It seems logical to conclude that a hard fought game won by a single point implies that the two teams are evenly matched. By the same token, it also seems logical that a hammering with five goals between the teams implies that the winning team is leagues ahead of their opponent. However, when we watch these matches, we’re only getting a single data point each time. How would the teams perform each time if they could play the same match over and over? Was that scoreline a fluke, or can the difference in points really be a strong indicator of the relative quality of the teams involved? In this article, we investigate whether or not being on the wrong end of a thrashing really means you’re that much worse than your opponent.
The premise for determining the importance of the winning margin was, similar to the article on home performance versus expectation, based on the site’s rating system. Using the current rating system (a modified Elo rating which also takes home advantage into account) as a control, a new rating system was created by adding in bonus points. The bigger a team’s winning margin, the greater the improvement in their rating. Similarly, a team that lost by a large amount will see a larger drop in their rating. Rather than following a linear increase in bonus points, this system instead followed a logarithmic curve, with the bonus points increasing more slowly when the score difference becomes more extreme: While we want to avoid having too much emotional motivation rather than pure logic when devising these systems, it’s fair to assume that neither team will continue to play flat-out when there’s already 15 points between them, and this doesn’t change much if that becomes 20 points. However, it’s much more impactful to go from 2 points down to being ahead by 3.
Both the new rating system and the existing system ran through each match, and the differences between expected result and actual result were recorded, in order to determine each system’s predictive power.
In the end, there was very little difference in how the system performed with and without these ‘bonus points’. There was a slight decrease in predictive ability when bonus points were included, but nothing more than a fraction of a percentage. While adding these bonus points didn’t damage the rating system, the conclusion which can be drawn from this is that a win is indeed a win, whether it’s a 16 point hammering or a one point nail-biter.
It’s intuitive to think that a bigger scoring margin would indicate a bigger gap in quality between teams, and while this can be the case in many occasions, there are also factors which can skew the results and render this factor irrelevant.
It seems dismissive to just say that certain matches are ‘flukes’, but sometimes, for whatever reason, one team just doesn’t show up, and the other fires on all cylinders. For example, last year Waterford managed to beat Tipperary in the league, and came extremely close to reaching the All-Ireland final in a semi-final against Kilkenny. However, in the second half of the Munster final, everything went wrong, and they lost by 21 points. Waterford played terrific hurling both before and after this match, and while probably the underdogs against Tipperary, it wouldn’t have been surprising going into that final if they had won. Good teams can have bad days, just as bad teams can have good days.
Another factor which can play into these results is that once a lead has been firmly established, the makeup of the match can change. Each team might try out some inexperienced subs, the winning team might start easing off, and some of the players on the losing side may start calming down and start hitting fewer wides, now that there’s nothing left to lose. The end result is a match that, while the result was never in question, can see the lead dwindle down to a much closer final scoreline, despite an obvious gap between each team’s real ability.
Finally, the results may have been impacted by limitations of the rating system used to test it in the first place: a team’s expected ability may be impacted by things like injuries or changes in management. This is not caught by the rating system, which only factors in match results and home advantage. In other words, there could indeed be a correlation between winning margin and team quality, but it couldn’t be seen through all the noise of other factors influencing how teams perform.
Which teams win big?
Ok, so we’ve found that we can’t correlate big wins with big gaps in skill between teams. However, what if we look at the teams that consistently win big? Additionally, what can we learn about teams from looking at how they perform in both their wins and their losses? Which teams will scrap to the bitter end, keeping their losses close? Which will rub in their victories, extending the lead as far as it will go? Let’s take a look at how each team has done in the last couple of years:
The above charts display the average winning or losing margin for each team, disregarding how many matches they won or lost, since 2015. This can give us some insight into how these teams play their matches. Let’s narrow the chart to just the current Division 1 teams:
As well as looking a bit like a bootleg Windows logo, this graph provides us with hints to each team’s situation.
In one corner we have Kerry standing alone; caught over the last few years between being too strong for division 2A, but still among the weaker teams in 1B: The end result is a team that has both won and lost by big margins as they transition from playing weaker teams to playing against some of the best.
On the opposite side we have Cork. While still a strong team overall, Cork have been inconsistent, and haven’t often been the definite favourites or the definite underdogs in their matches over the last couple of years. The end result is lots of close, competitive matches, won or lost by only a few points. Kilkenny also find themselves in this quadrant: perhaps their status as the team to beat has resulted in teams always trying to give their best against Kilkenny. The end result is Kilkenny not necessarily winning by massive margins, but their overall skill preventing any large losses. Finally, Waterford are the last team in this quadrant, a result of their tough defence but often patchy offence.
In the bottom left, we have Offaly and Laois. Two of the teams who have struggled the most over the time period studied, they have the unenviable position of losing badly when they lose, and barely scraping past the finish line when they win. Though this chart doesn’t reflect winning percentage, neither of these two teams can draw optimism from how the chart would look if it did. Closer to the average are Wexford who, though they have had more success than Offaly or Laois, still have a lot of room for improvement.
Finally, the top right corner shows us where most teams would like to be: Big wins, and close losses. Here we can see last year’s champions, Tipperary, who have the lowest average scoring difference in their losses, while also dominating most of their wins. The current Tipperary team do not give up easily, and will not relent once they’re ahead. Near to Tipperary are Galway and Clare; Galway have certainly been one of the top teams over the year’s examined, and will likely only find themselves further in the corner of this graph as the league progresses, due to their weaker competition this year. Being the top team in 1B last year, Clare appear to have benefited by this measure, though they also managed some high scoring against tougher opposition in 2016, most notably in the league semi-final against Kilkenny. Limerick also find themselves in this quadrant; possibly a result of usually being the second best team in division 1B, but also perhaps a sign of a team on the rise.
Overall, it’s hard to make a strong decision about the importance of scoring difference. Big wins are definitely more likely against much weaker opposition, but can still happen against tougher opponents, just as heavy favourites can sometimes struggle to scrape a win against underdogs. As we found, looking into a team’s average scoring margins gave us some insight into how they play and where they are as a team, but often teams could have similar results for very different reasons. All in all, the margin of victory can not be written off as totally unimportant, but its significance can very easily be obscured and overwhelmed by other factors. The correlation between winning margin and relative ability was certainly not enough to base a rating system on, though at the same time, it didn’t really hurt the system’s predictive ability either. This kind of indecisive result can be frustrating or disappointing, especially when trying to finish up an article like this with a clear message. In the end, a big part of both analytics and sport is accepting that, most of the time, there’s no clear, definite, one-size-fits-all solution to plan around.
And anybody who tells you differently is trying to sell you something.