So here’s the concept: we want to evaluate each touch by the difference between expectations and results.
- Input Contact: this serves as the expectation – given the situation you’ve been dealt, how well do we expect you to do?
- Output Contact: this serves as the result – looking at both terminal & non-terminal results
Serving, the example above, is a simple version of Efficiency Change. Because serving lacks a true input, we use a constant (orange line) that is the efficiency of the serving team winning/losing the rally on that first ball. Think of it as the inverse of FBSO efficiency:
(Serving Team Won Rally on 1st Ball – Lost Rally on 1st Ball) / (Total Serves)
In the graph, we see that the Input Efficiency is constant at -0.225 (the serving team loses on the 1st ball at a 0.225 eff – aka receiving team wins the 1st ball at a 0.225 eff). This passes the eye test as we assume the serving team is at a disadvantage on that 1st possession.
In the graph, we also see that the Output Efficiencies vary based on the Output Contacts (different R codes & service errors). Aces & errors are logically at +1 and -1 respectively for the serving team, while non-terminal ratings range from perfect pass at -0.307 (serving team is in trouble) to poor pass at 0.180 (where the serving team is actually at an advantage).
These efficiencies can be created with great variation and can reflect a combination of things we might think are important. We can include information for just the team touching the ball or we can include the data about the opponent. We can add rotational information, what kind of route the attacker is on, how intense the match is by using the relative scores, the location of the contact on the court, or any combination of context we choose.
Finally, Efficiency Change:
Output Eff – Input Eff
That’s it. What result do you get, minus what you had to work with. By subtracting these two, we can see value added or removed by a player making this contact.
Server A, expected to win rally at -0.225 and she forces a poor pass (R/). This poor pass has a value of +0.180 for the server. Output minus input. +0.180 – (-0.225) = +0.405. Server A has added 40% of a point by forcing a poor pass.
Server B, also expected to win at a negative rate (-0.225). He goes and rips a jump serve out of bounds, -1. Output minus input. -1 – (-0.225) = -0.775. Although this guy alone made the entire error, because his team was somewhat likely to lose anyway, he doesn’t carry 100% of the lost point.
Server C, still expected to win at the negative rate of -0.225. She goes and drops a short floater for an ace. Output minus input. +1 – (-0.225) = 1.225. This server, because she’s expected to lose even before she makes contact, is able to generate nearly a full point and a quarter for her team. This is why aces are super valuable. They take a slightly negative situation and score you a massive boost in efficiency.
The efficiency change numbers are in the blue bar chart at the bottom of the visual above, but as you can see, they’re merely the difference between the blue & orange lines in the top graph.
Naturally, this style of evaluating contacts can be applied to any skill – and with increasing complexity. If this sounds like your type of rabbit hole, then maybe these posts will be up your alley: