Understanding OPR and alliance statistics

Offensive Power Rating (OPR) estimates an additive contribution to alliance scores. Scoutless makes one row for each scored alliance in a finalized match. Teams listed in that alliance have a value of one in the row; other teams have zero. The current dataset service includes all listed alliance members, so inspect sitting-team flags when reviewing a result.

The calculation chooses team estimates that minimize the sum of squared differences between the observed alliance scores and the sums of team estimates. This is least squares. It is a model of alliance outcomes, not an observation of each robot’s scoring actions.

Fictional simplified alliance equations: A + B = 30, A + C = 24, and B + C = 18. These three independent equations give A = 18, B = 12, and C = 6. Adding A’s and B’s estimates reproduces their alliance total of 30.

Real datasets usually contain many rows that cannot all be matched exactly. Least squares finds a compromise. The toy example explains the arithmetic only; it is not a Scoutless team rating or a complete event. If A and B always appear together, the data may identify their combined contribution without identifying either one separately.

Defensive Power Rating (DPR) uses the same team participation rows but fits the opposing alliance’s score. Calculated Contribution to Winning Margin (CCWM) is OPR minus DPR. For an illustrative OPR of 18 and DPR of 11, CCWM is 7.

DPR does not isolate defensive skill: opponents, partners, match strategy, and the schedule also affect the score conceded. Likewise, CCWM is an estimate tied to its input matches, not a guaranteed margin in the next match. Negative estimates are possible in an unconstrained least-squares model.

Scoutless defines win rate as wins divided by wins + losses + ties. A fictional record of 6 wins, 3 losses, and 1 tie gives 6 / 10 = 60%. A tie stays in the denominator and is not counted as half a win.

Scoutless derives the team record from stored final matches with both alliance scores available. Official ranking records are separate source data and do not replace this calculation. Different match coverage can produce differences from a published ranking record. With no usable matches, win rate is unavailable.

OPR, DPR, and CCWM are unavailable when the requested team is absent from usable matches or the solver cannot separate team contributions. A singular or insufficiently informative system returns null, not zero. Scoutless does not currently calculate TrueSkill or a public Scout Score.

Compare the season, event scope, and freshness before ranking teams by a metric. A score scale can change between games, and a partial dataset can change an estimate substantially. Scoutless does not attach a confidence interval to these ratings; extra decimal places should not be read as extra certainty.

Review actual matches to investigate an interesting rating. Check whether the team’s observed role fits your alliance and whether the behavior repeats with different partners. The methodology below is the source for these Scoutless-specific definitions; other services may calculate similarly named statistics differently.