The league is over for another year, and so our attention now turns to the 2017 All-Ireland Championship. To mark the occasion, a new simulation has been built in order to try to predict who will win. While the results are interesting in of themselves, they have also drawn attention to certain aspects of the championship which could be worth criticising.
The simulation is based on the site’s rating of each team, based on their rating before their opening match, and works in more or less the same way as our league preview did (That particular simulation had eventual winners Galway as the third most likely to win, with runners-up Tipperary second most likely, so not a terrible result from that perspective, though it definitely underrated Wexford, giving them only a 10% chance of promotion). The simulation was run 1,000,000 times, and each team’s performance was recorded in each run of it. However, the championship has a few unusual traits which made it trickier to predict accurately, and trickier to simulate overall:
Firstly, we don’t actually know all the teams involved in this year’s championship. Due to a new rule change, the winners of the Christy Ring now play a pre-qualifier match with one of the Leinster quarter-finalists, as well as entering next year’s round robin. This meant that the Christy Ring championship also had to be simulated.
Secondly, the Leinster round robin is a different format to the simple knockout-style of much of the championship. Like the league, the simulation needed a criteria by which to decide between three teams on the same number of points, all of whom had won and lost to one of the other two. Though this was an issue in the league too, the reduced number of games make this tie-break scenario much more likely.
Thirdly, there is a lot of randomisation with who plays who in the qualifier rounds. Even if the simulation correctly guesses who loses each match in Leinster or Munster, it may not predict who then goes on to play against one another in each qualifier round match. Additionally, many venues are not yet decided, so in these cases the simulation either needs to guess at which county’s home venue it’ll be, or if it will be at a neutral location.
Finally, not all teams have the same number of matches, and not all teams reach the quarter or semi finals through the same means. More on that later, but for now: our predictions.
Initial pre-tournament predictions
The charts below display how likely the simulation felt each team was to win the Christy Ring, Munster, Leinster and All-Ireland championships, prior to their first match. As mentioned above, the Christy Ring winner now also has the opportunity to compete for the All-Ireland in the same year, hence its inclusion in the relevent 'feeder' tournaments.
Prediction time lapses
For those of you reading this on desktop (apologies to mobile users, this may not always display for you), we have also provided these interactive time lapses of how the odds change for each tournament. These time lapses shall be updated both here and on the main ratings page of the site, so feel free to come back and check these updated odds as the summer progresses.
As touched on earlier, and possibly hinted at by some of the odds above, the championship is a highly unfair tournament. If you look at certain teams’ ratings, compared to their odds of winning, there is not always a huge amount of correlation. For example, Offaly are only rated slightly higher than Laois, but are more than three times as likely to win Leinster, as they don’t have to get through the round-robin first. It could be argued that good teams have earned their seeding, and that these aren’t teams who would win much anyway, but the lack of balance is found throughout the tournament, even amongst potentially competitive teams: Limerick begin with a lower rating than Cork, but their odds of winning the Munster championship are more than double Cork’s odds. Similarly, they are considered more likely to win the All-Ireland championship outright. Clare are only rated slightly higher than Cork in the ratings, and yet have roughly four times the odds of winning the All-Ireland.
The reason for this is quite straightforward: Though there are a few factors, such as home advantage, or differences in how long different teams have to prepare, the key difference which can sway the odds dramatically for teams is the difference in the number of games. A team like Limerick could, potentially, win four matches* and be crowned champions. Under the most imperfect circumstances, a team like Laois could be required to play up to 10 matches** before they could be given the same accolade. To even reach a provincial final, Limerick need only to win their first match, while Laois’ best performance would total five wins. Both teams play in the same division in the league, but Laois have only a fraction of a percentage chance at winning the All-Ireland, only winning once every 250 runs, according to the simulation, whereas Limerick would win it roughly once out of every 25 times. Limerick are a better team than Laois, but not by that much. Laois, however, have so many opportunities to be knockout out, while only getting the same amount of leeway when they fall flat.
There are many criticisms of the championship structure. It’s confusing. It’s convoluted. Galway and Kerry are in the west of Ireland, not Leinster. However, this unfairness alone should be reason enough to address how it’s set up. There is a wonderful tradition to the championship, and to the Munster and Leinster championships, but should they be connected the way they are, giving some teams such large advantages? The Ulster championship is still held, but no longer results in a mismatched All-Ireland semi-final: should we consider doing something similar for Munster and Leinster, having them continue as independent tournaments, making room for a sensible championship structure?
Imagine a perfectly competitive championship, containing 22 evenly matched teams, the same number of teams that could potentially win the All-Ireland at the start of the tournament (this includes teams eligible to win the Christy Ring, and enter the All-Ireland that way). In this perfect world, each team could be expected to win, on average, 4.5% of the time, if the championship structure was fair. This perfectly competitive championship was run in the simulation, where each team started with the same rating, and each venue was set to neutral. By the rules of the All-Ireland, teams in the same position as Clare, Waterford, Limerick and Waterford won the All-Ireland at triple that rate, about 13% to 13.5% of the time each. Teams starting in the Leinster round robin would only have about 3% odds, just two thirds of what they aught to get, despite being as good as those seeded in the provincial semi-finals. Finally, those attempting to qualify through the Christy Ring may as well not show up, winning only about 0.4% of the time each, only a tenth of what they would get under a fair structure. There is a certain amount of variance in this simulation, and running it again would produce slightly different numbers, but nowhere near enough to counter this inherent bias towards teams in certain situations.
In other words, even in a perfectly balanced world, where everything else is equal, Kilkenny are still probably going to win it yet again. Typical.
*Win Munster semi-final, win Munster final, win All-Ireland semi-final, win All-Ireland final.
**Three Leinster round-robin games, win Leinster quarter-final, lose Leinster semi-final, win first qualifier round, win second qualifier round, win All-Ireland quarter final, win All-Ireland semi final, win All-Ireland final.