A Re-Evaluation of the Serve Ratings/ Weighted Average

May 17, 2021

8 min read

Introduction

An advanced way to measure the service performance is to grade each service attempt. The traditional way is to rate the serve on a 4 point scale ranging from 4 for a service ace and 0 for a service error. The way you define each point on the service scale is usually the inverse of the quality of the resulting pass by the opposition team, e.g. a reception error means the server gets an ace (click on link for more info). The weights we give to each serve quality is also usually just the inverse of the quality of the resulting pass by the opposition team. Once you grade each service attempt, you use a weighted average calculation to determine a service rating. This is where each serve quality is multiplied by a predetermined weight, which shows the degree of importance, before it is divided by the total service attempts. The traditional The traditional weights and calculation for each serve quality can be seen below:weights and calculation for each serve quality can be seen below:

We traditionally assign one extra point in weight for every increase in serve quality. However, the weight should reflect its actual importance but the above weights seemed to be assigned arbitrarily. There’s really no limit in how we assign the weights. We can extend the range from -10 for an service error to +50 for an ace if we wanted. We also don’t need to have equal increments between each serve quality, e.g. we could have a +5 increase between an error and a negative serve but a +1.5 increase between a negative serve and a neutral serve.

Weights used for serve quality is usually the inverse of weights for pass quality so we could simply inverse the recently re-evaluated weights for reception quality, which was based on the point scoring efficiency percentage after reception (or first ball sideout percentage) you would expected after each reception quality (see link to read more). I came up with the weights below:

An issue with this approach is there isn’t a lot of point scoring opportunities for the serving team outside of an ace (only the block kill). Also, it doesn’t take into account that higher quality serving may result in better counter-attacking opportunities because it would be easier to defend the receiving team’s limiting attacking options. A third issue is that an ace has a 100% break point winning rate but is only worth 1 point more than a positive serve. So more weight should be given to scoring an ace. Similarly, a service error has is a 100% break point losing rate but is only 1 point less than a negative serve. So a lot less weight should be given to commiting a service error.

Method

The measurement I decided to use is the break point efficiency percentage in complex I + II (BPE%). Break point refers to rally points was won by the serving team. This is regardless of whether the rally point was won via an action of the serving team, e.g. ace, or an action the receiving team, e.g. attack error. While I prefer performance indicators to measure actions that the team has control over, I wanted to know which team won the rally point. Efficiency is the net break point won, i.e. break points won minus break points lost. Complex I refers to the attack after reception phase, i.e. serve-reception-set-attack, and complex II refers to the immediate counterattack phase, i.e. serve-block/ defence-set-attack. So this performance indicators measure the percentage net break points won in the complex I and II phase.

I compiled data from the 2019-20 season for all teams plus 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 break point efficiency percentage in complex I + II for each quality of serve, I translated it into a scalable weight between 0 to 4, so it matches the original service weights.

ResultsResultsResults

Below is a table showing the break point efficiency percentage (BPE%) in complex I and II separated by quality of serve and by competition. You can see the better quality of serve, the better the probability scoring a net break point. This should not be a surprise. It also looks like you are likely to score a net break point after a positive serve (negative pass).

When I compiled all the data together, the weights I gave from negative serves to neutral serve to positive serve is in +0.3 increments (or +15% in BPE%) with a +0.4 increment (or +20% in BPE%) between a positive pass and overpass. This is much smaller than the traditional 1 point increments. Also, there a big jump of 1.6 between service error and negative serve and a 1.4 jump between a overpass serve and an ace. This is again much larger than the traditional 1 point increments. Also, it highlights the importance of aces and service errors, which makes sense they directly led to a rally point won or lost. from negative serves to neutral serve to positive serve is in +0.3 increments (or +15% in BPE%) with a +0.4 increment (or +20% in BPE%) between a positive pass and overpass. This is much smaller than the traditional 1 point increments. Also, there a big jump of 1.6 between service error and negative serve and a 1.4 jump between a overpass serve and an ace. This is again much larger than the traditional 1 point increments. Also, it highlights the importance of aces and service errors, which makes sense they directly led to a rally point won or lost.

The break point efficiency percentage in complex I + II separated by quality of serve and by competition. Click to enlarge.

Discussion

The new weights given for negative to overpass serves are within a range 1.6 to 2.6 points. By squeezing the weights of serves that did not end the rally to within 1 point, it means that aces and errors has a lot more importance in relative terms. This is demonstrated in the chart below. More aces can really drive up the service rating and more errors can really drive down the service rating. The increments in weight between the negative serve and overpass serve is also very similar (+0.3 or +0.4), which does reflect the passing increments in weight (see link).

So how does this look like in practice? Below is a table showing the Top 10 Servers in the Women’s Super League (2019-20) using the new serve ratings with at least 50 service attempts. The full list can be seen at the end of the post. The new serve rating is about 0.22 lower on average compared to the old serve rating. There’s a massive jump of 13 spots by #10 Molly O’Hagan. A big reason for this jump is because she only committed 4 service errors in 131 service attempts. This is a service error rate of only 3.1%, which happens to be the lowest in the league with 50 service attempts.

Limitations

A big disadvantage if we adopt these weights is it would change our reference points of what constitutes a good or poor level of performance. We might think serving 2.40 is great but is serving 2.10 any good now? I wasn’t so good with knowing the levels of serving performance with the old system so this will make it even messier. I don’t know what the average player or coach can decipher it.

Another danger in using these weights is 59% of the data is based on the Women’s Super League and only 15% of the data is based on men’s volleyball. If I am to find a global weight to use for men’s and women’s then I would definitely would need some more men’s data.

It could be argued that we should use specific weights for specific leagues rather than a global weight. This could mean we could discover the balance of whether to serve more “safely” or more “aggressively”. Again, we would run into the same problem of needing a lot of data to establish those weights, which we probably won’t have. Also, by having different weights for different leagues, reference points of what constitutes a good or poor level of performance just goes out of the window. I am involved in three different leagues so I would have trouble keeping up.

Servers Ranked by the New Serve Rating | Women’s Super League 2019-20

Rank (New)	Player	Team	Serve Rating (New)	Serve Rating (Old)	Rank (Old)
1	Georgia McGovern	Bristol VC 1	2.29	2.67	1
2	Nina VOGRIC	London Inter Orcas	2.27	2.57	2
3	Suzanne HORNER	TeamBU Wessex	2.24	2.52	4
4	Katerina BOLETOU	Malory Eagles UEL	2.18	2.55	3
5	Ramey GARDNER	Malory Eagles UEL	2.17	2.45	7
6	Brianna ROOT	Tendring (Essex)	2.16	2.45	8
7	Susana VEIGA	Durham Palatinates	2.16	2.50	6
8	Abby BECKER	Durham Palatinates	2.15	2.51	5
9	Michaele PING	Tendring (Essex)	2.14	2.37	13
10	Molly O’HAGAN	Nottingham Uni	2.13	2.25	23
11	Mason ROONEY	Durham Palatinates	2.12	2.38	12
12	Constanza ROSSI	Polonia SideOut	2.12	2.40	9
13	Ezster NAGY	Polonia SideOut	2.11	2.34	15
14	Yulia TOKARENKO	Polonia SideOut	2.11	2.38	11
15	Maria REVILLA	London Inter Orcas	2.09	2.29	21
16	Eleni KAKOURI	Nottingham Uni	2.09	2.40	10
17	Kathia RUD	London Inter Orcas	2.08	2.31	17
18	Molly VAN ESSEN	Tendring (Essex)	2.08	2.31	18
19	Stephanie SIERRA	Malory Eagles UEL	2.07	2.37	14
20	Angelika SIERGIEJ	London Inter Orcas	2.05	2.32	16
21	Katya TSANKOVA	Polonia SideOut	2.04	2.17	31
22	Bruna BON	Durham Palatinates	2.03	2.30	20
23	Kelsee SAWYER	Nottingham Uni	2.03	2.29	22
24	Duygu YETKINCAN	Polonia SideOut	2.01	2.18	28
25	Sierra WILSON	Malory Eagles UEL	2.01	2.12	37
26	Ana VILFAN	Polonia SideOut	2.00	2.31	19
27	Madison HUNZIKER	Tendring (Essex)	2.00	2.22	27
28	Lizz DONALDSON	Bristol VC 1	2.00	2.06	42
29	Shantell ARNOLD	Malory Eagles UEL	1.97	2.14	34
30	Madison WEG	TeamBU Wessex	1.96	2.24	24
31	Katarzyna PLUCINSKA	Polonia SideOut	1.96	2.17	30
32	Steffi NEUN	Cambridge & ARU	1.95	2.22	26
33	Jana PARMOVA	London Inter Orcas	1.95	2.24	25
34	Liene KNUPKE	Bristol VC 1	1.94	2.15	33
35	Elizabeth MANN	Nottingham Uni	1.93	1.96	49
36	Mechell DANIEL	Tendring (Essex)	1.93	2.04	44
37	Lara PICHT	TeamBU Wessex	1.92	2.16	32
38	Brianna GARDNER	TeamBU Wessex	1.91	2.18	29
39	Laura FUNK	Cambridge & ARU	1.91	2.14	35
40	Sophie SZYSZKO	Malory Eagles UEL	1.90	2.13	36
41	Federica VITALI	London Inter Orcas	1.88	2.11	38
42	Cara OLDEN	Tendring (Essex)	1.87	2.07	41
43	Marzia DI BELLA-NEGI	Bristol VC 1	1.85	1.78	59
44	Kendra ZWART	Durham Palatinates	1.85	2.09	40
45	Maria SVIRIDOVA	Polonia SideOut	1.85	2.05	43
46	Rebecca RICE	Nottingham Uni	1.84	2.02	45
47	Mattea GOALEN	Nottingham Uni	1.84	2.09	39
48	Francesca TANTUSSI	Cambridge & ARU	1.83	1.97	48
49	Nicole BISHOP	Tendring (Essex)	1.82	1.98	47
50	Agne VAICIULTYE	Cambridge & ARU	1.81	2.02	46
51	Ada ADAMEK	Polonia SideOut	1.79	1.93	51
52	Chiara RANZUGLIA	London Inter Orcas	1.77	1.88	57
53	Brittany WELSH	Durham Palatinates	1.76	1.89	55
54	Amber DAVIES	TeamBU Wessex	1.74	1.90	53
55	Gabriela DURCZYNSKI	Bristol VC 1	1.73	1.96	50
56	Anna KEEFE	Nottingham Uni	1.73	1.89	54
57	Petra LINDOVSKA	Cambridge & ARU	1.72	1.73	61
58	Malgorzata BIENIEK	Cambridge & ARU	1.72	1.88	56
59	Ebonie HICKS	TeamBU Wessex	1.66	1.75	60
60	Shona FRASER	Durham Palatinates	1.66	1.87	58
61	Lea GOSTINCAR	Nottingham Uni	1.63	1.92	52

Demolition crews are breaking ground on an exciting redevelopment project this week, marking the beginning of a bright future for the former restaurant site and surrounding properties, led by Carmel

Mastering the Markets: The Art of Algorithmic Forex Strategy Development

Get ready to dry your hair the right way! 💨 With customizable features and a sleek design, the Parlux 3200 Plus Hair Dryer gives you control like never before. Don’t wait—check it out via our bio