Deadlock Meta · Replay Lab
Understand the meta.
Understand your mistakes.
Stats show what works across matches. Replay Lab shows what happened in yours: the fight, the objective window, and the economy behind the decision.
Open private Replay Lab ↗Public stats stay free. Private coaching is ad-free, with one subscription and explicit usage limits.
Fight review
Before, during and after: who was involved, recorded damage and casts, health changes and the actual outcome.
Macro decisions
Observed objective spawns, readiness and movement, plus verified structure pressure—not assumptions about intent.
Economy & purchases
Compare net worth with both teams, inspect banked souls and purchase timing. Unknown counter-buy mechanics stay ungraded.
Team review
Review both rosters together: observed combat losses, multi-hero participation, objective availability and same-time resource gaps. Communication and intent remain unknown.
Real replay example · match 108776476 · Pocket perspective
A death before an observed Mid Boss respawn
- 28:48 — Pocket purchases Refresher. This is a recorded transaction, not a verdict that the item was correct or incorrect.
- 29:07.8 — Pocket dies to Dynamo, about 19.7 seconds after that purchase.
- 29:30.5 — The replay records a Mid Boss respawn, about 22.7 seconds after the death. This is a concrete preparation window to review.
Review question: how did the lead-up to this fight affect readiness for the upcoming objective? The timestamps establish the sequence; they do not prove Refresher caused the death or that the team should have contested.
This excerpt is deterministic replay evidence, not generated AI coaching. Player display names are omitted from this public example.
One replay. Every player and both teams.
Upload a supported .dem, wait for extraction, then choose a player and a coaching goal. The match is parsed once and cached reports are prepared for all participants and both teams. Switching to a team review uses the same match ledger, without another parse or model call. Optional AI critique runs only when you request it; unsupported output falls back to clearly labeled evidence.
What we won’t pretend to know
Recorded movement is not vision or intent. Current item stats are not historical item mechanics. We won’t invent cooldown readiness, aim quality or counter-buy grades that the available replay data cannot establish.