Irish Dance Data, The Value of "Leap, Love"
Whenever I’m in a situation where I have to share a “fun fact,” my typical answer is that I used to be a competitive Irish dancer. When I shared this fun fact during my introduction at the NYC Data Science bootcamp, I reminisced about the curly wigs, the lively tunes, and the voice of my former master dance teacher screaming, “LEAP, love!”
Two weeks later, I turned my fun fact into my fun first bootcamp project. I thought, “What can I do with Irish dancing data? What can this industry gain from my analysis?” Many times the performing arts and creative industries are a bit overlooked in the data science world, and having background knowledge about competitive Irish dancing, I thought I could possibly extract something of value. I decided to scrape www.feisresults.com, which posts results from top Irish dancing competitions around the world. Using the Python package Scrapy, I collected the data from every listed competition.
There are some limitations to my scraped dataset to keep in mind. These results show the rank of competitors who placed (usually the top fifty) and not everyone who competed. Also, not every competition’s results are posted for each year. For example, the results for the All Irelands Championship in 2017 were missing, and no competitions from Australia or New Zealand were listed at all. Nevertheless, the Word Championships are, no doubt, the most prestigious, and the results were available for the past 11 consecutive years, so I focused on analyzing this subset for this project.
What did I find?
As I’ll demonstrate in this post, by analyzing the World Championships results from 2008 to 2018, I am able to:
- provide visualizations to describe the geographic distribution of the elite Irish dance market.
- graphically represent school’s strengths and weaknesses.
- identify marketing opportunities for vendors and dance schools.
- propose a method of tracking individual success over time, which can be used by stakeholders to better achieve their goals and consequently make results a little easier to digest for the consumer.
This figure outlines my thought process for this project. First, I thought about entities who are financially connected to the World Championships each year. Second, I thought about what their goals might be, such as establishing brand awareness for vendors or attracting new dancers for dance schools. Lastly, I thought about how these data, which are primarily success or geographic measures, could be used in ways to achieve those goals. For example, vendors could use these geographic measures to catch a wider proportion of the existing market and identify emerging markets around the globe. Alternatively, performance professionals can use success measures to easily identify top talent for their next touring season. The value lies where these categories intersect.
Therefore, my goal for this project was to create insights for stakeholders by answering these questions:
- Who are the top schools and dancers?
- What is the geographic distribution of the elite Irish dancing market?
Analysis and Discussion
In reference to geography, I used the plotly package to make a choropleth map of the competitors who placed in the last ten years, shown below. Unsurprisingly, the majority of competitors come from Ireland, USA, and the UK. However, it's interesting to note that there are competitors from Brazil, The Netherlands, Poland, Mexico, Russia, and others.
To make their products more accessible to a broader audience, vendors could offer their website in multiple languages. Additionally, they could compare how the proportion of dancers change per year to look for emerging markets of elite dancers. Since elite female dancers buy new dresses once or twice each year (which cost well over $1,000 each), catering to this growing subset of the market can become fruitful with repeated business.
The other category of useful data relies on measuring success, which is not as trivial as it sounds. As seen below, solo and team competitions are separated into specific categories based on age and gender.
The winners of each category receive a first place prize, resulting in over twenty first-place solo winners and nearly the same number of team champions for each year. Placing in any of these competitions is noteworthy. However, if you’re trying to hire the next lead dancer for Riverdance or trying to decide which school to enroll your dance-obsessed daughter in, it’s not easy to compare using singular wins as a metric.
Insider tip: Once students enroll in classes under any particular Irish dance school, they almost always compete under that name for the rest of their dance career. Switching schools does occur, but it is very rare due to the tribe-like intensity of the competitive ecosystem. Therefore, it is imperative for schools to attract dancers early on or to expand their enrollment through personal referrals.
The Carey Academy is notorious for their team dance wins. Below is a bar graph that tallies the total number of podium wins over the past eleven years.
This method can be improved by using a heatmap, seen above, showing which schools did well in each team dance category. Here the heatmap shows that The Carey Academy has done particularly well in junior team dance competitions, while The Claddagh Dance Company has done well in senior ladies competitions. Dance schools can use this information as selling points to attract dancers for their classes and summer workshops.
When considering solo wins, bar charts are less effective for showing cumulative success. A first-place medal is equivalent to a third-place medal on this graph. This is why Jordan McCormick is listed as second even though he has only won first place once, compared to Jack Quinn’s seven first-place wins. Furthermore, bar charts do not provide much insight regarding time series data. Notice that there is a nine-way tie for 5th place when ranked from top to bottom. To solve these issues, I have used weighted ranks and exponential decay as described in the next section.
Cumulative Irish Points with Exponential Decay Method
That is why I propose another approach: the use of “Irish points.” The concept of Irish points is already used during championship-level competitions which help normalize the variance in raw scoring between judges.
To give a bit more background: When soloists compete at Worlds there are three rounds. Each round, dancers receive a raw score from multiple judges. Dancers are awarded points based on each judge's ranking; a first place vote receives 100 Irish points, second place receives 75, et cetera. After the first two rounds, competitors above the 50th percentile advance to the third and final round.
Since the concept of Irish points is accepted by The Irish Dancing Commission as a way to rank dancers across judges, it seems natural to use a similar method to informally rank dancers. If applicable, this provides a simple way to quantify success over time across age and gender categories.
In addition to Irish points addressing the question on longitudinal success, an exponential decay function can be applied to each year’s Irish points to address the issue of relevance. For example, the Irish points of a dancer who had won 12th place in 2018 would be worth more than those of a dancer who won 12th place in 2014. For this project, the decay constant, lambda, was set to 0.2 arbitrarily so that the half-life of Irish points was around three years.
Implementing the Irish Points Method
The concept of ranking athletes changes considerably when accounting for depreciation over time. Those who have not competed in the last few years dropped in ranking. For example, Jack Quinn goes from 3rd to 17th due to the fact that he hasn’t competed at Worlds since 2016 and has been touring with Riverdance and Heartbeat of Home as a professional dancer.
Measuring Success Over Time
The 2014 to 2018 time series graph below shows the trends of the top-ten dancers using the Irish points method with depreciation. One could note, Shannon Bradley continued to rise above the others while Michaela Hinds dropped in the rankings after a year of not competing.
These data may be useful for dance companies to scout dancers at the top of their game for their next touring season. Additionally, vendors may gain insight to identify talent for sponsorship or brand association. For example, Alliyah O’Hare seems to be the dancer to watch, having won every World Championship she has entered starting at the youngest available age category. As an Irish dance social influencer, she has marketing potential to reach a wide audience of aspiring Irish dancers for the next five to eight years.
From a broad perspective, concentrating the leaderboard gives enthusiasts a way to promote the sport in a more digestible way. In other words, it is easier to keep track of the top ten overall rather than the leaders of the numerous categories of solo competitions. Sports, including tennis and golf, rank professionals based on previous performance. In fact, given that the World Championships have adopted a double recall system for large categories starting in 2019, the Irish point method can be used to ensure a more even split than the method proposed.
The Value Summarized
Discussed in this post, this dataset can be valuable to stakeholders in a variety of ways. Vendors can use geographic data to track the market and identify potentially neglected areas. They can also use success measure trends to target potential influencers that they can use for online marketing. Schools may also use success measures to help promote their classes or workshops, solidifying school loyalty before it's too late. Similarly, performance professionals looking to hire, can use the Irish points method to easily see who are the best dancers across competition categories over a period of time. The Irish Dance Commission may even consider this approach for the newly adopted double recall system of competition. Personally, the intrinsic value of this data goes beyond the metrics I have collected. I have witnessed the dedication of these competitors and experienced the thrill of the competition myself.
To continue this project further, I would scrape a compete dataset from all World qualifying competitions, where available. This would provide data required to answer questions about the distribution of world qualifiers and the proportion of those who place at Worlds. In addition, I would add a column in my dataset for birth year and create a Shiny web app for individuals to adjust these graphs according to variables including age, location, and school. Lastly, this dataset could be reanalyzed in regards to more specific geographic location.
If you'd like to "LEAP" into my web scraping code or the jupyter notebook file of my exploratory analysis, check out my github repository.