This blog was written by Stephan Lewandowsky and crossposted from the Psychonomic Society’s blog.
Those of you who are soccer fans may find the following passage easy to follow: “In a new age for football, AZ had a goal against Cambuur disallowed for a foul on the keeper after the decision was reviewed. Stijn Wuytens thought he had won the game for his side only to be called back after replays showed a foul by Levi Garcia on Leonard Nienhuis.”
Translation to follow, but let’s first look at the video that captured this event:
Now to the translation: AZ refers to AZ Alkmaar which is a Dutch professional football club from, you guessed it, the city of Alkmaar. SC Cambuur is a Dutch professional football club from, you didn’t guess it, the city of Leeuwarden. The dramatic event in the video occurred during the semi-final of the Dutch Cup in 2017.
The Sun explains the drama surrounding the decision:
“Referee Kevin Blom initially gave AZ the goal to the jubilation of the home crowd, but after complaints from Cambuur players Blom gestured that the video ref should review the goal. Video official Dennis Higler, who watched the match from the city of Hilversum where most of the Dutch broadcasting studios are located, signalled Blom that he saw a clear foul by AZ’s Levi Garcia, who pushed Cambuur goalie Leonard Nienhuis with his hands in the run-up to the goal.”
We have blogged on the expertise of soccer referees here before. That earlier research discovered that elite referees were able to correctly call between 64% and 77% of incidents during actual games. A subsequent article by the same authors showed that slow-motion replays yield more accurate technical decisions, especially for corner kick situations which tend to involve a bewildering number of players jostling for position.
The existing research has left open several questions that researchers Jochim Spitz, Pieter Moors, Johan Wagemans, and Werner Helsen tackled in a recent article in the Psychonomic Society’s open access journal Cognitive Research: Principles and Implications.
Unlike previous research, the stimuli in this study were 60 realistic video clips of actual tackle incidents selected from matches of the UEFA (Union des Associations Européennes de Football). The figure below shows a screen snapshot of a stimulus video:
Two versions of the video clips were created. One that displayed the incident in real time (for 3 seconds on average), and one that presented the same incident in slow motion (12 seconds). A pair of professional referees inspected each incident until they reached consensus about what the “correct” reference decision should be. Out of the 60 videos, 4 were determined to contain no foul; 2 a foul with no penalty card; 36 a foul that should attract a yellow card; and the remaining 18 a foul with a red card.
The participants were 88 elite referees who were all active in professional European soccer. Each participant evaluated all 60 incidents and determined (a) whether a foul had occurred, and if so, (b) what sanction to hand out (no card, yellow card, or red card). Half the incidents were viewed in slow motion and half in real time. Each incident was presented once at each speed across participants.
Turning first to the accuracy of decisions, there was no difference in accuracy between the two speed decisions. The scatterplot below shows the real-time (X-axis) and slow-motion (Y-axis) accuracy rates for each of the 60 incidents, averaged across participants.
It is clear that the judgments were correlated across speeds: not surprisingly, calls that were easy to make in real time were also easy to make in slow motion, and vice versa. The orange dot with 95% confidence intervals denotes the mean.
Turning to the severity of the disciplinary measure that was awarded for a foul, the data revealed two effects: first, the “correct” (reference) decision mattered, and the speed at which the incident was presented also affected severity. The two effects are presented jointly in the next figure.
Each panel in the figure shows the results for a different reference decision, where NC stands for no card (no sanction); YC for yellow card (warning); and RC for red card (removed from game). The small dots represent the raw data and can be ignored for now. The larger plotting symbols (black circles for real time and orange triangles for slow motion) show participants’ mean responses, conditional upon each reference decision.
Ignoring the video speed for starters, it is clear that participants mirrored the reference decisions: when the “correct” decision was a red card (rightmost panel), the participants handed out a red or a yellow card but rarely opted to forego a sanction. When the “correct” decision was a yellow card (center panel), participants rarely awarded a red card and rarely let the incident pass unpunished, and so on. This suggests that the participants were well calibrated to the reference decisions, as one would expect from elite experts.
The second notable aspect of the results is that the penalties handed out by participants in the slow-motion condition tended to be harsher than those handed out by participants in the real-time condition, irrespective of the reference decision. Specifically, when the reference decision was a red card (right panel), the same sanction was applied considerably more frequently with slow-motion presentation than in real time (and correspondingly, fewer yellow cards were awarded with slow motion than in real time). Likewise, when the reference decision was to withhold a sanction (left), the slow-motion video triggered more yellow cards than the real-time version. Finally, when the reference decision was a yellow card (center), slow motion entailed more red cards than real time.
If accuracy did not differ with presentation speed, why was the punishment so much harsher in slow motion?
Key to this result is the recognition that referee decisions—much like judicial verdicts—involve a component of judging the intentionality underlying an incident. An accidental slip will attract a lesser sanction than a wilful kick. By extending the time of an incident (by a factor of 4 in the present study), slow motion can facilitate the perception that the offender has much more time to contemplate and plan his actions than he actually does. In consequence, slow-motion physical contact may seem to reflect greater intentionality than when the event is witnessed in real time.
One implication of these results is that, theoretically speaking, a player would be better off committing an obvious foul that does not require a slow-motion replay to be nailed down. Being subtle may not pay off if one does get caught.