I looked into giving appropriate weights for the reception (see here and here) and serve (see here) that truly reflect its importance. Rather than giving arbitrary weights of 0 to 4, I linked them to the point scoring rates you would expect after each serve and pass quality as a proxy of its importance, since it is associated to success in volleyball. The performance indicator I used the for reception is the Point Scoring Efficiency Percentage After Reception (similar to First Ball SideOut %, i.e. net points scored after reception). The performance indicator I used for serving is the Break Point Efficiency Percentage in Complex I and II (i.e. net points scored after reception and the immediate counter attack). See below for resulting weights used for passing and serving.
Pass Quality | Weight (2019) | Weight (2020) | Weight (Proposed 2021)
Perfect | 3 | 4 | 4
Positive | 2 | 3 | 3.8
Neutral | 1.5 | 2 | 3.5
Negative | 1 | 1 | 3.3
Overpass | 1 | 0.5 | 2.3
Reception error | 0 | 0 | 0
Serve Quality | Weight (2019) | Weight (2020) | Weight (Proposed 2021)
Ace | 4 | 4 | 4
Overpass | 3 | 3.5 | 2.6
Positive | 3 | 3 | 2.2
Neutral | 2 | 2 | 1.9
Negative | 1 | 1 | 1.6
Service error | 0 | 0 | 0
Limitations of Using Weights of 0 to 4
I noted in those blog posts that a major issue of using a new set of weights is it changes your reference points of what rating constitutes a good level of performance. For example, I know that when using the 2019 weights, the very best passers in the Women’s Super League will reach or approach 2.00 passing while the worst passers will have 1.50 passing. My reference points have changed when using 2020 weights. I figured out a conversion rate of just subtracting 0.50 (or 0.60 for the top passers) off the new reception rating to gage the old reception rating. However, it still takes a bit of time to work it out. If I was to adopt the proposed 2021 weights then my reference points would change yet again. If I am having difficulty in understanding what number constitutes a good or poor performance then I don’t know how the average coach and player are going to fare.
Using a Different Performance Indicator – Expected Sideout/ Break Point Percentage
An expected sideout percentage is another way to measure passing performance. It is the sideout percentage, or percentage of points won as the receiving team, you can expect from a team’s or player’s passing. For serve, you can use an expected break point percentage. When I looked into the actual calculation, it is simply a weighted average calculation. So rather than using an abstract weight of between 0 to 4, it actually uses the sideout percentage that would occur for each grade in passing quality as the weights themselves. So if the sideout percentage after a perfect pass is 60% and after a positive pass is 50%, then a player who passed a perfect pass and positive pass has an expected sideout percentage of 55%. This was genius to me because now you have a value that players and coaches can relate to. If a player’s passing performance has an expected sideout percentage of 55% then you can expect, based on their passing quality, that you can expect a 55% sideout rate.
Those who read my blog will know that I don’t use sideout percentage in my analyses so why am I using this performance indicator that involves a sideout percentage. This is true and this was actually my initial reaction to not use this performance indicator. However, I figured that any arguments I had against using it is cancelled out because those percentages are applied to all teams and players equally, so it was fair to everybody. Those weights do not change and is not affected by what the actual sideout percentage is. So a player who makes five perfect passes but the outside hitter makes five attack errors (0% sideout) is not punished for it.
I want to compile the league-specific expected sideout rates to measure reception performance and league-specific expected breakpoint rates to measure serving performance.
I will use the sideout percentage and break point percentage for the given quality of serve and pass, respectively. Sideout refers to scoring points as the passing team. Please note that service errors are not counted towards the sideout percentage. The main reason is for this is because you want to know how likely are you to win the rally after a reception attempt and there’s no passing attempt after a service error. Break point refers to scoring points as the serving team. Its calculation can be seen below.
Sideout % = SideOut Points Won/ Total Reception Attempts
Break Point % = Break Points Won/ Total Service Attempts
The resulting performance indicator you can use is the expected sideout percentage (ESO%) and expected break point percentage (EBP%). Expected refers to the percentage you would expect to sideout or break point for any given quality of reception. The calculation is as follows:
Expected Sideout % = [(Perfect Pass * SO%)+(Positive Pass * SO%)+(Neutral Pass * SO%)+(Negative Pass * SO%)+(Overpass * SO%)]/ Total Reception Attempts
Expected Break Point % = [(Ace * BP%)+(Overpass Serve * BP%)+(Positive Serve * BP%)+(Neutral Serve * BP%)+(Negative Serve * BP%)]/ Total Service Attempts
I compiled data for the Women’s Super League (185 sets) and BUCS Tier 1 Women (61 sets). This is because Mark Lebedew stated that expected sideout percentages should be specific to a particular league and I had enough data in those leagues to reduce variability in the the data. I only had a maximum of 19 sets for the other competitions I have been involved in. I only included the 2019-20 season data as that’s when I started to stat neutral passes.
Women’s Super League
Below is the sideout and break point percentage in the Women’s Super League, 2019-20 season. The SO% and BP% after a negative serve is just the combination of after a perfect pass and positive pass. I find it quite interesting that even after a negative pass, there’s a 44% of siding out.
Below is the players’s expected SO% for the top 5 and bottom 5 passers. It looks like the best passers will have an ESO% of around 50+% while the worst passers will have an ESO% of the low 40%. The league average is around 47% expected sideout percentage.
Below is a table showing the best servers in the Women’s Super League (2019-20) with at least 50 service attempts. You can see the best servers has an expected break point % of 58% while the worst has 41%. You may notice that the expected break point % is lower after reception, which may lead you to think it could be better to serve first. However, the expected sideout % does not take into account service errors, which will would increase the team’s overall sideout %.
BUCS Tier 1 Women
The sideout and break point % are actually very similar in the WSL and BUCS with a 1% to 3% difference. The exception is the overpass, with the SO%/ BP% is 11% higher/ lower.
Below is a table showing City University London player’s ESO%. It is generally lower due to the lower level.
Below is a table showing the best servers at City University London (2019-20).
The big surprise is that the best expected sideout rates is around 50% to 54%. This is only a little more compared to the league average of 47%. That’s only slightly better than calling head or tails on a coin toss flip. However, this could be due to the level as the best passers in the PlusLiga has an expected sideout % of 63%. Also, if you simply just get the serve in play, you will still expected to win 40% of break points (43% for City University London).
I’m also not a fan that the difference between the best and worst passers and servers are quite small. It’s only 14% in the Women’s Super League (10% for City University London) for passers and 17% for servers (9% for City University London). If the differences are so small then selecting for the best players, based on reception or serving, is going to tough.
Despite this, I think I will start to use the expected sideout and break point percentage as a measurement for reception and serving performance, respectively. The biggest benefit is that players can understand what it means. However, I can only use it for the Women’s Super League and BUCS Tier 1 Women. If I was to use this performance indicator in the Men’s Super League or in beach volleyball then I would need to collect more data.