How Parse Percentiles Work
Every parse on Epog Logs has a colored percentage badge next to it — a fire mage's 87 in green-orange next to a warrior's 42 in blue. Those numbers are how the site compresses "how good was this performance" into a single comparable figure. They're not arbitrary: there's a specific formula, and once you know what it's measuring you can read a parse instantly. This guide explains the methodology and what each color tier represents.
The formula
For a given boss and class, every parse has a percentile equal to your value divided by the class top, times one hundred. So if the best fire mage on Onyxia in the entire dataset is doing 4,000 DPS, and you parsed 3,000 DPS as a fire mage, your percentile is 75. If you exactly tied the top, you'd be 100. If you doubled the top, you'd be 200 — and yes, that can happen when you're the new ceiling.
This is called a value-based percentile. The alternative — used by most legacy WoW analyzers — is a rank-position percentile, which orders all parses from worst to best and reports your position in that ordering. Rank-position percentiles tell you how many people you beat. Value-based percentiles tell you how close you got to the ceiling. The two answer different questions, and value-based is more useful for theorycrafting because it cleanly separates "how well-played" from "how many tries you had at it".
Why class-scoped
The pool you're being measured against is only other parses of the same class, deduplicated to one entry per distinct player (the player's best). A retribution paladin doesn't get a worse percentile because hunters out-DPS them on Onyxia — they're being compared to other ret paladins. This is the only fair way to do it, since class ceilings differ widely depending on raid composition and encounter mechanics.
Spec is intentionally not part of the comparison pool. A frost mage and a fire mage compete in the same pool. If frost outperforms fire on a particular fight, frost's percentiles will skew higher and fire's lower — that's information, not a bug. It tells you the spec ranking on that fight.
Minimum thresholds
Tiny pulls and joke parses are filtered out: a parse needs to be at least 50 DPS or 10 HPS to count for ranking. Without this, an early-fight death producing 8 DPS would inflate the percentile pool with nonsense.
The color tiers
Percentiles are color-coded the way WoW item rarities are colored, which makes them readable at a glance:
- 100 — artifact gold (●): the literal class ceiling on this fight. The benchmark every other parse is compared against.
- 95+ — legendary orange (●): elite play. Within 5% of the ceiling. Anyone holding multiple legendary parses across a tier is among the very best players in the dataset.
- 75+ — epic purple (●): very strong. Comfortably above average for the class — typically only achievable with a focused build, good consumables, and a clean execution.
- 50+ — rare blue (●): solid. The middle of the ceiling — a clean run by an attentive player.
- 25+ — uncommon green (●): fine. Possibly under-geared, possibly a learning night, or a fight where the player ate a mechanic mid-pull.
- Below 25 — grey (●): well below the ceiling. Often means a death, a fight spent on adds instead of the boss, or genuine room to improve.
Caveats worth knowing
- The ceiling moves. When a new top parse lands on a fight, every other parse's percentile drops slightly (because the denominator just got bigger). This is why your historical 95s can quietly become 89s as the meta improves — the absolute number didn't change, the pool did.
- Class composition matters. A mage in a raid with no curse-of-elements warlock will parse lower than one in a raid that has every debuff up. The site's percentiles don't normalize for buffs — they measure raw output. Two parses with the same percentile aren't necessarily the same skill level if one had perfect buff coverage and one didn't.
- Fight length affects HPS less than DPS. Healers smooth their output across a fight; DPS is more bursty. On very short fights, the top parses come from cooldown-stacking — be wary of reading a 10-second fight as representative.
- Per-boss vs. raid-overall. The site computes percentiles separately for each boss and for each raid as a whole. A player who consistently parses 80+ on every boss in a raid will have an overall raid percentile of around 80; the two are independent calculations of the same data.
Where percentiles show up
You'll see them on the per-boss damage tables in the log viewer, on the player parse modal (click any name), on the rankings page, on guild pages, and on individual player profiles. Wherever the badge appears, it's the same calculation — the same color means the same thing across the site.
Next step
Want to put this into practice on a specific fight? Onyxia's Lair — Epog strategy walks through the only end-game raid currently available on Epog and what parsing well there looks like.