Effects of visual occlusion by the wall on goalkeeper performance during free kicks in football 5 6

22 Free kicks are an important goal scoring opportunity in football. It is an unwritten rule that the 23 goalkeeper places a wall of defending players with the aim of making this harder for the attacking 24 team. However, the defensive wall can occlude the movements of the kicker, as well as the initial part 25 of the ball trajectory. A ball coming into view later will likely delay movement initiation and possibly 26 affect performance. Here, we used virtual reality to investigate the effect of the visual occlusion of the 27 initial ball trajectory by the wall on the performance of expert and novice goalkeepers. We showed 28 that movements were initiated significantly later when the wall was present, but not by the same 29 amount as the duration of occlusion (~200ms, versus a movement delay of ~70-90ms); movements 30 were thus initiated sooner after the ball came into view, based on less accumulated information. For 31 both novice and expert goalkeepers this delayed initiation significantly affected performance (i.e., 1.532 3.6cm larger spatial hand error, 7-8% fewer interceptions). These performance reductions were 33 significantly larger for shorter flight times, reaching increased spatial errors of 2.8-4.5cm and 11-13% 34 fewer interceptions. Further analyses showed that the wall-induced performance reduction did not 35 differ significantly between non-curved and curved free kicks. The wall induced early movement 36 biases, but only for ball trajectories with curve in the same direction as the required movement; the 37 effect of these biases was negative, away from the final ball position. In two additional experiments 38 (with variations in initial forward and sideward ball position, respectively) largely consistent results 39 were found (i.e., the wall-induced performance reduction was smaller and failed to reach significance 40 with variations in initial sideward ball position). These results may suggest that, certainly for expert 41 free kick takers who are very good at shooting hard free kicks that avoid the wall (reducing its function 42 as a block), goalkeepers could consider omitting the wall. 43 44 45

Introduction effects on performance. 81 The effects of occluding the initial part of the ball flight using a wall may be apparent to 82 goalkeepers, but their effects have never been scientifically investigated -but see Brault   It is hypothesised, that due to advanced visual and action capabilities, expert goalkeepers will be able 87 to overcome some, if not all, of the negative consequences of ball occlusion caused by the wall. Our 88 main goal was to examine the effect of the occlusion; we evaluated the effects separately for novices 89 and experts, without explicitly, statistically comparing the groups. As this study focused uniquely on 90 the occlusion of the ball trajectory; our simulator did not include a visual representation of the kicker. 91 We tested the effects of the wall across a realistic range of ball trajectories, which varied in terms of 92 their flight time, spin-induced sideward curve, and the position they entered the goal. We explored 93 how the effects of visual occlusion by a defensive wall on performance was modulated by these ball 94 flight parameters. In two additional experiments we varied the initial forward and sideward ball 95 position to assess the effects of the wall across a wider range of trajectories.  Experimental set-up 105 The study took place in a large laboratory designed to capture human movements of large amplitudes. 106 All experiments were conducted using an HTC VIVE immersive, interactive virtual reality system. 107 This system consists of a head-mounded display (one AMOLED screen per eye, 3.6'' diagonal,  (https://osf.io/s2bc9/). Note that the simulations were based on a ball with a 0.22m diameter, which 133 deviated slightly from the used visible ball size (see above). The simulations ended when the ball was 134 around 2m past the goal line. All 3D ball trajectories were saved in separate text files (90Hz) loaded 135 prior to each trial; the associated rotational velocities to be applied to the virtual ball (fixed values per 136 trial) were saved in (and thus loaded from) individual trial sequence files (txt format), which also 137 contained the relevant details for each trial. This loading, as well as the trial sequence, environment 138 physics, sounds, ball motion, wall visibility, and data saving were controlled using C# scripts attached 139 to virtual objects in Unity.

140
Free kicks originated from a distance of 23 meters, exactly in the centre of the goal -which 141 tends to be harder for the goalkeeper to stop [1]. There were 18 unique, experimental ball trajectories: 142 3 sideward arrival positions (1.5m, 0m, -1.5m [all entering the goal at a height of 1.75m], three levels 143 of sideward curve (leftward, none, rightward), and two flight times (1s and 1.2s, defined as the duration 144 between ball release and the ball centre passing the middle of the goal line). Sideward curve was 145 generated using ball spin. Leftward curve involved counter-clockwise spin, no curve evidently 146 involved no spin, and rightward curve involved clockwise spin. The spin axis was determined as 147 follows: 1) tilt a vertical axis 15deg away from the goal around the x-axis, 2) rotate the resulting axis 148 around the vertical y-axis towards the final lateral ball position (angles: -3.73deg., 0deg., 3.73deg.).     360 trajectories (note that due to a glitch in the Unity code, the first practice trial was never saved).

193
The order of the 360 trials was randomized for each participant. Breaks were taken after each 120   We only analysed the 180 experimental trials of which some trials were excluded for various 207 reasons. To account for frame updates being missed, we required all inter-frame intervals to be .5 and 208 1.5 times the ideal inter-frame interval (1/90s; 90Hz being the targeted Unity frame rate). To remove 209 trials where the tracking of either of the hands failed temporarily (which would result in a big position 210 jump once tracking was successful again), we required the 3D distance covered by the hands between 211 successive frames not to exceed 25mm (which visual inspection indicated consistently removed 212 erroneous trials). An additional requirement was that these distances should not be ≥5 times larger than 213 the average of the distances for the preceding and next inter-frame intervals (applied separately for 214 each hand). The latter algorithm was optimized by trial-and-error based on visual inspection. To   considered, and because AE is a more adequate, continuous measure of performance. This is apparent 258 in the Results section, which presents success rates for all comparisons involving AE.

259
For Prediction 3 (i.e., larger wall-induced performance reduction for shorter flight times) we 260 separately calculate and compared the change in AE due to the presence of the wall (i.e., AE with a 261 wall minus AE without a wall) between the two flight times (1.0s and 1.2s).

263
The effect of the wall for different types of free kick 264 We further explored how the behavioural effects of the wall varied across ball trajectories. We

328
This study aimed to examine how goalkeeper behavior in free kick scenarios is influenced by     The effect of the wall for different aspects of the free kick 386 It is unlikely that the discussed effects of the wall are the same for all free kick trajectories, as  Table 1 for success rates).  As discussed, we observed that the wall negatively affected performance for both novice and 421 expert goalkeepers. We also wondered whether there would be a difference between non-curved and 422 curved free kicks in this respect. We found that the effect of the wall on performance did not differ Both novice and expert participants showed a smaller early sideward bias in the direction of the curve 432 with a wall, compared to without (novices: t(14) = 9.48, pHS = 2.9•10 -6 , average 6.0cm reduction, Fig   433   6A; experts: t(9) = 8.54, pHS = 1.7•10 -4 , average 5.8cm reduction, Fig 6B). for AE appeared to follow the early bias: the wall had a more negative effect on performance for 461 congruent than incongruent trials (novices: W = -114, pHS = 0.0037, median difference 6.4cm, Fig 7C   462 and see Table 1 for success rates). For experts, however, AE did not differ between the congruent and  Table 1 for success 464 rates). The early movement is affected more by the wall for congruent trials and novices were unable 465 to correct for this during the movement. The success rates, however, do highlight that free kicks with 466 incongruent directions of required movement and curve were the more difficult ones to block in 467 general.  Both novices and expert goalkeepers delayed their movement onset when the wall was present.

506
Although the occlusion lasted ~200ms, the delay was in the order of 70-90ms, implying that 507 movements were initiated sooner after the ball came into view (and were thus based on less 508 accumulated visual information). This, combined with the shorter time to intercept, resulted in 1.5-509 3.6cm increase in the spatial error (7-8% fewer interceptions). Performance reductions were even more  This underscores the relevance and consistency of our findings, also given that the degree of curve 551 used was realistic but relatively large.

552
It is interesting to examine by what mechanisms the wall affects performance. Clearly, not 553 seeing the initial ball trajectory means at any moment the goalkeeper has less accumulated visual 554 information. Seeing the ball later also affected kinematic details of the movement. Initiation was 555 delayed due to the wall, which means less time was available to perform the interception. Besides 556 initiation, the early spatial features of the movement were also affected by the wall for both novices 557 and experts, and more so for trajectories curving in the direction of required movement. For novices 558 these early biases translated into significantly greater performance reductions for these specific 559 conditions (for experts they did not). As evident from the discussion above, we found that the effect  In conclusion, we used a virtual reality goalkeeping simulator to show conclusively that visual 590 occlusion by the wall in free kick scenarios in football negatively affects goalkeeping performance.

591
The effect was substantial and consistent enough to warrant the suggestion that goalkeepers could 592 consider removing the wall in some instances, specifically when dealing with expert free kick takers 593 whose hard shots are mostly on goal and hardly ever blocked by the wall. Scientifically, the interesting 594

708
In the following, we will describe only those aspects of the methods that differ from the main 709 experiment (see the main manuscript for all the other details).