The fall of the field goal in men's hockey

And the rise of penalty corner defence?

Introduction

Bavo Bruylandt asked a good question about the decline in goals scored in the men’s game at the Olympics. Here is the comment and question:

“Assuming more players in circle means less time in mid-field build-up, and more actions in 23m and circle, would that also not mean more PC's are granted? And PC's having a success rate around 20% would then be more prevalent in the goals scored compared to field goals.

That would easily be testable with this dataset: how did the absolute number of PC change, does it correlate (negatively) with goals scored? And does indeed the relative number of PC versus FG change as well?”

Let’s recap quickly because the question flows from this previous article which supported Shane McLeod’s assertion that goal scoring decreased at the last Olympics. The reason McLeod gave for the decline was the tactical change to deep defending in order to counter the threat of attacking aerials thrown to land in or around the circle.

We’ll focus on the last three tournaments (Eurohockey 2023, ProLeague 2023-24 and of course the 2024 Olympics) and pool the other tournaments¹ because the decline in goals scoring was apparent in the Eurohockey and ProLeague competitions and especially in the Olympics.

Figure 1

Field and Penalty corner goals.

What Bavo focused on was the difference between field goals and penalty corners and penalty corner goals. The assumption I think is simply that with more players defending low on the pitch, and more attackers amongst them, the chance of scoring a field goal declines. But, by contrast the chance of winning a corner (and so the chance of scoring a corner) in that crowded environment increases².

As a first step we take the data from Figure 1, drop the “Earlier tournaments” to focus on just the last three competitions and simply separate out the field and the penalty corner goals.

Figure 2

That’s interesting. We know overall that the number of goals declined markedly in the last Olympics but this summary of goals per game very strongly indicates that while field goals did decline (averages were 3.1 (± 0.47),  2.9  (± 0.18) and 2.0 (± 0.24) field goals per game for the Eurohockey, ProLeague and Olympic tournaments respectively) penalty corner goals remained roughly the same (average penalty corner goals per game were 1.5 (± 0.36), 1.7 (± 0.14) and 1.6 (± 0.19) for the Eurohockey, ProLeague and Olympic tournaments respectively).

With most data of this sort there can be a fair amount of variation around the average values (see the spread of the data points in Figures 3 and 5 for example) so it is important to check that what we think we see in Figure 2 ( a decline in field but no decline in penalty goals) is supported by a formal analysis³.

Table 1.

And the analysis clearly supports what we might infer from Figure 2. With the 2024 Olympics as the reference tournament (i.e. the tournament to which the other two are compared) there is a significantly lower number of field goals at the Olympics  (p = 0.01 and 0.02 compared to the Euros and ProLeague respectively which is lower than the accepted cut-off of p = 0.05). By contrast, there is no reason to think that the number of penalty corner goals differed at all across these three tournaments (p = 0.76 and 0.71).

The ratio between field and corner goals

Another point Bavo made was that there may be a change in the ratio of field goals to penalty corner goals - the goal difference between the two. We can already guess at this relationship from Figure 2 but here it is summarised.

Figure 3

A clear trend towards there being a smaller difference between the number of field and penalty corner goals across the tournaments. If we pool this data we can ask whether there is an overall difference (rather than a per game difference) in the proportion of field goals to penalty corners.

Figure 4

There is not (p = 0.17 on the top line) and this is mainly due to the variation in data across the three tournaments (the spread in points in Figure 3). But of course, the trend is still there and one can indetify a change by looking at the proportion test statistics above each column. For the Euros and the ProLeague significantly more field goals were scored than penalty corner goals (p < 0.01 for both) whereas there was no difference in the proportion of field goals to penalty corner goals at the Olympics (p = 0.20).

In terms of just the goals then, the result is quite intriguing. There is a decline in goals scored culminating in a lower number of goals overall at the Olympics in comparison to recent tournaments and this decline is wholly attributable to a fall in field goals. But, I’m getting at Bavo’s comment a little backwards because up front he asked about penalty corners. As mentioned at the beginning it is reasonable to assume that lower and more concentrated defence would provide more opportunities to win penalty corners. And if we are looking at penalty corners as well as penalty corner goals we should, to be complete, also look at the number of shots to see if the absolute number changed or the conversion rate of shots to goals changed. Unfortunately shooting metrics are not commonly collected by the FIH during tournaments. The last Olympics was unusual in that it collected not only shots but also possession and circle entry data but there is no shot data for the other tournaments. So we are stuck looking at penalty corners in isolation.

We already know that the number of penalty corner goals hasn’t changed between tournaments so our initial expectation would be that the number of penalty corners awarded hasn’t changed either even in the face of this ‘low block’ defence. What does the number of penalty corners look like then in comparison to the number of penalty corner goals for each of these tournaments.

Figure 5

Well that’s interesting, and slightly odd. The number of penalty corners awarded is higher at the 2024 Olympics (supporting Bavo’s point) but as mentioned above the number of goals scored from these corners has not increased. This obviously suggests that, for at least the last Olympics, the penalty corner conversion rate might have been lower. The individual points are quite spread for each tournament so again we need to check our assumption about a change in conversion rate. What does the analysis say?

Table 2

The corner conversion rate was significantly lower at the last Olympics in comparison to the 2023-24 ProLeague (p = 0.01) and was also lower than the Eurohockey conversion rate (p = 0.073) but the result is not quite significant. Penalty corner conversion rates were around 21% for both the Euros and ProLeague but fell to 15% for the Olympics.

Conclusion

Men’s teams seem to have adapted to the aerial ball. The defences’ set during the Olympics decreased the chance of conceding a field goal but the trade-off may be an increase in the number of penalty corners conceded. Interestingly, this increase does not result in a concomitant increase in the number of penalty corners scored. What’s the explanation? It can’t have anything to do with the original question because scoring from penalty corners is a microcosm of attacking and defending techniques and tactics outside the hurly-burly of open play. Whatever happens to create a corner (lots of defenders feet in the circle for example) does not affect whether that corner will be scored. So the only reasonable interpretation is that corner defence improved significantly in the men’s 2024 Olympics in comparison to the previous two tournaments. The result is accentuated here in the data because of the higher number of penalty corners won. The men have simply got better at defending corners.

There, that was fun. Thanks Bavo for the comment and there will be more analysis from me coming soon.

1. See the previous article for details of the tournaments analysed

2. A simple question. Is the probability of winning a corner correlated with the number of players in the circle? I think it is a general assumption we coaches make, and also one that is easy to confirm or disprove though I have never seen it formally tested.

3. The type of analysis is often dictated by what kind of data one has. In sports analysis the data is often counts (goals, shots, circle entries) and that lends itself to what’s called general linear models with poisson error structures. When analysing rates (penalty corner conversions for example) the overall model incorporates an offset to account for the number of corners from which the goals are scored.