In the summer of 2018, the national lacrosse teams of 46 countries from all across the globe travelled to Israel to battle it out to named the best team in the world.
Over the course of the tournament, there were 170 games with a total of 6,366 shots taken, of which 3,287 resulted in goals. 7,142 ground ball situations were contested, 3,701 Faceoffs taken and 1,595 minutes spent on the penalty bench. The result was that the USA were crowned champions for the 10th time after defeating Canada 9-8 in the final.
All data has been taken from the publicly available information on worldlacrosse2018.com, whose website seems to be no longer operational. Many thanks to them for gathering such high-quality data that makes the subsequent analysis in this blog possible.
Unsurprisingly, the USA came out on top in almost every category. They scored the most goals, won the highest proportion of faceoffs and ground balls and spent fewer minutes on the penalty bench than most teams. This is in addition to the fact that every game but one that the USA played was against teams who finished in the top 5 in the tournament, making their statistics all the more impressive. Statistically, if nothing else, the USA were the best team at the championships.
On the flip-side, the statistics will be tough reading for anybody from Uganda. Uganda were consistently at the wrong end of every chart, scoring among the fewest goals, winning fewest faceoffs and having the most penalties of any team in the tournament. This was however their first major international tournament, and their participation should be celebrated as they were the only African nation competing at the World Cup. This meant that teams from 6 continents* were represented at the championships, showing the growing popularity of the sport worldwide.
* Just waiting for Antartica to submit a team now...
Some of the observations that I personally found more surprising:
There is an old saying in Lacrosse that "ground balls win games". For those unfamiliar with the sport, a ground ball refers to a point in the game where neither team is in posession of the ball. This can be due to a number of reasons - perhaps a misplaced pass, or a defending player successfully knocking the ball out of the stick of an attacking player.
With the granularity of data provided, we can test this theory. For every game, the total number of ground-balls won by each team can be compared to the final score. The values have been converted to percentages to make them comparable between each game and each team.
Note that on the above chart, there are two dots for every game (one dot per team). Note also that any dot in the top half of the chart represents a team that scored more than 50% of the goals scored in that game i.e. the team that won that game.
It is immediately clear that there is a positive correlation between winning ground balls and winning games. The actual correlation value is 64%, which is very high for a real-world dataset in scenarios as complex as live sporting events.
Looking at the chart in a bit more detail shows that no team that won 70% or more of ground balls lost any game at the World Cup. Conversely, no team that won 30% or less of ground balls won any games. If we analyse the data a little more closely, we can calculate the actual chance of winning a game based purely on the proportion of ground balls won.
This chart shows the proportion of games that were won for every proportion of ground balls won. The red line is the average, and the black lines are 1.5 standard deviations above/below the average. It's a basic method of estimating upper/lower bounds, outside which results could be considered unlikely*.
* Getting technical: for a normally distributed dataset, it covers 90% of outcomes. This approach has been used rather than directly calculating percentiles as it is more robust when working with small sample sizes (as we have at the highest / lowest ends of the axes).
The chart shows that teams winning 75% of ground balls should expect to win 80% of games (red line), and are very unlikely to win less than 65% of games (black line). On the flip-side, teams that win only 30% of ground balls can expect to win around 25% of games. The best results that such a team could expect is to win half of their games. That's not a good position to be in.
So - it seems to me that there is a lot of truth in the old saying. Ground balls do win games.
Since we also have the data available, let's look at Faceoffs too. For those unfamiliar with the sport, a Faceoff is taken to start every quarter of a game, and is also taken after every goal is scored. It's similar to a faceoff in ice hockey, or a throw-in in basketball. Both teams contest a 50:50 position, and the winner ends up with possession of the ball, from which their team can start an attack. Facing off is a highly specialised skill that typically only 2-3 players on any team will specialise in. Some teams even train a single player whose sole responsibility is to win face-offs and who will then substitute off the pitch to allow other players on to attack or defend.
Being good at faceoffs can mean an extra 10-15 possessions per game, in which time you are more likely to score and unlikely to concede. Therefore, we would expect to see a positive correlation between winning faceoffs and winning games. Repeating the analysis we did before for ground balls, we can create the following chart:
The correlation is there, but there is significantly more scatter. Some teams won 90%+ of face-offs but still ended up losing the game. The correlation value is 37%, which shows that there is a link but it's not the strongest or most-dominant factor.
This chart shows that winning more Faceoffs does still increase the likelihood that you will win the game, but the effect is less than it is for ground balls. Furthermore, the likely range of results is far greater than it was previously, as indicated by the distance between the black lines. A team that wins exactly 50% of faceoffs could realistically expect to win as much as 80% of games, or as little as 20% of games.
Having said that, it is clear that winning faceoffs is still advantageous to any team. A team that wins 80% of faceoffs could expect to win 65% of games, with all other factors being equal. Note that the downturn at 85% is unlikely to be real and is most likely caused by exceptional results (such as the team that won 90% of faceoffs and still lost).
The final attribute I have assessed - does getting a lot of penalties significantly reduce the chances of your team winning?
Answer: not really. The correlation is -5%, which is close to negligible. It correctly identifies the negative correlation (more penalties = less chance of winning). The chart below shows the effect of averaging over all games.
This chart shows that there is no real change in the team's likelihood of winning due to the number of penalties. So long as the penalties are roughly evenly distributed between both teams (30-70%), there is no real effect on your team's chances. The effect is most pronounced if one team is getting significantly more penalties than the other. A team that gets 75%+ of penalties reduces their team's likelihood of winning by 5-10%, but even then, there is a lot of scatter in the results.
So - while it may seem counter-intuitive, the statistics seem to show that spending time on penalties doesn't have a significant impact on your chances. However, I really wouldn't advise this as an excuse to start making late hits...
Going back to a point made earlier, this could explain why Jamaica ended up finishing 13th. Without being rude, I wouldn't personally have expected Jamaica to finish in such a high position. The stats seem to suggest that Jamaica won a high proportion of ground balls, but also gave away a lot of penalties. Combining these two effects, the positive effects of winning ground balls and the negligible effects of conceding penalties, the net effect ensured that they had a very successful World Cup campaign.
Finally - because it was available, I have crunched the numbers for all players for all teams for the available statistics. They are given in the charts linked below.