Go to any team site and you’ll see Team, Individual, Match by Match, and Leader stat links sub-divided into Overall, Conference, Team, Opponent, Offensive, & Defensive categories. Click on any to see only tables. Many times, tables embedded within other tables, even. I started counting cells of data from one site recently and quit when I passed 3,000. The first three rules of data analysis might be: [1] Plot the Data [2] Plot the Data [3] Plot the Data, but what don’t you see when you click? Plots! Presenting results of analysis in numbers rather than pictures is widely considered bad practice.
Volleyball stats are either real time to inform decisions in the moment or intended to be summative after matches to inform growth opportunities in subsequent practice environments. They’re most pertinent encapsulated by a single match or multiple matches against the same or similar opponents. In other words, attempts to combine statistical measures across a whole season of varied opponents can be as futile as using win-loss records of teams to determine their rank order in the landscape. Strength of opponents matter because the skills that drive statistics are not performed in a vacuum. i.e. MLB hitters are far more skilled than those from 50 years ago, but still, they average about .250 at the plate because pitchers are better, too.
Most ironic is data drawn from a single volleyball match lacks veracity because of too small sample sizes, while that determined by accumulating season-long tallies to provide large enough sample sizes can often be invalid to use for comparing player performance. Many have a tendency to overlook or ignore stats which contradict their beliefs and others show a willingness to respond, only to act prematurely. In the hands of either of them, misinformation can rule the day. This why so many coaches bring on data scientists who’d be less likely to fall prey to these polar and fallacious responses to incomplete or insufficient evidence.
Every statistic measured in volleyball relates to one of two things. The first, a measure for the quality of any individual player’s touch of a ball as it relates to placement & speed to what transpires directly after, or some combination therein as demonstrated by a number or proportional rate. (Serve, Pass, Dig, Set & Assist, Attack & Kill, Block) The second always relates to measuring the success of team outcomes determined by a composite of these touches. (Side-Out Rate, First Balls, Rotational plus/minus, etc.) The former usually an attempt to separate the contribution of any individual component(s) to the latter. Unfortunately, some stats are suspended on the blurry edge between. Assists for example. Does an assist measure an individual player’s touch even though it gets assigned to one? Or is it a measurement for team effectiveness being dependent on the pass or dig before it and the swing coming after?
Perhaps some are wondering why this rate is a function of only live-ball side-out opportunities. First, about 80% of all assists happen in side-out. Second, I am of the belief any assist that comes with the benefit for a point score deserves to be treated as a “bonus” in the calculation. No team intends to generate an assist when they serve a ball. It hopes to serve in a manner to hinder the receive from scoring. (It being a defensive act even though I always see it categorized in Offensive tables.) It can happen in the form of forcing a receive error or blocking the attack, but only after these have failed to be accomplished does the gift of a potential transitional point present itself. Even in the very moment I watch this unfold on the court my emotional response treats it like a bonus, if I’m being honest! Oh! And being easily calculated from data offered in box scores certainly doesn’t hurt, either! LOL
If you look at the rate of change of my preferred assist rate as it pertains to the #1 (NYU) & #12 (Loras) team’s setters above, you can see the slope of the NYU’s assist rate is about half that of Loras. (Roughly a 0.3% per ranked position compared to 0.6% per ranked position) A lesser slope in this context simply means the skill level of the opponent factors less on this setter and his team to effectively execute the skills necessary required for success. The actual data points on the graph also go a long way to inform the strength of schedule the players on these teams have competed thus far. The plots narrate a story not possible to be truthfully told by an assist table or God-forbid any assist per set ranking. Now that’s an offensive statistic and I don’t mean it in a way that suggests a volleyball point was scored, either. LOL
However, consider the following that would be less confounding:
Over a year ago I made a case for why Season-long stats which come in the form of a ratio of tallies can be insufficient in comparing performance of players. I’d like to share the case of Simpson’s Paradox as it pertains to a couple OHs on the same team over a month of matches again. The next two slides will show how a combination of tables and rates across the season can offer misinformation and then take the exact same data to demonstrate how looking at it as a plot related to the strength of an opponent can make all the difference.
You read that right. One dude hit 0.100 better every single time and his season long H% rate still lags behind the other by 0.030. Now look below for the plots across the strength of the opponents in these matches to have your eyes see the truth in a way your brain might miss in tabular form.
The slopes of these lines are identical because each player responds equivalently to the strength of an opponent to thwart their attempts to kill the ball. If one was less steep than the other having played with the same setter and a similar supporting cast, it would have indicated a higher resilience for that player in executing the skill. Most importantly, though, the vertical space between the lines is consistently 0.100, just as the premise of this case study began.
This week I focused on a couple of “Individual” statistics to see how plots across the strength of an opponent spectrum can be richer in content and more precisely informing. Next week I intend to offer a couple of “team” statistics as they relate to strength of opponents. These will be based in reality as the first will encompass almost 4 years of St. John Fisher efficiency data as it compares to opponent strength. (One of the reasons when I state in a previous article that Fisher is playing an equivalent SOS this year than last, you most certainly can take it literally.)
I think the second one I share next week is interesting in how it came about. Maybe you will think so, too. Two months ago, I had a conversation with an opposing coach after a scrimmage. One I have known for almost a decade because he coached my son about that long ago. He was very excited to track live first-ball side-outs and set reasonable goals for his team performance in this category. He mentioned the team’s goal was to terminate 35% of live balls before their opponent could gain possession. (1BSO excluding terminal serves by the opponent.) I asked him, “Shouldn’t that goal be variable, dependent on who you are playing? And wouldn’t you want to be slightly more concerned with the rate of those originating from in-system passes? These questions and his willingness to go down the rabbit hole with me set the stage for what I think is a pretty cool data tracking system I will share as the second in a two-part series related to volleyball statistics next week.
The last couple weeks in March I have chosen to focus on volleyball statistics because I can only write about the NPI so much! LOL Many will be pleased to hear I will only go there just one more time the rest of the 2025 season even though I continue a due diligence in studying its behavior in the interim!