Basic raw and compound stats
If you haven’t read the introduction post, be sure to check it out.
In this post I will outline some of the raw and compound statistics that the analysis yields and hint at what these mean and how they can be used for further analyses and how they can guide decision-making regarding pre-game line-up selection and in-game playing tactics.
For this post I will use data from the last major championship for which I have collected data: The Central European Championship 2019. I will focus on the mixed team, since I have data for more games than for the men’s team. We played several games, and I have recorded data from three of them (against Switzerland, Croatia and Austria in the final). I have de-identified the players on the team. For the starting sixes we used a flexible line-up system: two fixed lineups (6+6 players), one with the 6 best players (for sets that we had to win that were against better opponents), and one with the other six players (for sets that were easy or where we had nothing left to lose, so that they would get more playing time). In the fixed line-ups, I played as the caller/playmaker on position 1, and in the best line-up I was the right-winger on position 6. The analyses are performed, not only on a player- and team-level, but also for the starting sixes to give insights into which line-ups are more successful (bearing in mind that the best line-up will also go up against the toughest opponents and vice-versa).
This is the scoring system used in the analysis:
Eliminations include line infringements, invalid throws, and missed suicides, and thus also decrease the player’s points by 1.
TEAM and starting six stats
Total number of sets won
Swedish Mixed Team 24 – 15 All opponents
Total playing time
4490 seconds
Total score difference
+18 points
Number of played sets
40 sets
Team results
Total Points 65 points
Hit percentage 35.6%
Defence percentage 70.1%
Average difference in number of players on court at end of set +1.55 players
Points scored per minute of playing time 0.87 points/min
Results for starting sixes
Playmaker
Wins 11
Draws 0
Losses 2
Score 18
Sets 13
Points/set 1.4
Points 33
Points/min 1.8
Hit% 45.4%
Defence% 67.6%
Playtime 1126s
Relative playtime 25.1%
Avg set time 87s
Players surviving set 3.2
Player difference end of set +2.5
Not playing
Wins 11
Draws 1
Losses 3
Score 16
Sets 15
Points/set 1.1
Points 22
Points/min 0.6
Hit% 29.7%
Defence% 73.8%
Playtime 2132s
Relative playtime 47.5%
Avg set time 142s
Players surviving set 2.7
Player difference end of set +1.5
Right Wing
Wins 6
Draws 1
Losses 5
Score 2
Sets 12
Points/set 0.2
Points 10
Points/min 0.5
Hit% 34.9%
Defence% 65.7%
Playtime 1232s
Relative playtime 27.4%
Avg set time 103s
Players surviving set 2.5
Player difference end of set +0.6
Event counts
These are the event counts for all the players, which form the basis for many of the subsequent analyses.
Playing times
Total play time for match or tournament
Percent of match time, in case you want to make sure time on the court is distributed between players in a particular way
Number of sets played
Average relative playing time, which is determined by how quickly you get eliminated in the sets you play. This is a stat I use for one of the ways to estimate how important a player is for their line-up, which I will explain more about in another post
Point scoring
Raw points, with positive scores indicating a net positive effect on the number of players on court. I’m currently developing an index that looks at how this net score varies when the team is trailing or leading in score or number of players on court, indicating whether the player scores “easy” points, or the points are actually needed for winning, as well as indicating how the player performs under pressure.
Points per minute, which is derived from total points divided by total playing time. This is one of the main outcome measures I use for the other analyses, as it sums up much of what it means to be successful at dodgeball. Team points per minute almost entirely predicts whether the team wins or loses.
Technical skill
Hit percentage is calculated from the number of eliminated opponents divided by the number of throws
Defence percentage is one minus the number of times one is eliminated divided by the number of times one has been thrown at
Obviously, both of these measures are more stable the higher the number of throws. However, they are inherently unstable since a player can be left alone on court for a one-on-one with a lot of exchanged single throws that decrease hit percentage and increase defence percentage. To mitigate that effect, I’ve created an index for overall technical skill by taking the average of the two, which partly decreases the effect that such situations can have on an individual player’s percentages.
I have extracted the hit percentages for all players for different situations, such as planned attacks (both when one ball is thrown and when several are synchronised), improvised attacks, and against defenders with and without a ball. I have most of them memorised, and before tournaments I usually go over them again if I am going to be the caller. I then use these stats to determine who will throw depending on the situation. For example, if we have four balls and are more players on the court, then I may call the players who have weaker arms, but are good at synchronising, on one of the key opponent players. If we instead have fewer players on court, 2-3 balls in possession, and my teammates have good arms, I try to increase the tempo on the court to score points by countering without necessarily making calls, but allowing for improvised plays. There are also players that have amazing arms but are terrible at improvising and making snap decisions, which preclude that tactic. There are also those that have bad arms, but incredible timing, accuracy and ability to read the game, and these players can fit into a countering tactic, but may require drawing attention away from them with pump fakes so that they can get further up on the court. Most callers intuitively understand these aspects, but the subjective sense of teammates’ individual stats and abilities can be miles off what is actually the case in terms of stats and net effects on court.
I estimate the player’s technical skill by taking the average of attack and defence percentage.
I’ve also created a score for corrected technical skill which has a higher correlation with point scoring ability and probability of winning the set/match. Read this post to see how it is derived, but suffice to say that I did a linear regression analysis to see how each of hit and defence percentage determine point scoring ability, and then weighted the variables accordingly.
As you can see, Male5 has by far the highest technical ability with a corrected value of 67%. Despite this the number of points per minute is low (and in particular the tactical skill below). I will write a case report on what is going on here in a future post. Dodgeball is a non-linear game, where a higher percentage does not automatically mean more points, but rather there is a sweet spot where the rate of return is optimal, and this can be estimated at least to some extent. Stay tuned…
Tactical skill
By regressing out the effect of technical skill on the ability to score points, one is left with “everything else”, which is rather broad, but can be deconstructed into specific predictable stats. A large part of this is determined by the ability of the player to make the right decision regarding whether to throw a ball or not when attacking (knowing the difference between one’s own attaching/dodging percentage relative to that of the opponent), and positioning on the court during defence. I will devote a future post to explain this in more detail.