Wing vs. Middle position
Effect of position on attack and defence success rates
Take Home Messages
– Playing on the wing improves hit percentage by 4.8%
– Playing in the middle is associated with a 3.3% lower likelihood of getting eliminated
But, wing players have a more favorable position than those in the middle…
From having showed some of my analysis results and discussing the findings with dodgeballers, one comment I often receive is that the results are biased toward different positions. The idea is that wing-players (those on the far left and right) are advantaged compared to players in the middle, either because they are far away from opponent throwers, or because they are more likely to hold a ball to block with when defending. This is then assumed to inflate the statistics for wing players compared to those in the middle.
As I’ve shown in a previous post, attack and defense percentages are major determinants in terms of game outcome. Players that hit more and get eliminated less win games. With that in mind, we can compare the differences for these percentages for different positions.
Most people assumed that wing players will have much better stats for both attack and defense. Since they are more likely to have a ball in defense, they may have a higher defense percentage, all else equal. Similarly, they may also have a slightly higher attack percentage, since having a ball allows them to counter attack and eliminate an opponent while the opponents are retreating and are still unorganized, which has a higher scoring probability than when they have reached the back line and organized.
Assumption 1) Wing players are less likely to get eliminated in defense
Assumption 2) Wing players have a better attack percentage
I’ve also heard the argument that the statistic “points per minute” will be biased toward wing players. This may be true for positive points (scored in attack) since they may be able to counter and are generally have more active roles. However, this ignores the fact that they are also thrown at more frequently, and their counter attacks can also be countered, leaving them more vulnerable. In the end, a positive hit and defense percentage is what is what will determine the difference between these positions. Having a faster playing style will only increase the magnitude of “points per minute”, but will not influence its direction (positive or negative).
My impression has always been that the differential effect of position is much smaller than expected because of the availability bias, speckled with a bit of substitution bias, fueled by some confirmation bias: the best players are more often on the wings, which is why the stats for wing players are better, and also why people subjectively think the wing position is advantageous. Using the data I’ve collected, I will estimate the objective difference between these positions.
My method for identifying the difference between positions
In my dataset I’ve gathered meta-data for the number of players left on court as well as the position of the player for each action that takes place. Using this I can find out whether an event occurred when the player was playing on the wing position or in the middle. Counting from the left, position 1 is always a wing player. If the player is in position 4, and there are 4 players left on court, then that player is also a (right) wing player. If there are 5 players on court, then they are a middle player.
I’ve used the data from the Mixed team from the Central European 2019 Championships. Most players have played both positions, since when the wing player gets eliminated, a middle player becomes the new wing player. This means that each game contains data for all players for both positions, allowing for an intra-individual comparison between the two positions. In other words, rather than comparing a couple of wing players with a couple of middle players, which would be biased due to confounding variables such as skill, one can directly compare how each position affects each player. Calculating the difference between the positions, one gets an unbiased conclusion regarding the difference between the positions.
Identify the position for each player, whether in the middle or on the wings
For each position (wing and middle) and each player, calculate:
The number of throws
The number of times thrown at
The number of times hitting an opponent
The number of times getting eliminated
To increase the stability of the data, exclude all the players that had less than 5 throws or were thrown at less than 5 times, either on the wing or in the middle
This excluded 4 out of 12 players from the analysis
Calculate for each player
Attack percentage as wing
Attack percentage as middle
Defense percentage as wing
Defense percentage as middle
Subtract percentage for middle from wing for each player
Attack percentage difference = attack as wing – attack as middle
Defense percentage difference = defense as wing – defense as middle
Calculate the average percent differences for the included individuals
The main outcomes are percent differences between wing and middle positions, for attack and defense. A positive score means that the player has a higher success rate for the wing position than for the middle position.
Conclusions
Surprisingly, the original assumptions were not corroborated, and the difference between positions is lower than what most people assume.
Finding 1) When playing on the wing, players have a 4.8% higher hit percentage than when playing in the middle
Finding 2) When playing in the middle, players have a 3.3% higher defense percentage than when playing on the wing
When including the excluded individuals, the numbers barely differ with +2.1% and -4.3% for attack and defense in favor of the wing position. Note that these findings are intra-individual, so there is no bias in terms of player skill. One possible confounding variable may be that middle players become wing players at a later stage in the set, when the team compositions differ, but this is likely to have a very small effect.
Regarding the improved hit percentage on the wing, I assume this is because of the higher probability of ball possession in defense, giving the possibility to counter attack which has a higher probability of hitting an opponent than an organized attack. The wings had 2.8 throws for each throw from the middle.
Regarding the improved defense percentage in the middle, this is despite the fact that I haven’t excluded the effect of wingers having a blocking ball in defense, which should lead to the opposite results. I assume the improved percentage in the middle is because the best players throw and sync more at the wing players, since they tend to be the best players on court. The middle players instead get thrown at more often by un-synced individual throws by worse throwers, causing them to be relatively “safer”. Also, the wing players got thrown at more frequently, with 2.2 times for each throw on a middle player, causing them to be eliminated 90% more.
These small differences in outcome run counter to what people usually assume regarding the different positions. It is clear that any such assumption stems from an observation that is biased due to the fact that the best players are usually placed on the wings.
These findings can be used when determining which position a player should play. Some players are better both at attacking and defending in the middle than on the wings, while others are better in the middle. Also, one can use this to remove the effect of position when comparing the technical skills of players. Subtracting 4,8% from the attack percentage of a player that plays 100% on the wing, will give a result that is comparable to that of a player that played 50% on the wing and had 2,4% subtracted. I have not made this correction for my estimate of technical skill since it already averages the two scores and both positions are advantageous for different skills.
Let me know if you have other assumptions about the game of dodgeball, or if there are tactical aspects that you think are interesting, and I will let you know if I have collected data that allow them to be tested.