.png)
Could An Attack Rating Be a Better Version of the Hitting Percentage?
Limitation of the Hitting Percentage
The hitting percentage, calculated as H% = (Kills – Errors)/ Total Attack Attempts, is a widely used volleyball performance indicator and it is highly correlated with success in volleyball. It is one of my go-to statistics that I use for my analyses. Corey Bagley, Director of Analytics at Volleymetrics, made an Corey Bagley, Director of Analytics at Volleymetrics, made an argumentthat a limitation of the hitting pct. is it only measures points ending actions and it ignores the other attack attempts. In real terms, roughly 55% of all attack attempts ends the rally so we are ignoring the other 45% of attack attempts (it’s 52%-48% for the Women’s that a limitation of the hitting pct. is it only measures points ending actions and it ignores the other attack attempts. In real terms, roughly 55% of all attack attempts ends the rally so we are ignoring the other 45% of attack attempts (it’s 52%-48% for the Women’s Super League, 2019-20). I have since seen this argument being made here and here. Super League, 2019-20). I have since seen this argument being made here and here.
When I heard that, it really gave me food for thought. I agree that more weight should be given to an attack attempt that results in the defending team chasing down the second ball than a soft attack attempt that resulted in an easily defended ball which puts the defending team in system for all attack options. You may want to start a player with a slightly lower hitting pct. but consistently puts the defending team out of system more often over a player with a slightly higher hitting pct. but the continuation attack attempts are easily defended.
Using an Attack Weighted Average
A weighted averageperformance indicator can overcome this limitation because it takes every attack attempt into consideration. Every attack attempt is multiplied by a predetermined weight, which shows the degree of importance (e.g. 5 points for a kill, 0 points for an error), before it is divided by the total attempts. It has a similar calculation for the serve and reception. This has the obvious benefit that because now it will reward players who will be able to put teams out of system more often because more weight is given to it, which the hitting pct. does not take into account.
As an example, we can use the suggested weights described below where we can simply assign one extra point for every increase in attack outcome.
Weight | DataVolley Code | Action (Quick Definition)5 | A# | Kill (point won)4 | A+ | Positive Attack (results in a poor dig or overpass)3 | A! | Rebounded Attack (attack is blocked/ rebounded but play continues)2 | A- | Negative Attack (results in a neutral, positive or perfect dig)1 | A/ | Attack Block Killed (attack is blocked killed and point is lost)0 | A= | Error (point lost in error)
Attack Rating = [(A#*5)+(A+*4)+(A!*3)+(A-*2)+(A/*1)]/ Total Attack Attempts
Establishing the Correct Weights
An important part of using the weighted average is assigning appropriate weights that truly reflect the importance of the outcome of each attack. The weights I assigned above are actually quite arbitruary. We can assign any weight in reality. We could assign 11 points for a kill and -3 for an attack error.
The important question is how would you quantify the importance, and thus the weight, of each attack outcome. For the reception, I linked it to the point scoring efficiency percentage you would expect after each grade of reception quality (click on this link to read more). You could possibly do something similar for the attack. You could measure the percentage of winning or losing the point after the next chance to attack, i.e. how well do you block/ defend (to avoid losing a point) and then attack again (to win a point). Or the percentage of winning the rally after each grade of attack, e.g. which is the percentage of winning or losing the rally after a positive attack? We can then use that percentage as an “expected” probability and turn that into an appropriate weight.
I would like to test this idea out with real data but unfortunately I don’t currently stat negative attacks and knowing if an attack is negative (those attacks are currently statted as positive attacks by default) is probaby critical if the attack rating has any value. Also, I don’t know how I would actually export that data off DataVolley as it involves several sequences of play.
Attack Rating or Hitting Pct?
I think I will be sticking with the hitting pct. for now. This is still a popular performance indicator that most players and coaches understand and it is correlated to success in volleyball. The other issue with the attack rating is even if I was able to establish appropriate weights for each grade of attack, the resulting number would be difficult for coaches and players to understand, at least initially. For example, what would a 3.58 attack rating even mean in regards to performance level?
Comments