How Scores Work

The current automatic Raid Night Role Performance model is validated for DPS and healers. Role Parse Blend V2 uses the Warcraft Logs overall role percentile at 80% relative weight and the item-level bracket percentile at 20%. If one ranking component is unavailable, the available component is proportionally reweighted rather than replaced with a neutral value.

Raid-night damage or healing activity remains visible as diagnostic context, but it does not add or subtract Role Performance points. Active time already affects the output that produces the Warcraft Logs percentile, so scoring it again as a separate 25% component could count overlapping performance evidence twice and affect roles differently.

Tank and Support evidence is still collected, but NRO does not turn a tank's damage parse or a support player's damage ranking into a Role Performance grade. Those roles remain N/E until their own mitigation, defensive, control, and support-performance models are validated.

A Warcraft Logs percentile is accepted only when it belongs to a ranked kill included in the current Calculation Run and matches the specialization recorded for that pull. If the ranking or spec identity cannot be established, NRO keeps the local output evidence as context instead of manufacturing a score.

Boss-level reports may also show local output percentile, active time, and boss-target contribution for progression context. That encounter analysis is separate from the Raid Night Role Performance headline and does not substitute for a missing Warcraft Logs ranking.

Today's logs reliably prove some useful actions, but they do not yet enumerate every player-owned chance that was missed. After the first score-eligible contribution, the provisional index uses a 20-point evidence floor. Normalized contribution frequency can add up to 75 more points with diminishing returns. It never invents a failure.

NRO does not yet have a complete class, specialization, and talent capability denominator. Because of that, the provisional score does not award points simply for using more distinct kinds of Utility. Breadth remains visible evidence until NRO can prove what tools the player's actual build had available. This prevents naturally broader class kits from receiving an automatic scoring advantage.

Pulls normalize the amount of positive evidence, but attendance never becomes a failure. Assignment-owned work remains in Assignment Execution. A raw cast that depends on its target, strategy, timing, or effectiveness remains visible evidence but does not automatically add provisional Utility credit until that context is validated.

Once an evidence source can certify the complete opportunity denominator, NRO will use the formal model: successful opportunities earn full credit, partial successes earn half, and confirmed player-owned misses earn zero. The four lanes below are reweighted when a lane truly had no opportunity.

Two comparable pulls are enough for a provisional Consistency score. Evidence strength remains Limited at first, becomes Moderate with five comparable pulls, and becomes High at eight or more.

Execution Stability uses the actual earned credit from Mechanical and Assignment opportunities. A 70% Assignment timing result therefore contributes 70% to that pull's Assignment execution rather than disappearing or being converted back into a binary pass.

Performance Stability measures how much the player's DPS or HPS moves around their own progression-matched local baseline. This is repeatability, not an absolute quality ranking. Role Performance remains responsible for answering how strong the player's actual output was.

Dependable Floor looks for unusually weak output relative to that same local baseline and begins contributing after five comparable performance pulls. Before then, the supported components are proportionally reweighted. Missing evidence is never replaced with a neutral score.

Tank and Support do not receive an invented DPS or HPS stability model. When no validated role-output signal exists, their Consistency score can come from Execution Stability alone, with reduced component coverage recorded in the evidence.

Cross-Night Repeatability belongs to the Team Season view. The Raid Night processor never pretends that one night contains multi-night evidence.

Assignments are weighted by importance:

We also account for timing, difficulty, whether the player was primary or backup, whether the assignment was actually possible, and whether another failure prevented them from completing it.

A documented or confirmed assignment always takes priority. NRO should not invent a missed cooldown, interrupt, or soak that the player was never clearly given.

For strict timed assignments, use within ±5 seconds earns full credit. Credit then decreases continuously: 10 seconds off earns 80%, 15 seconds earns 60%, 20 seconds earns 40%, 25 seconds earns 20%, and 30 seconds earns zero. Missed assignments and cooldowns consumed outside the valid assignment window earn zero.

AMZ, Darkness, Rallying Cry, and Zephyr use a separate effectiveness model. NRO measures how much of the intended incoming-damage window the actual defensive covered, so a useful defensive is graded by protection effectiveness instead of timing alone. If the planned window contained no meaningful damage, the opportunity is N/E rather than an artificial success or failure.

NRO first has to prove that a mechanic created a fair, player-owned opportunity. A formal Mechanic Parse requires at least three scored opportunities. One or two observations remain coaching evidence and do not become a formal 0 or 100.

Every validated scoreable opportunity currently counts equally in the formal success-rate score. The mechanic-level breakdown still shows the exact mechanic names, successful outcomes, failures, exclusions, and evidence used to defend the result.