Guide

Game red parry systems explained

Harbor Siege’s pressure meta leaned on long blockstrings: safe lights chained into plus frames, chip damage stacked, and defenders who mashed reversal got frame-trapped every time. Hold-block was the only honest option — but it never flipped advantage. After balance review, engineers shipped a red parry layer: while still in blockstun from a blocked hit, the defender could press forward (or a dedicated parry input) within a tight window to parry the next strike in the string. Success cancelled remaining blockstun, granted +6 frame advantage, and opened a jab or throw check. Blockstring respect rate climbed from 22% to 51%; chip-damage support tickets fell 38%. Defense gained a readable skill hook without turning every matchup into full parry footsies.

A red parry (blockstun parry, guard parry, or SF3-style red parry) is distinct from neutral parries and just guard. You are already blocking; the input fires during blockstun, not at raw impact on the first frame. That constraint makes red parry a blockstring counter rather than a neutral anti-poke tool. Done well, it gives defenders a high-skill escape from infinite pressure without deleting offense. Done poorly, it is either unusable (window too tight) or degenerate (every string becomes unsafe). This guide explains what red parry adds to guard systems, timing-window models, frame advantage and cancel routes, economy coupling, blockstring interaction, the Harbor Siege refactor, a technique decision table versus just guard and full parry, pitfalls, and a production checklist.

What red parry adds beyond hold-block and just guard

Hold-block resolves each hit with defined frame disadvantage. Just guard rewards impact-timing on individual hits but still leaves the defender minus unless they perfectly time every strike in a five-hit string. Red parry targets the gap between blockstun end and the next active frame — or the late blockstun frames of the current hit — letting a defender interrupt a continuing string.

Blockstring break point — offense must respect a defensive option mid-string, not only on wake-up or after knockdown.
Skill expression without neutral parry risk — you already blocked the first hit; failed red parry usually means you keep blocking instead of eating a full counterhit.
Readable telegraph for spectators — red parry success often tints the guard flash or plays a distinct SFX; the string visibly stops.
Counterweight to chip and guard-break meters — defenders who read strings can escape before posture breaks.
Roster-neutral defense layer — every character gets the same red parry rules; grapplers cannot rely on infinite safe chains.

Red parry does not replace fuzzy guard or throw tech — it answers strike blockstrings. Document which move classes are red-parryable (strikes yes, command grabs no, projectiles sometimes) in one table visible to players.

Red parry timing window models

Engines implement red parry through a few recurring FSM hooks. Pick one primary model and document it in your combat bible — mixed per-character rules without UI explanation feel arbitrary.

Late blockstun window (classic red parry)

During the last N frames of blockstun on hit K, accept a forward+block or parry input. If hit K+1 would connect while still blocking, the parry succeeds: blockstun clears, attacker enters parried recovery. Typical window: 2–4 frames at 60 FPS. SF3-style games tie the window to the final third of blockstun so defenders react to rhythm, not raw reaction to startup.

Between-hit gap window

After blockstun ends but before the next hit’s active frames, open a 1–3 frame parry state if block is held. Works when strings have visible gaps; fails on true blockstrings with frame advantage > 0 unless you also grant parry during the gap. Pair gap data with frame data exports so designers see which strings are theoretically unred-parryable.

Per-string once (budget parry)

Defender gets one red parry attempt per blockstring sequence; input consumes the budget whether it succeeds or whiffs. Prevents red-parry fishing every frame at the cost of higher cognitive load. Show a small UI pip that refills when the string ends or after N seconds of neutral.

Meter-gated red parry

Spend 0.5 super meter or one defensive stock for a wider window (5–6 frames) or automatic red parry on the next blocked hit. Use when your roster has extreme plus-on-block tools; gate prevents every-round usage.

Frame advantage, cancels and counter routes

Red parry success is worthless if the defender cannot punish. Define these constants in data:

Advantage on success — typical range +4 to +8; must exceed the attacker’s fastest abare option after parried recovery. If jab is 4f startup, +5 is minimum for strike checks.
Allowed cancels — jab, throw, special, or only light normals. Full special cancel from red parry can delete neutral if damage is high; cap to lights or scaled damage.
Parried attacker state — fixed recovery animation (like a whiffed heavy) vs generic stagger. Fixed recovery makes matchup knowledge matter.
Pushback reset — snap to mid-range or preserve corner spacing; corner red parry into throw is oppressive if pushback is zero.
Chip and guard meter — on success, refund chip taken during the string segment and pause guard-break drain for 30 frames.

Regression test: for each character’s benchmark blockstring (e.g. light ×3 into special), compute whether a frame-perfect red parry on hit 2 yields a confirmed punish at midscreen and in corner. If only one archetype can punish, widen advantage or shorten attacker recovery uniformly.

Blockstring and offense interaction

Red parry reshapes how designers build pressure, not only defense:

String termination points — end blockstrings on minus frames or jump-cancel before red-parry windows on hit 3+; offense chooses risk.
Frame trap baiting — delay mids in strings to catch mashing red parry into counterhit; traps must be minus on block if red parry fails so defense is not strictly dominant.
Strike-throw after respect — once red parry is common, shimmy and tick throw layers rise; ensure throw range matches post-parry spacing.
Multi-hit specials — decide per hit: first hit blockable, later hits red-parryable individually or only once per move.
Projectile blockstrings — usually exclude red parry or require parry input against physical shield (different skill layer).

Publish a red-parry vulnerability tier for benchmark strings in patch notes: Tier A (gap exists, red parry easy), Tier B (tight window), Tier C (true blockstring, must jump or reversal). Players stop guessing.

Harbor Siege pressure refactor

Pre-patch ranked duels showed:

Top-tier rushdown blockstring respect: 22% (defenders rarely countered)
Chip damage support tickets: 2.4× baseline per week
Guard-break wins: 31% of rounds (posture meter overtuned vs passive block)
Reversal mash rate on block: 44% (frame traps everywhere)

Shipped red parry rules:

3-frame late-blockstun window on hits 2+ of any blocked string (hit 1 excluded to preserve neutral footsies)
Input: forward + block within window; one attempt per string
Success: +6 advantage, chip refund on string segment, attacker 18f parried stagger
Command grabs and multi-hit supers immune; each super hit red-parryable once only
Training mode overlay: red zone on blockstun timeline + success rate stat

After 9,800 post-patch ranked duels:

Blockstring respect rate: 51% (was 22%)
Chip damage tickets: −38%
Guard-break win rate: 19% (was 31%)
Red parry success rate among attempts: 34% (skill ceiling exists)
Average round length: +4% (more reset neutral, fewer infinite strings)
Rushdown vs zoner win-rate delta: −6% (was −11%)

Technique decision table

Your design goal	Prefer	Over
Break oppressive blockstrings without neutral parry complexity	Late-blockstun red parry on hit 2+ with +6 advantage	Full SF3-style parry on every hit in neutral
Reward individual hit timing on block	Just guard per impact	Red parry only (no per-hit bonus)
Casual-friendly defense	Wider 5f window in casual queue or tutorial auto-red-parry once	Removing plus-on-block strings entirely
High-level footsies focus	Neutral parry + no red parry	Red parry on hit 1 (erases poke meta)
Prevent red-parry fishing	One budget per string + whiff does not consume on miss	Unlimited attempts every blockstun frame
Corner pressure balance	Red parry success adds pushback toward midscreen	Zero pushback into instant throw loop

Common pitfalls

Window on hit 1 — neutral pokes become unusable; rushdown dominates anyway from different angle.
Advantage too low — red parry succeeds but jab loses to 3f abare; players stop trying.
Silent rules — grabs red-parryable by bug; players feel cheated when experiment fails.
No attacker recovery — parry stops string but attacker is still plus; defense gained nothing.
Online input lag ignored — 3f window at 120 ms rollback feels impossible; add 1–2f leniency online only.
Conflicts with just guard — same input, different outcomes; separate inputs or priority table in FSM.
Red parry into infinite — special cancel damage unchecked; apply combo scaling from parry confirm.

Production checklist

Document red parry window type (late blockstun vs gap) in combat design doc.
Publish which move classes are immune (throws, grabs, armor, supers).
Frame-advantage table on success vs each roster jab and throw range.
Benchmark string tier list (A/B/C) for red-parry vulnerability in patch notes.
Training mode: blockstun timeline overlay with red window highlighted.
One budget per string or explicit unlimited policy in UI tutorial.
Online vs offline window leniency documented for netcode team.
FSM priority: red parry vs just guard vs fuzzy guard input resolution.
Replay tag “red-parry punish confirm” for balance review.
Chip refund and guard-meter pause rules on success.
Corner pushback on success to prevent throw loop degeneracy.
First-match tooltip when defender eats a 5+ hit blockstring without counter.

Key takeaways

Red parry is a blockstun-time defensive option that breaks continuing blockstrings — distinct from neutral parry and just guard.
Late-blockstun windows, gap windows, and per-string budgets are the three dominant implementation models.
Success must grant enough frame advantage and attacker recovery for real punishes — measure against benchmark strings.
Harbor Siege raised blockstring respect from 22% to 51% with a 3f hit-2+ window and +6 advantage.
Pair red parry with published string tiers, training overlays, and throw immunity rules — not hidden per-move exceptions.