/293886905_Observation_of_Gravitational_Waves_from_a_Binary_Black_Hole_Merger" >}} ## NumPy’s Role in the detection of Gravitational Waves - + Gravitational waves emitted from the merger cannot be computed using any technique except brute force numerical relativity using supercomputers. The amount of data LIGO collects is as incomprehensibly large as gravitational @@ -111,13 +112,14 @@ speed. Here are some examples: * Visualization of data - Time series - Spectrograms -* Compute Correlations +* Compute Correlations * Key [Software](https://github.com/lscsoft) developed in GW data analysis such as [GwPy](https://gwpy.github.io/docs/stable/overview.html) and [PyCBC](https://pycbc.org) uses NumPy and AstroPy under the hood for providing object based interfaces to utilities, tools and methods for studying data from gravitational-wave detectors. + ## Summary GW detection has enabled researchers to discover entirely unexpected phenomena diff --git a/static/images/content_images/cs/cricket-pitch.png b/static/images/content_images/cs/cricket-pitch.png new file mode 100644 index 0000000000..7d3d4c20e1 Binary files /dev/null and b/static/images/content_images/cs/cricket-pitch.png differ diff --git a/static/images/content_images/cs/ipl-stadium.png b/static/images/content_images/cs/ipl-stadium.png new file mode 100644 index 0000000000..a54ede3336 Binary files /dev/null and b/static/images/content_images/cs/ipl-stadium.png differ diff --git a/static/images/content_images/cs/numpy_ca_benefits.png b/static/images/content_images/cs/numpy_ca_benefits.png new file mode 100644 index 0000000000..843aa0131e Binary files /dev/null and b/static/images/content_images/cs/numpy_ca_benefits.png differ diff --git a/static/images/content_images/cs/player-pose-estimator.png b/static/images/content_images/cs/player-pose-estimator.png new file mode 100644 index 0000000000..6e0d8b5c2c Binary files /dev/null and b/static/images/content_images/cs/player-pose-estimator.png differ