Paste your combat log. See where the gap comes from.

Guests can try 5 free analyses, tracked by a cookie in this browser. The sixth analysis requires sign-in. Signed-in users get 5 free analyses daily, resetting at 00:00 Beijing time (UTC+8) without rollover. Each newly registered user who verifies their email through your invitation earns you 10 permanent credits, used only after the daily allowance. Analyses created while signed in appear in your history; earlier guest analyses are not imported automatically. Guest reports remain protected by the original browser credentials. Verification and necessary security limits still apply.

The site provides four isolated analysis entries: Retail Midnight raids and Mythic+, Titan Reforged raids, MoP Classic raids and Challenge Modes, and BCC Anniversary raids. It identifies the character, specialization, and role from the log before calling the matching edition-specific engine. Beta specializations still show validated differences and timestamps, but coaching must pass a stricter evidence gate.

This page accepts report links from classic.warcraftlogs.com and cn.classic.warcraftlogs.com, preferably with fight and source parameters. Public reports and unlisted reports available by link are supported. Private reports that require a signed-in user's OAuth access are not currently supported.

With one log, the backend searches for a comparable benchmark automatically. You can also choose an explicit benchmark in the advanced settings and paste a second log. An explicit benchmark skips search only; it does not bypass edition, specialization, boss or dungeon, difficulty, team size, partition, or route checks.

The backend scans leading public rankings for the same edition, region, specialization, and content, but it does not simply select rank one. It applies non-negotiable filters first, then scores item level, duration, trinkets, talents, set bonuses, build, and target or route opportunity. Other logs from the same character are excluded.

If the link has no source parameter and the fight contains several supported players, you need to confirm the subject. If no candidate meets the hard conditions after every allowed relaxation, the task completes with no_benchmark. This is not a service failure. You may provide a genuinely comparable explicit benchmark instead.

No. This page analyzes MoP Classic raids and Challenge Modes. Raids, Mythic+, and Challenge Modes use separate comparability rules. Automatic and explicit benchmarks must match edition, content type, season or partition, region, and specialization. Raids also match boss, difficulty, and team size; dungeon content also matches level or medal, affixes, route, and pull opportunity.

A damage difference is an observed fact, but its cause may not be player execution. The number remains visible. Coaching is generated only from controllable evidence with medium or high confidence and a concrete timestamp. Gear, set bonuses, talents, deaths, target opportunity, positioning, assignments, unresolved remainder, or low confidence can close the action gate.

The backend uses a versioned boss mechanics library to reconstruct phases, target lifespan, burst windows, and forced downtime. Evidence is labeled observed, inference, requires_external_context, or not_supported. Logs rarely prove voice calls, exact positioning, or that night's assignments by themselves, so the difference stays visible while confidence in its cause is reduced.

The current unified analysis extracts event evidence and generates reviews in code, without AI calls. It presents major differences, key moments and evidence-approved actions. Observed differences are not proven mistakes or recoverable gains. Mechanic coverage varies by specialization, and missing data is not invented. The interface and FAQ support Chinese and English; source names and some historical diagnostic text may retain their original language.

The result page connects each major cause to backend finding IDs, spell and target statistics, the phase timeline, and approved actions. Expand the related evidence to verify buffs, resources, casts, target lifespan, and upstream events at exact timestamps instead of relying on the summary alone.

Guides aggregate leading public logs from the current edition and partition for kill pace, deaths, damage taken, targets, roster, and routes. Each page shows its collection date, sample size, and source links. Speed rankings are an auditable sample, not a universal progression strategy. Positioning, assignments, roster, and affixes still need team context.

The backend does not store complete raw events, OAuth tokens, AI API keys, or the Turnstile secret. Tasks, derived evidence, public character and server metadata, and final interpretations are retained for 7 days. The browser remembers up to 10 recent task IDs. Administrative analytics may retain submitted public log URLs and explicit benchmark URLs beyond 7 days, but no additional character-name field is recorded.

Turnstile protects guest analysis submissions and sign-in code requests. Signed-in users do not need another challenge to submit an analysis. For CAPTCHA_FAILED, complete a fresh check in the current form before submitting. If the widget cannot load, check whether your network or content blocker blocks challenges.cloudflare.com. Reading reports and history does not require another challenge.