Guide

Game guard crush systems explained

Harbor Siege ranked duels averaged 94 seconds in Platinum tier before patch 2.1.0. Telemetry showed block rate on mids at 78% with zero posture drain on standard guard — defenders could hold block indefinitely while chip damage ticked at 2% per blocked special, never reaching chip lethal thresholds. Rushdown characters abandoned blockstrings after the second blocked normal because no move forced a guard reaction beyond “keep holding back.”

Patch 2.1.0 added explicit guard crush properties to two specials per archetype: startup guard crush on a slow overhead special and active guard crush on a plus-on-block low follow-up. Blocked hits now apply stance damage that collapses guard on the third consecutive crush without invoking the full posture meter from guard break systems. Average duel time dropped to 79 seconds; block rate on mids fell from 78% to 61%; aggressive opener rate rose 19%. This guide covers guard crush as a per-move property, block bypass tiers, stance-damage stacking, interaction with perfect guard and frame traps, rollback determinism, the Harbor Siege refactor, a technique decision table versus posture-only and chip-only anti-turtle design, pitfalls, and a production checklist.

What guard crush is (and is not)

Guard crush is a move-level flag that changes what happens when an attack connects while the defender is blocking. It is not the same as a global guard-break meter, though the two systems often interact.

Three distinct anti-turtle layers appear in modern fighters:

Chip damage — life loss on block; threatens lethal at low health but does not force action at high health.
Posture or guard-break meter — cumulative resource depleted by blocked hits; collapse enters a punishable break state.
Guard crush property — this specific move applies crush rules on block: stance damage, guard stagger, or full guard break regardless of remaining posture.

Guard crush answers a design gap: “I need this move to threaten block without giving every normal the same power.” Designers tag slow overheads, charge releases, and blockstring enders rather than flattening all block interaction through one meter.

Guard crush property taxonomy

Document every crush move in a data table keyed by move ID. Typical columns:

Property	On block effect	Example use
Startup guard crush	Crush active during startup frames only; blockable after active begins	Slow overhead that beats mashing jab during wind-up
Active guard crush	All active frames crush; whiff punishable if baited	Plus-on-block low ender that still threatens guard collapse
Stance damage crush	Adds crush meter per blocked hit; guard breaks at threshold	Blockstring conditioning without full posture system
Instant guard break crush	Single blocked hit enters break state immediately	Rare super or EX-only anti-turtle tool
Chip-through crush	Ignores block damage reduction; applies full or elevated chip	Special that makes block still costly at high health

Frame window placement

Startup crush windows must be visible in frame data overlays. If crush frames 1–12 of a 28-frame startup, defenders learn to interrupt after frame 13 or jump before crush ends. Active-only crush on a 4-frame low is nearly unreadable — pair with distinct animation telegraph (glow, sound cue) or accept that crush becomes a conditioning tool rather than a reaction test.

Stance damage and crush stacking

Many games implement guard crush as a secondary meter independent of life and posture:

Defender starts at 100 stance (hidden or shown as a thin bar under health).
Each guard-crush blocked hit deals 25–40 stance damage.
At 0 stance, next blocked hit or a dedicated crush ender triggers guard stagger (8–20 frames of blockstun extension with guard disabled).
Stance regenerates at 5–10 per second when not blocking crush-tagged hits.

Stacking rules matter for blockstring design:

Same-move stack cap — repeated crush low only counts once per string; prevents infinite stance drain from one button.
Crush decay on reset — stance refills partially when pressure ends; defender gets a breath after escaping corner.
Crush carry between hits — stance damage persists through entire blockstring; third crush ender in a sequence breaks guard reliably.

Publish stance damage values per move. Hidden crush math feels arbitrary when guard suddenly collapses mid-string.

Interaction with perfect guard and frame traps

Guard crush must coexist with defensive skill tools:

Perfect guard immunity

Most games grant perfect guard immunity to guard crush — a timed block within 2–3 frames of impact negates stance damage and may reflect crush property. Without this escape, crush becomes unavoidable damage and defenders quit. With overly generous immunity, crush moves never threaten high-level play.

Frame traps after crush threat

Crush conditioning creates new frame-trap branches:

Opponent respects crush overhead → delay mid catches button after crush startup ends.
Opponent mashes after two crush lows → third hit is plus frame trap, not crush, catching jab.
Opponent perfect-guards crush → attacker is minus; defender gets counter-punish window.

Telemetry pairs: crush attempt rate, perfect-guard rate against crush, guard stagger rate, and average blockstring length before reset.

Block bypass tiers and invulnerability

Some engines treat guard crush as partial block bypass rather than stance damage:

Tier 1: Guard stagger — defender stays blocking but enters extended blockstun; no damage; enables attacker re-approach.
Tier 2: Guard break on block — single crush hit opens guard for one hit punish window (distinct from posture collapse).
Tier 3: Unblockable during crush frames — block input ignored; functionally identical to unblockable but only during tagged frames.

Crush moves do not bypass invincibility or armor unless explicitly flagged. A common bug tags crush as hitting through backdash invuln — test throw-invuln and upper-body invuln frames separately.

Harbor Siege refactor

Patch 2.1.0 guard crush rollout per archetype:

Rushdown: standing heavy overhead — startup crush frames 1–14 of 22-frame startup; 35 stance damage on block.
Rushdown: crouching medium special follow-up — active crush on all 6 active frames; +2 on block; 30 stance damage.
Zoner: charged projectile release — crush only if held 30+ frames; 40 stance damage; minus-8 on block if not crushed.
Grappler: armored approach special — instant guard break on block if stance below 50%; otherwise 45 stance damage.
Stance threshold for guard stagger: 0 (third crush hit in 4 seconds triggers 18-frame guard stagger).
Perfect guard within 2 frames: nullifies stance damage; attacker minus-4.
Training mode: orange outline on crush-active frames; stance bar visible.

Over 28,000 ranked matches post-patch: average duel time 94s → 79s; mid block rate 78% → 61%; perfect-guard rate against crush 12% → 19% (players learned the counter); rushdown win rate vs zoners 44–56 → 51–49. Survey item “Blocking feels interactive” rose from 38% to 71% agree.

Technique decision table

Approach	Best when	Watch for
Stance-damage crush (3-hit threshold)	Blockstrings need escalation without full posture meter	Hidden math feels random; show stance bar or telegraph third hit
Startup crush on slow overhead	Defenders mash during obvious wind-up	Unreadable crush on fast overheads deletes neutral
Active crush on plus-on-block low	Low blockstrings stall without guard threat	Infinite stance drain if same move stacks uncapped
Instant break crush (super/EX only)	Rare comeback or meter spend anti-turtle	Every round opens with crush super; defense obsolete
Posture meter only (no crush flags)	Unified resource; Soulslike or SF6-style	All blocked hits feel identical; no move-specific threat
Chip damage only (no crush)	Simple fighters; lethal threat at low HP	Full-health turtle meta like pre-patch Harbor Siege
Perfect guard nullifies crush	Skill-based counter for high-level defenders	2-frame window too tight online; crush feels unavoidable

Common pitfalls

Every normal has crush — blocking becomes useless; defenders always jump or backdash; neutral slows to a crawl.
Crush with no whiff recovery cost — safe on block crush lows dominate; offense requires no commitment.
Uncapped same-move stance stack — one button drains guard in a loop; degenerate blockstring meta.
Crush ignores perfect guard — removes skill counterplay; high-level players abandon defensive lab work.
Hidden crush frames — players cannot learn when crush starts; frame data distrust spreads.
Crush hits through throw invuln unintentionally — backdash escape stops working; grappler matchups break.
Rollback desync on stance meter — one client sees guard stagger, other does not; serialize stance in every snapshot.
Crush plus chip lethal on same move — double punishment on block; defenders die at 15% from one blocked special.

Production checklist

Tag crush type per move: startup, active, stance damage, instant break, chip-through.
Publish crush-active frame ranges in frame data and training overlays.
Define stance meter max, per-move damage, regen rate, and stack caps.
Specify guard stagger duration and whether counter-hit rules apply.
Implement perfect-guard nullification with explicit attacker frame disadvantage.
Test crush against invuln, armor, backdash throw invuln, and cross-up block direction.
Balance blockstring length: third crush hit should break guard, not first.
Serialize stance and crush flags in rollback netcode snapshots.
Telemetry: crush rate, stagger rate, perfect-guard vs crush, duel duration.
Tutorial: recognize startup crush telegraph and perfect-guard timing window.
Compare rushdown vs zoner win rates before and after crush patch.
Ensure crush moves have whiff punish windows longer than crush reward on block.

Key takeaways

Guard crush is a per-move property, not a substitute for chip damage or global posture meters.
Stance-damage stacking lets blockstrings escalate toward guard stagger without every normal breaking guard.
Startup crush rewards reading mashing; active crush on plus lows keeps pressure honest.
Perfect guard nullification preserves skill counterplay against crush-heavy offense.
Harbor Siege duel time dropped 16% and mid block rate fell 17 points after targeted crush properties on two specials per archetype.