The 2017 senior hurling league is almost upon us. While it may be considered less prestigious than the All-Ireland, or even the Munster and Leinster titles, it still remains as one of the important tournaments for intercounty hurling, as it provides each team their first proper challenge against their peers. It can be questionable how much effort each county is putting in at this stage, with the league simply offering a period of experimentation and conditioning before the ‘real’ test of the championship for many. Nevertheless, the league provides teams with an opportunity to demonstrate what they can do against opposition of similar quality, and can give insights into where teams have improved or where they still need to improve.
As is customary for anybody writing about sport before a major championship begins, it’s time to begin making predictions. In order to prevent emotion and personal bias into this prediction, we will be making use of our rating system to produce these predictions. Otherwise, the article would simply be, ‘Clare to repeat, followed by winning everything forever’. Additionally, by comparing the results of the simulation with the initial ratings, we can take a look at how fair the structure of the league is: For example, is it an advantage to start in Division 1B, as the past two winners have done, since this provides an easier road to the qualifying stage? Are you better off starting in Division 1A, playing quality opposition from the beginning, and being rewarded with an easier match in the quarter finals? Or, do these effects nullify one another?
A quick explanation of the simulation
Teams were given initial ratings based on the rating system produced in previous articles. A computer program was written to simulate each of the league fixtures for the 1A and 1B Divisions. The odds of a team winning a match was determined by the relative rating of its opposition, and after a result was generated, each team’s rating was updated. The simulation was run one million times, and each team’s overall performance was recorded: whether the team won the league outright, was a runner-up, reached the semi-finals, reached the quarter-finals, failed to reach the knockout stages but remained in the group, or if they were demoted to division 1B (if a 1A team) or in the promotion/relegation match against the winner of 2A (if a 1B team). The simulation took home advantage into account. It assumed that the semi finals and finals would be on neutral ground, but that the quarter finals matches would be played at the grounds of the team with the fewer home games up to that point. If each team had played an equal number of home matches, a ‘coin-flip’ (50% random chance) was used to determine who had the home advantage. One limitation of the system was that it did not predict any draws; the system would only ever assume one team beating the other. Nor did it factor in scoring difference: if two teams had equal points, the team who won their head-to-head match-up would be ranked ahead in the league table. If three teams were on equal points, with a rock-paper-scissors scenario of each team winning and losing against one other, the rankings would be decided by the team’s relative ratings.
The results of the simulation are displayed in the table below, ordered by the number of simulated leagues won:
This table shows the percentage of times each team was relegated to 1B (if a 1A team) or in the 2A/1B promotion/relegation match, (if a 1B team), remained safely in their division but did not qualify for the quarter finals, was knocked out in the quarter-finals, was knocked out in the semi-finals, was runner-up, and won the league outright.
Additionally, for the 1B teams, this table indicates how often in the simulation they were promoted to next year’s 1A division.
Fairness of the league structure
If we compare the predicted winners and losers of the league with the initial ratings, there is a pretty high level of correlation. There was a reasonable amount of fuss last year to the Galway management stating that it wouldn’t be the end of the world if they were demoted to 1B, and while that’s certainly not a winning attitude, there’s truth to it: They’re predicted to win an appropriate percentage of the time based on their rating. If a team has the ability to win outright, they shouldn’t care if they play the other best teams in the knockout portion or the early stages of the league. All in all, the league is a fair competition, at least within its division; a later article may be dedicated to whether or not the relegation/promotion matches between divisions make upward mobility too difficult. With this fairness established, let’s get back to the numbers, and get back to making wild predictions about the teams.
The chart below displays how often each team won the league in the simulation:
It’s hard to argue with some of the teams favoured by the simulation. Like many pundits, it favours teams such as Tipperary, Kilkenny and Waterford. Galway, playing in division 1B this year, are also given a good chance, as they’ll have an easier path than most to the knockout stages, similar to the paths of Clare in 2016 or Waterford in 2015, both eventual winners. Cork and Dublin are noticeably less favoured than most 1A teams. While it’s still very possible that either of these teams could do well, they were also the two teams most favoured for demotion to 1B by the system.
From 1B, Limerick are unusually low down in the winners table, despite being predicted as the second likeliest to be promoted to 1A next year. Whether this is just a quirk of the randomisation in the simulation, or a result of how the system interprets Limerick’s likeliest path to the final, it definitely seems to underrate them.
The chart below gives the percentage of simulations teams found themselves in the 1B/2A relegation/promotion match:
The simulation does not like Laois one bit. Not only does it give them the lowest rate of success to win the league, it also gives them the highest probability to go to the promotion/relegation match. Which, in fairness to the simulation, is understandable, as they did just that last year, and followed it up with a dismal All-Ireland campaign.
Offaly are also viewed with some pessimism by the system, and after last year’s poor performance they’ll be seeking to improve, particularly if Kerry continue to build on their own improvement, which the simulation possibly overrated.
Changes for 2018
The above tables display who the simulation sees going up to 1A next year, and who will replace them in 1B. The simulation agrees with most pundits that the promotion battle will be between Limerick and Galway, though again it appears to overrate Kerry. It’s likely that the system had underrated them early in the database, so their relatively strong performances last year have boosted their rating excessively. Of course, nothing is certain in sport, and maybe the system’s right, and we humans have no idea of the beautiful hurling Kerry have in store for us.
Dublin are in the unfortunate position of being deemed most likely, by a pretty large margin, to be demoted next year. Behind them are Cork, who came off a poor year last year, and Waterford, last year’s runner-up. Though Waterford are rated poorly here, it’s worth remembering that they only have two home games in the group stages of this year’s league.
Limitations and conclusions
Of course, while it’s fun to make predictions about any competition, there are always limitations to how accurate these can be. The entire betting industry depends on it. The rating systems fail to take in a wide variety of factors. Many analysts are tempted to assume that because a factor’s impact is hard to measure it doesn’t actually have any impact. This is obviously untrue. There have been matches at the club level, and pre-league inter-county matches, which are not in the rating system’s database. There have been changes in management for half of the division one teams, including the move of last year’s winner’s manager, Davy Fitzgerald, to Wexford. There have been changes in personnel, and will be further changes that we can’t foresee, due to injuries or personal issues or simply poor performance. All of these make forecasts difficult even for the best minds in the sport, nevermind a simulation based on a very basic rating system. So, with all that in mind, are there any real conclusions we can draw from all this?
Clare to repeat in 2017. Check the data again, I’m pretty sure that’s what they’re hinting at