An Elo rating system for hurling

Introduction

The Elo rating system is a system for estimating the relative strengths of different players or teams in a game. Although it originated in chess, it has been adapted to try to put a number on how much better one team is than another in a variety of sports. Here at Take Your Datapoints and the Goals Will Come, we strive to bring together the best in statistical analysis and armchair postulating on the sport of hurling. As such, creating an Elo rating system for the intercounty teams seemed like a natural starting point. If you just wish to see how we’ve rated your team versus others, feel free to just check out these figures, and then follow us on twitter so that you can tell us to tell us how wrong we are. However, if you’d like to learn more about the system, then read on.

Ratings at the beginning of 2017; even with the All-Ireland loss, years of consistency mean Kilkenny are still the top rated team.

Ratings at the beginning of 2017; even with the All-Ireland loss, years of consistency mean Kilkenny are still the top rated team.

The previous chart, in map form; note that Fingal had acquired a rating of 1315, but have withdrawn from competition in 2017, so are not included in this map

The previous chart, in map form; note that Fingal had acquired a rating of 1315, but have withdrawn from competition in 2017, so are not included in this map

What is the Elo rating?

First of all, what exactly is the Elo rating system? We’re not going to pretend that we’re above using Wikipedia as a source, so here’s the relevant page. Still, the following is my quick explanation: In short, it’s a scoring system in which a team is rewarded more for defeating a higher rated opponent, and punished more for losing to a lower rated opponent. How a team wins or loses in unimportant, as is the frequency of matches; all that matters is the result, and how the teams involved were rated to begin with. Each team is given a starting rating. Two teams with the same rating should have a 50/50 chance of winning when facing off. Following a match, the winning team’s score will rise, and the losing team’s score will decrease by the same amount. In the case of a draw, the team with the lower starting rating will see their score increase. How much the scores increase is a factor of a constant, k, and the relative scores of the teams. As such, a shock underdog victory will result in a much bigger swing than a top tier team beating a weak team. With me so far? Let’s look at some real word examples from our rating system:

Waterford 2014-2016

By the end of 2014, Waterford looked like a team in crisis. In 2016, Munster finals aside, they were a threat to win both the league and the championship. Their remarkable turnaround can be seen as we look over their change in rating over the past two years.

Cork 2012-2016

For an example of a reverse trend, we can look at Cork. In an All-Ireland final replay in 2013, and the Munster champions in 2014, Cork seemed to be re-emerging as one of the top teams in the country, as befits their historical ‘Big-Three’ status. Then, following a shakier 2015 campaign, Cork had, like many of us, a disastrous 2016. Narrowly avoiding relegation to division 1B in the league by winning only their demotion match against Galway, and failing to get past the qualifiers in the All-Ireland, Cork’s dramatic plummet over the last year can be seen clearly in their rating changes.

Westmeath 2014-2016

As a final example, I’d like to draw your attention towards Westmeath. An odd choice, perhaps, but that’s exactly the point. While the previous examples emphasised how the ratings responded to the results, I’ve selected Westmeath as an example of the ratings showing us something we might not have noticed. It’s rare that anybody mentions any of the non-traditional hurling counties, or really anything that’s going on in the lower divisions. However, the numbers don’t discriminate, and can occasionally reveal trends long before we do, often in places where we weren’t considering looking. Westmeath are a team on the rise that nobody has really mentioned, but the data shows that they have been improving very steadily over the last few years, with a particular improvement over the past year. They already came close to replacing Laois in division 1B in the league this year, and went on to defeat two other 1B teams in the Leinster round robin, before being knocked out by eventual semi-finalists Galway. Only time will tell if this trend continues or if last year was their peak, but with some luck on their side, the ratings hint at Westmeath being the next team to make the transition from ‘Average’ to ‘Pretty good’.

 

How was the hurling rating created?

So, how does one create their own rating system from scratch? In my case, I started by creating a database of the last few years’ intercounty matches in the National League, All-Ireland Senior Championship, Christy Ring Championship, Nicky Rackard Championship and Lory Meagher Championship. At this point, I’d like to thank various regional newspapers and, once again, Wikipedia, for keeping records of the scores, especially when it comes to tracking the elusive lower division results. Once this database was created, starting scores were assigned to each team. These were based on the league division each team was in at the start of the database, with higher division teams being assigned higher scores. A computer program was written to iterate through the matches and calculate the changes in score. The k value mentioned above, which determines how much the score changes, was determined after some trial and error. This value is important: if it’s too small, the ratings will take a long time to catch up to the actual quality of the team. If it’s too large, than it can end up overestimating how good a bad a team is on the basis of only a handful of results. The k value was eventually selected based on which seemed to produce win/loss predictions most in line with the final results.

 

What’s the point?

At this stage, it should be worth asking: why go to all this trouble? What is the point of creating such a system? Clearly, the answer is high-stakes gambling.

That was a joke. Do not bet on matches based on this system’s predictions. In fact, you probably shouldn’t gamble at all, the bookies always have an edge. Hell, they probably have their own rating system to help them judge the odds.

Still, it can be interesting to see how this system, based on nothing more than a simple equation, can produce predictions which are in line with, or better than, those of many pundits. Free of human biases and sentimentality, if can often detect an upswing or a decline in teams before fans are willing to admit it. It also gives us a method to track how teams have changed over time, and tracking the history of a team’s rating versus their championship performance can be fascinating. Though not a replacement to one’s own judgement, in can be a powerful tool when used in conjunction with the traditional ‘eye test’ of a team, and will hopefully aid you all in many an argument down the pub about who would beat who.

 

Future innovations in the rating system

A final note before wrapping up this article: where to go from here? Well, as the new hurling season starts, I intend to regularly update the ratings based on the results of matches. More than that, however, I intend to tweak and hopefully augment the site’s rating system over time, beyond the limitations of the Elo rating. Though powerful, there are many factors the Elo rating doesn’t take into account. In future articles, I hope to investigate the actual impact of some of these factors, and fold these insights into the algorithm. Until then: thank you for reading this article to the end, and I hope you enjoy tweeting angrily about how underrated your particular county is.