I wrote about serve reception in a post a couple of years ago (read it here) and my thoughts there were a bit more philosophical than technical. I want to build on some of the ideas in that post and add some data before giving some ways to integrate the ideas I present. So here goes...
I have been thinking a lot about how we evaluate serve reception since at least when I wrote that previous post and this spring and summer finally yielded some productive ideas that I am looking forward to integrating into my team's training and competition. I'm not changing how I grade receptions but I am expanding how I think about serve reception to include its contribution to scoring points. This expansion is a product of treating all non-terminal skills (receiving, setting, digging) not as isolated skills but as opportunities to either make it harder or easier for our team to score on the next attack. In the case of serve reception, pass average and in system percentage both treat passing as an isolated skill so how can we incorporate scoring into our passing evaluation?
Before I get into that, I think it is important to quickly look at how I grade passing. To be clear, I don't think that my way of grading is better than other ways, it is just an expression of what I think is important and that can vary from one program or scout to another. I think it is important to explain my grading because it influences the data that underlies everything else I'm writing about. First, I grade on a four-point scale so a "four" is a ball passed within a step or so of the setter's "perfect" location while also allowing the setter to be in a desirable posture. Threes, twos, and ones are basically determined by how many options I think the setter reasonably can set on the pass (there's a difference between can and should, which I wrote about here). I use Data Volley's R/ grade for "one-half" receptions, which are passes that are kept in play but the receiving team cannot take a swing. This grade is useful to make "one" grades more connected to scoring points without being affected by the noise of shanks and overpasses that aren't aces.
The data set that I'm using is the last three years of Pac-12 matches for the University of Colorado, where I am the Technical Coordinator. While the data visualization below contains the data for Colorado's opponents as well, I am going to focus on my team. The graph is built in Tableau, which is a really fun and powerful data visualization tool. It is highly interactive, so click around and enjoy.
Let's remember that I said scoring is what drives my work here. Reception is a step towards scoring, which is what really matters. Passing well is nice but good passes are ultimately useful because they make scoring easier. So let's approximate how well my team scores after different reception grades. The graph below is built on reception grades (x axis) and expected first ball efficiency (y axis). To find expected first ball efficiency (xFB), we'll need the number of times each attack outcome (K, 0, E/B) occurs following a particular reception grade and the number of times that reception grade occurs. xFB
will be calculated in the same way we calculate attack efficiency with one crucial difference. I am using reception attempts as the denominator rather than attack attempts because this calculation is about the passer rather than the attacker so I want to include receptions that don't have an attack that follows. Each reception grade has its own calculation so each team has five points on the graph and these points correspond to the xFB for that reception grade for that team. I then asked Tableau to show curves to relate each team's data points to one another so each team has one curve that roughly links their five points together, giving us a sense of how the values change as we move from grade to grade. The xFB changes from grade to grade are what triggered
my thinking around reception evaluation.
The conclusions I draw from this visualization are not exactly universal but I think that they are similar enough that my conclusions about my team can be useful for many other teams. Let's start at the top end of the grade scale, fours and threes. Better than almost any other team in the conference, Colorado shows that there is little difference between these two kinds of receptions in terms of how well we attack after such a reception. That makes sense because there is probably less difference between what fours and threes look like than the difference between any two other grades. (This could be an argument against scoring on a four-point scale but that's not what I want to focus on here.) We see a big gap between xFB on threes/fours and xFB on twos, which is important because that means twos are clearly different than threes and fours in terms of our ability to score. We score less on our first swings when we pass twos than if we pass better. Just like with threes and fours, this makes sense but it is important to see how large a difference in xFB there is (almost 100 points). It is worth noting that, even though there is a drop in xFB, we can still be reasonably successful hitting .250 in our first ball offense but we start to put more pressure on other aspects of our game. There is an even larger drop between twos and ones (from .250 to .104) and now we have entered dangerous territory, it will be really hard to win if we are only hitting .100 in first ball. Not every team in the graph shows the same grouping of xFB values but there is almost always some kind of grouping that is apparent. I think that if I used more than 5-6 matches of data for the non-Colorado teams, the groupings would be more clear but I'm focusing on my team and extrapolating out from there. The important thing to take away from this is that there are reasonable ways that I can group reception grades together when I consider how well teams attack after those reception types.
I see three groups of receptions for CU: ones after which we hit well over .300, ones after which we hit mid-.200s, and ones after which we hit .100 or worse. I see these three groups as being either favorable for, slightly below average for, or poor for scoring. I could easily stop here and just assign new number values to my reception grades but that doesn't connect the skill to scoring in any meaningful way. Why not just use In System Percentage (IS%) to express the same idea? I think there are two reasons, IS% doesn't connect to scoring and IS% ignores that we can still win passing twos. IS% is certainly useful but it doesn't accomplish what I'm looking for. After thinking about alternative scales, I arrived at Green-Yellow-Red (G-Y-R). This scale separates us from reception average-type numbers and gives us a sense of the situations that attackers find themselves in after a reception: favorable, questionable, and difficult. G-Y-R allows me to continue grading serve reception in the same way I have been, as an expression of the number of front row attacking options that are available, but now I have a way to talk about how reception affects scoring.
Here are two different visualizations that are showing almost the same thing, G-Y-R reception frequencies for opponent passers. Each bar in these stacked bar charts represents a different passer and the colors represent how often that type of reception occurred for that passer. The numbers at the top of each bar are numerical expressions of each passer's G-Y-R. The difference between the two graphs is that the first shows gross counts so we can see which passers received the most serves while the second shows receptions as a percentage so that each passer's performance can be easily compared to that of another passer. I show both because I think it is interesting to see if there are particular passers that are above/below average in terms of their number of attempts and I also think it is important to be able to make comparisons regardless of usage rates. I created these plots in R and I am happy to share the code with anyone interested.
So what does a good passer look like in G-Y-R? The obvious answer is that more green is better, as is less red. But that generalization, like IS%, ignores the yellow, the in-between cases, that can make or break teams' first ball success. Compare Player 13 and Player 80. They have almost identical G% but Player 13 has 10% more Y than Player 80. That means that Player 80's team is going to be hitting around .100 10% more often than Player 13's team when each of them passes. That's a difference in siding out that I want to be aware of and I wouldn't see it if I only looked at IS%. But that comparison doesn't answer the basic question of what a good passer looks like in terms of G-Y-R. I think that 50-30-20 would be the sign of an elite passer. In the sample above, Player 41 and Player 89 are, in my opinion, the best of the bunch. There were passers in my sample that met my criteria of 50-30-20 but they had fewer than 40 receptions, likely because we thought they were good passers and tried not to serve them much.
Let's look at data from a single match to see what G-Y-R can look like. The image at right is of a simple Data Volley worksheet I built to compare reception average, xFB, and G-Y-R. The data make it pretty clear that all three are pretty tightly related. The upper team in the worksheet passed 50% of their receptions at a grade of one or lower and that distribution was reflected in the low xFB and reception average values. Meanwhile, the lower team had 50% G and had xFB and reception average values that were in line with that. If the numbers are all so closely aligned, then aren't xFB and G-Y-R superfluous? I don't think so because I don't think that reception average gives us a clear indication of how a team should score. I don't like relying on xFB because it is too easy to conflate its value with reception averages. (Look at how closely the values of the two measures can seem.) Both xFB and reception average boil a passer's performance down to a single number which removes valuable context. Passers are not going to pass every ball the same but xFB and reception average give us the sense that every ball will be the same because they each give a single value. G-Y-R helps us very quickly understand that passes will be different and gives us a sense of how the passes will be distributed. Compare two passers on the upper team, one at 1.97 and the other at 2.00. If I was deciding who to serve at based only on those reception averages, I might be inclined to pick the lower average but the G-Y-R gives me an interesting insight. The 2.00 passer passes 5% more R than the other so I can give my team a much better chance to earn a point 5% more often if I concentrate my serves on that player instead of on the 1.97 passer. Compare that player to the 2.00 player on the lower team whose G-Y-R is 20-60-20. The dramatic differences in Y% and R% are reflected in a 30 point difference in xFB. These are important differences that affect a team's ability to win points in first ball.
I think G-Y-R can also have an impact on how we coach our athletes. I don't think that it changes how we want our athletes to look and move but I do think G-Y-R should shift our interest and energy when working on reception. Knowing that there isn't much difference between a three and a four in terms of scoring, how much energy should we put into improving a pass' location from 6-7 feet off the net to 1-2 feet off? But what about improving pass location from 11-12 feet off to 6-7 feet off? That's a potential difference of 100 points in hitting efficiency. How important is the height of a pass? A low pass to the center of the court may not prevent our setter from getting to it but it may prevent her from setting it with her hands. Passing a ball to that same location but higher can therefore mean a 150 point increase in xFB. To me, this is a different perspective on improvement. Improvement can mean raising the ceiling on our performances, that our average performance improves because our top scores consistently become a little better. G-Y-R gives us a way to quantify the benefit of raising the floor, making our lowest performances better. G-Y-R suggests that raising the floor would have a much bigger effect on our ability to side out than just seeking general improvement that is reflected in reception average.
G-Y-R also give me an alternative way to think about scoring reception games in practice. We play a serve receive offense game at Colorado where points are awarded based on the quality of our pass and the outcome of the first ball attack. We have been using traditional reception grades for scoring but that's going to change so that our team can more clearly see the effect that reception quality has on side out offense.
I am looking forward to exploring G-Y-R with my team in the coming season. It could be the beginning of an important shift in our thinking towards valuing non-terminal skills in relation to scoring.