Whatever happened to home advantage?

At time of writing, we are now slightly more than halfway through the group stages of the 2019 All-Ireland, with 18 of the 30 group games played across the Leinster, Munster and Joe McDonagh championships. Though it’s still early days, and many of the early conclusions will probably soon be proven wrong, a few narratives have already emerged: Tipperary have shaken off their underwhelming league with a very strong Munster campaign (something that was hinted at by this site’s article analysing the league). Waterford have followed up on their strong league with a disastrous Munster campaign (something that was not foreseen). Offaly have failed to rebound from their relegation from Leinster, as many expected, and now face a successive relegation to the Christy Ring if Kerry pick up even one more point in their next two games.

However, teams performing above or below what we expected is something that happens every year. There is one notable stat, however, that does feel very strange: What has happened to home advantage?

In this article, we will quickly explore whether or not it's too early to say if this championship is a genuine outlier, and speculate about why this is happening.

Home advantage

Home advantage is not an imagined phenomenon: In this older article, the effect of playing at home was analysed, and it confirmed that there was a tangible improvement for most teams when playing at their home stadium. Though the rating system used on Take Your Datapoints has changed significantly since then, the BAINISTEOIR v2 system still puts quite a large emphasis on whether teams play at home, away or at a neutral ground. How important this is depends on the team: over the last couple of years, for example, Clare and Kilkenny have led the pack in how much their performance improves at home (though Clare’s rating in this regard has taken a bit of a hit after losing by 13 points to Tipperary last Sunday) versus how they’d play in other stadia.

However, according to the linked article, the home team wins about 54% of the time, drawing another 6%, while teams on neutral grounds win about 47% of the time, and draw about 6% of the time. These figures date back to 2017, but have been reasonably consistent with the All-Ireland and League competitions since then.

How unusual is this year’s championship?

As mentioned above, the BAINISTEOIR v2 rating system used on this website accounts for home advantage when determining whether or not a team is expected to win or lose. While it doesn’t always correctly predict the winner, over the course of enough games it does a good job with the odds, in that the favourite wins roughly the percentage of the time the system predicts. In other words, if the average favourite across 100 games is given a 60% chance to win by the rating system, you could safely assume that roughly 60 of these matches will be won by the favourite.

So far in this year’s championship, the rating system expected 10 of the 18 home games to be won by the home team (draws counting as half a win). Instead, there have been only 7 home ‘wins’ (6 home wins and 2 draws). In Munster specifically, the rating system expected 3.5 home wins from 6 games, but instead there has only been one: Tipperary vs Waterford in Semple Stadium. Just so you don’t think the rating system is broken, here’s a comparison of the results of previous competitions with how the BAINISTEOIR v2 system expected things to go:

This definitely looks dramatic; The previous couple of tournaments have gone almost exactly as predicted: Rounding the percentages to the nearest number of wins, the rating system prediction was exactly right for the 2019 league, and only predicted one more home win than actually happened in the 2018 All-Ireland. Less than halfway through, and it’s already 3 wins off for the 2019 All-Ireland, and 2.5 of this comes from the Munster championship.

Now, it’s important to temper this with some important qualifiers: The sample size is simply too small to say that this is significant. Statistical power is a measure of probability that we can reject the null hypothesis (In this case, that the proportion of home wins expected by the rating system is the same as what is observed).

Performing a chi-square test, we still have only a 31% probability that there is something not being accounted for by the rating system when it comes to home games in this year’s All-Ireland. When Munster is removed, the rest of the All-Ireland really doesn’t look strange at all, and there’s only about a 7% probability that the null hypothesis is rejected, and is only slightly higher than previous tournaments (3% power for the 2019 league, 5% power for the 2018 All-Ireland).

Only including games in Munster, the power value goes up to 51%; this is still too low to really make a judgement, but is impressive given the tiny sample size of 6 games. Depending on how the rest of the home games go, Munster could finish with anywhere from an 89% probability that something strange is happening (if every home game is lost), all the way down to a 6% chance (if every remaining home game is won).

In other words; it’s too early to say whether or not this is a genuine anomaly, or if it will revert to the mean soon. As is often the case when trying to analyse a complex game like hurling with only a handful of intercounty matches per year, the sample size simply isn’t enough. However, there are actually quite strong odds that Munster’s home games are significantly unusual considering there have only been 6 matches so far. Insufficient to say we should scrap the home games, but enough to sit up and take notice, especially since it’s not something that we’ve seen in the last few tournaments.

Why is this happening?

So, while far from being mathematically certain, and not quite mathematically probable, it is possible that there’s something wrong with playing at home in Munster. However, we will take the cautious approach, and assume that things will revert to the mean and teams will start winning at home again. However, it still compels you to ask: why hasn’t it done so yet? What is it about the schedule so far that has resulted in so few home wins, especially in Munster?

It may not be the whole story, but there are two main hypotheses that I have:

  1. A schedule so far weighted towards those without a consistent home venue and

  2. Overrated teams getting home games

Irregular home games

Of the 15 teams playing in Leinster, Munster and the Joe McDonagh, I have identified 6 where they have had some issue or another relating to their home venue, meaning it’s either not always the same venue, or it hasn’t always been used in championship games. These 6 are distributed equally across the three groups:

  • Joe McDonagh

    • Antrim’s saga with trying to get the new Casement Park built has dragged on for years, meaning they’re unable to play in their biggest population centre of Belfast. Instead, they have been forced to use club pitches in rural areas of the county, or use the home venues of neighbouring counties.

    • Kerry usually play their hurling in Austin Stack Park in Tralee. However, they have still occasionally had matches in Fitzgerald Stadium in Killarney, further away from their small pockets of population that favour hurling over football.

  • Leinster

    • Officially, Parnell Park is the home stadium for Dublin. However, there have been plenty of occasions where they’ve used Croke Park too.

    • Galway have had Pearse Stadium in Salthill as a consistent venue, but they weren’t allowed to use it in the championship until last year. Instead, they had to play in a neutral venue. In fact, since they joined Leinster in 2009 up until they were allowed their own home games in 2018, Galway only played 2 fewer championship games in Offaly than Offaly did!

  • Munster

    • Cork have spent a lot of the last few years waiting for the new Páirc Uí Chaoimh to be built, instead using the much smaller Páirc Uí Rinn. This inconsistency was exacerbated this year when, due to issues with the sod, Cork had to switch back and forth between both stadia.

    • Waterford have had multiple issues; as well as not being allowed to use their home venue of Walsh Park in the championship until this year, they also sometimes use Fraher Field in Dungarvan instead, taking them to the opposite side of the county.

When you break down how these 6 counties have performed at home over the last 5 years versus the other 11 who all have consistent stadia, there is a significant difference; those with consistent championship venues have won 5.9% more often than their overall average, while those without win only 2.5% more often. In Munster, possibly because it’s the only group where all counties’ fanbases follow hurling more than football, the difference is even more drastic: Combined Clare, Limerick and Tipperary have won 12.2% more of their home games than their overall average in the last five years (Clare lead all teams over the last five years, with an astonishing 28.9% improvement in winning percentage at home versus overall). Combined, Cork and Waterford have only won 2% more.

This alone could explain why home wins have been so rare in Munster rather than the other groups, but the schedule has exacerbated this. In Leinster and the Joe McDonagh, the teams with inconsistent venues have had 47% of the home games so far, despite only accounting for 40% of the teams. In Munster however, 60% of the home games have been in these inconsistent venue counties.

Though the rating system accounts for each team’s usual performance at home, it doesn’t know the difference between a championship match and a league match, with the former typically having much bigger crowds than the latter, and it also doesn’t distinguish when a county is using a secondary ground, such as Páirc Uí Rinn or Fraher Field, rather than the primary one. In this way, it could have massively overestimated the odds of certain teams when playing at home.

Overrated teams

The other, more simple and straightforward hypothesis is just that some teams have been overrated or underrated going into this year’s championship, and it’s just so happened that the overrated teams have played more at home, while the underrated teams have not.

The chart above plots how many more home games than away games each team has had, versus how much their margin rating (the winning or losing margin the rating system thinks this team would score above or below the average) has changed since the end of the league; the thinking being that the teams whose ratings improved since then were underrated, and the teams whose ratings declined since then were overrated. The plot does appear to correlate: the teams who have played more home games have almost all seen their rating go down, while those who have played more away games have seen their ratings improve. Of course, the most drastic examples again happen to be in Munster: Tipperary are the most improved, and Waterford the most disimproved.

Now, this chart isn’t perfect: the ratings can only adjust to the current ability of the team after they’ve played matches, so the results are affected by what they’re trying to measure. However, if we look into some of the other stats, especially those that aren’t measured by the rating system, we can see how some teams were incorrectly rating coming into this year’s championship. As mentioned above, Tipperary had the biggest difference of any team in this year’s league between their actual winning percentage, and the winning percentage you’d expect given their points scored versus conceded (Pythagorean expectation). They were considered massively unlucky not to do better, and it seems like now they’re reverting to their very impressive mean.

Similarly, it’s worth considering that the competition is quite different between league and championship. Until the knockout stages, Waterford didn’t encounter any other team in Munster. They had very easy games against the majority of teams in division 1B, and even the stronger teams, like Galway, may have been overrated coming into the start of the year (Galway’s margin rating has gone down by 3.5 points since 2018, and are another example of a team that’s struggled at home compared to their expectations). This caused Waterford’s rating to inflate somewhat: Waterford had the second best defensive rating (score conceded per game) in the league, but when you removed their games against the weaker teams of Carlow, Laois and Offaly, Waterford’s defensive rating was significantly below average. Using another example, Galway’s offensive rating dropped off relative to the rest of the pack, and their usual scoring has been somewhat absent since the Leinster championship began.

It can be difficult, until a sufficiently large number of games have been played, to tell whether a team’s rating improvement is a bubble or a genuine change. In this year’s league, Waterford were far and away the most improved team, while Tipperary continued to decline after an underwhelming 2018; looking at the Munster table now, it does seem like this was more bubble than real change. It’s very possible that, for everything else that’s been said about it, the real story of why teams in Munster haven’t played well at home is simply that Tipperary are much better than we thought, and have mostly played away, while Waterford are much worse than we thought, and have mostly played at home.

What conclusions can be made?

Analysing the statistical power of the unusual results, as well as talking through the hypotheses, it does seem like the home advantage should resume normal service soon. As more games are played, and as all teams have played as much at home as away, I would expect things to revert to the mean.

There are other insights that can be drawn from looking into the numbers around this: for example, teams benefit from having consistent home venues for championship games, and I’m sure any fans of Antrim or Cork would be much happier if they ever get all the issues with building their new grounds resolved. Home advantage doesn’t actually count for that much in counties where the home fans can’t (or simply don’t) show up in large numbers.

Another thing worth noting is the limitations of rating systems, or any other means by which people determine odds. They can’t account for everything, and you always lose details when trying to reduce everything to one or two numbers. It seems clear now that most people did not expect such a surge from Tipperary after an iffy league, and that maybe we all expected Waterford to recover from last year’s slump a little bit too soon.

Finally, the key thing to remember from all this is that we can overreact too quickly: it seems that pundits have a new favourite to win the All-Ireland after each game is played, and it’s absurd to hear conclude, after two games in Walsh Park, that Waterford should return to playing in Limerick or Thurles instead of at home. When you see a couple of unusual results close to one another, there’s a tendency to take one or two of these datapoints and extrapolate wildly into the future. People need a story to talk or write about, but most of these will seem ridiculous in retrospect as things balance out once enough games are played. As the old Irish saying goes: It’ll be grand.

Of course, if it’s not grand, and if the home losses continue to build up, I may need to revisit this article later on. And if it comes to it, I hope I’ll be able to figure out what the hell is going on, because we Clare supporters depend on home wins more than anybody!