Wow, I just posted a +/- stat as a scouting basis in another thread, before I noticed this one.
Is +/- perfect? No, absolutely not? Is Blake Wheeler the best player in the NHL? Hell no! Is Alexander Ovechkin 2.4x better than Sidney Crosby? Yes, but that's a different story.
But the core idea is what actually matters here. Four stats should be tracked to get an ideal measurement (one of which can be done after the match by looking at scores), "team" +/-, "robot" +/-, alliance +/-, human player balls scored. Potentially you can even have a super cell +/- category for each of those if you wanted.
Additionally each of these stats should be tracked for the event and each individual match.
Robot +/- is how many balls the robot scored against how many are in it's trailer. Team +/- is the robot +/- with the human player balls scored (by their HP) added to the "+" column.
Alliance +/- for an individual match would simply be the score (before penalties, possibly removing super cells if you wished). For an event you may replace this with an "OPR/DPR" style calculation if you prefer.
The ideal tracking solution would be 12 scouts per match. One scout is tracking each robot to see how many balls it scores. At the end of the match they note how many balls were scored in the trailer. That scout also notes some basic traits of how the match was played (robot was pinned, played defensively, etc.).
One scout is also assigned to each human player to track how many balls they score (accuracy can also be noted if you wish). General notes here also apply (throwing into field vs. "feeding" robots).
Both the average +/- and the individual match +/- are going to be very important data. The average should (theoretically) give a fairly imbiased estimation of which robots are best against the field of competitors. The individual match one will show outliers. What happened in those outliers is quite possibly the most important data of all.
Find what went right and what went wrong (based on the quick notes taken) and you ca figure out which teams are reliable, what strategies work for/against them, and what types of robots they match up well against.
That data is going to be a quicker and less subjective method of estimating the abilities of teams.