.png)
Establishing Weights For the Quality of the Pass | Addressing the Weight of Reception Errors and Overpasses
Introduction
I looked into re-evaluating the weights given used in measuring the passing performance (see link). Rather than giving out artbitrary weights from 0 to 3 or 4, I linked them to the Points Scored Efficiency % after reception [PE% After Reception = (Points Scored – Errors)/ Total Reception Attempts] as a proxy of each pass quality’s importance (I called it the Modified First Ball SideOut Efficiency % in the original post but calculation is really the PE% After Reception because it only considered points won and lost by the receiving team only). It worked out quite nicely that for every increase in pass quality, there was about a +9% increase in PE%, which translated to 1 point increments, as seen below.t worked out quite nicely that for every increase in pass quality, there was about a +9% increase in PE%, which translated to 1 point increments, as seen below.
PE% | Weight | Pass Quality 24% | 4 | Perfect 19% | 3 | Positive 10% | 2 | Neutral 1% | 1 | Negative 1% | 0.5 | Overpass -100% | 0 | Reception error
One thing that bothered me, which I did note in the original blog post, was that reception errors only had a 0 point weight. Reception errors is actually a -100% PE% since you lose a direct point by definition. If 1 point is the equivalent of a 9% difference in PE% then a -100% is actually worth -11 points in weight (-9% * 11). Although a -11 weight is unworkable and probably unsellable to coaches and players, we still need to address the fact that a reception error is really bad. Even with a negative pass, you still have a 1% chance of scoring a net point straight after reception.
The other issue is with overpasses. With the PE% calculation, you can still score points by block killing the overpass attack (and lose points via block errors), thus why you can get a 1% PE% after an overpass. However, the chances of losing the rally itself through an opposition overpass attack is quite high but it is not taken into consideration with the PE% calculation. I did address it by giving a 0.5 weight but that is an arbitrary weight, something I wanted to avoid, beacuseI did address it by giving a 0.5 weight but that is an arbitrary weight, something I wanted to avoid, beacuse I could have given it a 0.25 weight if I wanted to. I could have given it a 0.25 weight if I wanted to.
So I wanted to revisit this topic again to see if I could address these two issues. I could have gone back to the data and change the weights to take into account the -100% PE% for the reception errors, e.g. +25% in PE% is the equivalent to 1 point in weight. However, this doesn’t take into account the overpass issue. I then thought that I could use a true First Ball SideOut Efficiency Percentage (FBSOE%), i.e. which team won the rally point after reception. I don’t use the SideOut Percentage in my analyses because it involves opposition errors that the receiving team can’t control for (see link for more information). However, I can justify it, at least in my mind, because the weights would be applied equally to all players so it is fair. So by using the FBSOE%, this shouldn’t change too much for negative, neutral, positive and perfect passes (opposition block errors that wasn’t kills are now included, i.e. easily defended roll shots but the refereee called for a net touch on the blocker) but it should change for overpasses since now kills by the opposition is now included as part of the calculation. he FBSOE%, this shouldn’t change too much for negative, neutral, positive and perfect passes (opposition block errors that wasn’t kills are now included, i.e. easily defended roll shots but the refereee called for a net touch on the blocker) but it should change for overpasses since now kills by the opposition is now included as part of the calculation.
Method
I will use the First Ball SideOut Efficiency Percentage as a proxy measure for the importance of each pass quality. The sideout refers to rally points won by the passing team. This is regardless of whether the passing or serving team was responsible for winning the rally point. The first ball refers to the point scoring after reception only and not beyond. I didn’t want to go beyond that beause the longer the rally, the less of an effect the first pass will have at the end result of the rally. For the overpass, first balls also include the immediate overpass attack by the serving team. I excluded attacks by the serving team after a digging an overpass, i.e. the sequence of overpass-dig-set-attack. The efficiency percentage refers to the net rally points won, i.e. rally points won – rally points lost. This is because I generally find that points won AND points lost are better associated to success in volleyball. The FBSOE% is calculated as follows:
FBSO% = (Points Won After Reception – Points Lost After Reception)/ Total Reception Attempts
Points Won After Reception = Kills + Free Ball Kills + Block Kills + Opp. Attack Errors
Points Lost After Reception = Attack Errors + Attack Block Killed + Free Ball Errors + Free Balls Block Killed + Setting Errors + Block Errors + Reception Errors + Opp. Attack Kills
I compiled data from the 2019-20 season for all teams with 2 Champions League matches in the 2020-21 season. I did not go further back because I did not stat neutral passes. Once I calculated the first ball sideout efficiency percentage for each quality of pass, I translated it into a scalable weight between 0 to 4, so it matches the current passing weights.
Results
The First Ball SideOut Efficiency Percentage can be seen below. The FBSOE% is very similar to the PE% from negative passes to perfect passes. However, the FBSOE% for overpass is -26%. This means there’s a 26% of immediately losing the point through an immediate overpass attack.
When I scaled the weights to 0 to 4, a 0.1 of weight translates to around 3.1% difference in FBSOE%. You can see the negative passes to perfect passes are bunched up from 3.2 to 4.0. The weight drops -0.9 to 2.3 for an overpass, since you are more likely to lose a point. Then there is a very steep drop of -2.3 to 0.0 for a reception error. This means reception errors will really drop a player’s and team’s overall reception rating.
Discussion
These weights are super interesting. It gives an indication that it is a lot more important to avoid reception errors than it is to have perfect passes. A handful of reception errors will really hurt your overall service rating. If you just keep the ball in play on your side, i.e. negative passes, you would have a minimum service rating of 3.2. This is of course easier said than done. Also, increasing the quality of the pass from a negative pass will only give you small increases in serve rating since there’s a +0.3 and +0.2 in weight.
So how does these weights look like in practice. Below is the top 10 passers in the Women’s Super League 2019-20 with at least 50 passing attempts. The full list can be found at the end of the post. It still looks like there are big differences between the best passers (3.53) and the worst passers (2.76). The new rankings are largely similar to the old rankings but there are some big jumps. For example, Mechell Daniel of Tendring (Essex) was the 11th best passer under the old weights but now she’s is ranked the third best passer under the new weights. The main reason was Mechell had the lowest reception errors (7 reception errors in 143 reception attempts, 4.9% error rate) which will boost her overall reception rating. The league’s average pass rating is 3.14, which means the average pass is just below a negative pass (3.2).
LimitationsLimitationsLimitations
The data is heavily centred on the Women’s Super League (60%, 185 sets out of 307). This means the weights should work well in that league, while the men has very limited data (15%, 45 sets out of 307). Although I did want weights that could work globally, it may be better to have specific weights for specific levels. However, that would require much more data, data which I don’t have.
Another issue is your reference points of what constitutes a good level of performance goes out of the window when you change weights. For example, would you know if 3.05 passing constitutes a good level of passing? I know I would have difficult in understanding what the new reception rating means in terms of level of performance and if I have difficulty, I don’t know how the average player and coach can interpret it. Also, if we go with my above suggestion to have specific weights for specific levels, this will only add to the confusion.
Use in Practice
I think I will continue to use the 0, 0.5, 1, 2, 3, 4 weights for errors, overpasses, negative, neutral, positive and perfect passes, respectively (see link). There is a nice symmetry and it is somewhat backed up by data. I think more thought is needed on finding these specific weights or whether another measurement on reception performance could be better, e.g. expected sideout rate.
Rank (New) | Player | Team | Pass Rating (New) | Pass Rating (Old) | Rank (Old) |
1 | Elizabeth DOWD | Durham Palatinates | 3.53 | 2.76 | 1 |
2 | Bruna BON | Durham Palatinates | 3.48 | 2.61 | 2 |
3 | Mechell DANIEL | Tendring (Essex) | 3.39 | 2.24 | 11 |
4 | Guilia RASTELLI | Polonia SideOut | 3.38 | 2.47 | 3 |
5 | Ada ADAMEK | Polonia SideOut | 3.38 | 2.35 | 8 |
6 | Mia MERINO | TeamBU Wessex | 3.35 | 2.45 | 4 |
7 | Adriadne SIERRA | Malory Eagles UEL | 3.34 | 2.42 | 5 |
8 | Maggie STEWARD | Cambridge & ARU | 3.32 | 2.38 | 6 |
9 | Angela VAMVOUSKI | London Inter Orcas | 3.29 | 2.21 | 13 |
10 | Anna TSOUMI | London Inter Orcas | 3.28 | 2.37 | 7 |
11 | Katarzyna PLUCINSKA | Polonia SideOut | 3.27 | 2.15 | 18 |
12 | Jana PARMOVA | London Inter Orcas | 3.26 | 2.27 | 10 |
13 | Ausra MOHANDAS | Cambridge & ARU | 3.26 | 2.04 | 25 |
14 | Lara PICHT | TeamBU Wessex | 3.23 | 2.17 | 16 |
15 | Martha BULLEN | Nottingham Uni | 3.23 | 2.10 | 23 |
16 | Sophie SZYSZKO | Malory Eagles UEL | 3.22 | 2.05 | 24 |
17 | Petra LINDOVSKA | Cambridge & ARU | 3.21 | 2.20 | 15 |
18 | Liz LEAHY | Tendring (Essex) | 3.21 | 2.24 | 12 |
19 | Ramey GARDNER | Malory Eagles UEL | 3.20 | 2.14 | 20 |
20 | Chiara RANZUGLIA | London Inter Orcas | 3.17 | 2.14 | 19 |
21 | Elizabeth MANN | Nottingham Uni | 3.16 | 2.21 | 14 |
22 | Yulia TOKARENKO | Polonia SideOut | 3.15 | 2.27 | 9 |
23 | Kelsee SAWYER | Nottingham Uni | 3.15 | 2.13 | 21 |
24 | Nina VOGRIC | London Inter Orcas | 3.15 | 2.17 | 17 |
25 | Susana VEIGA | Durham Palatinates | 3.11 | 2.00 | 28 |
26 | Madison WEG | TeamBU Wessex | 3.09 | 1.95 | 30 |
27 | Madison HUNZIKER | Tendring (Essex) | 3.05 | 1.90 | 31 |
28 | Molly O’HAGAN | Nottingham Uni | 3.04 | 2.12 | 22 |
29 | Heather REITZFELD | Bristol VC 1 | 3.04 | 2.02 | 26 |
30 | Emily BELL | Tendring (Essex) | 3.00 | 1.84 | 32 |
31 | Eleanor BARNETT-WRIGHT | Durham Palatinates | 2.99 | 1.98 | 29 |
32 | Gabriela DURCZYNSKI | Bristol VC 1 | 2.97 | 2.02 | 27 |
33 | Rebecca RICE | Nottingham Uni | 2.97 | 1.81 | 33 |
34 | Laura FUNK | Cambridge & ARU | 2.90 | 1.80 | 34 |
35 | Lizz DONALDSON | Bristol VC 1 | 2.90 | 1.58 | 41 |
36 | Brianna GARDNER | TeamBU Wessex | 2.88 | 1.75 | 37 |
37 | Shona FRASER | Durham Palatinates | 2.86 | 1.68 | 39 |
38 | Gemma MACIA ROMERO | Bristol VC 1 | 2.81 | 1.75 | 36 |
39 | Anna KEEFE | Nottingham Uni | 2.80 | 1.70 | 38 |
40 | Francesca TANTUSSI | Cambridge & ARU | 2.78 | 1.80 | 34 |
41 | Ana VILFAN | Polonia SideOut | 2.76 | 1.59 | 40 |
Comments