Guide

Game setplay and frame kill systems explained

Harbor Siege's sweep ender granted +32 advantage on knockdown — enough for real offense, on paper. In ranked data, 71% of those knockdowns ended in neutral resets: attackers guessed meaty timing, whiffed, and ate wake-up reversals. Oki damage sat at 14% of round damage. Players with strong frame-data literacy still could not convert advantage because the game offered no repeatable routes — only a timing minigame with no training wheels.

The refactor introduced setplay: predetermined sequences that spend knockdown advantage across filler moves so a follow-up attack's active frames align with the defender's wakeup (a frame kill). Sweep into dash-stop into low meaty, launcher into backdash into overhead, and throw knockdown into micro-walk strike-throw became documented routes with training-mode overlays. Oki damage share rose to 36% without raising single-hit damage. This guide covers setplay taxonomy, frame-kill math, route trees and branching, neutral setplay off plus frames, the Harbor Siege case, a technique decision table vs random oki, pitfalls, and a production checklist.

What setplay is (and why frame kills matter)

Setplay is offense built around situations where the attacker knows spacing, timing, and available options in advance — usually after a knockdown, wall splat, or guaranteed plus-on-block string ender. The attacker is not reacting frame-by-frame; they execute a rehearsed route whose branches cover defender habits (mash, delay tech, jump, block).

A frame kill is the mechanical core: using filler actions (dash, backdash, whiffed normal, jump landing, run stop) to consume exactly N frames so the next move becomes a meaty on wakeup frame 1, or a safe jump that lands before a reversal activates. Without frame kills, players must manually time every meaty — high skill floor, high failure rate, and poor spectator readability.

Setplay vs generic okizeme

Generic oki — attacker improvises meaties, throws, and jumps at knockdown; success depends on manual timing skill.
Setplay — designer or community documents routes with fixed filler; execution is mostly sequence memory and spacing discipline.
Frame-kill setplay — filler frame count is exact; the final move's active frames are guaranteed to align with a target wakeup frame absent external pushback changes.

Frame-kill math and filler taxonomy

Given knockdown advantage A and a target move with startup S and active frames starting at frame S, a frame kill finds filler totaling F frames such that:

F + S = A + wakeup_frame_offset

The wakeup_frame_offset accounts for delayed stand-up animations, tech rolls, and hard vs soft knockdown timing differences. Common filler types:

Filler	Typical frames	Design notes
Forward dash (first active)	12–18	Changes spacing; must stop before meaty whiffs
Backdash	14–22	Creates overhead or far throw range frame kills
Whiffed light normal	3–7	Fine-grained tuning; low commitment if defender early rises
Run stop / micro-walk	1–4 per step	Tekken-style precision; sensitive to input noise online
Jump empty landing	18–35	Safe-jump family; pairs with arc geometry
Special move with long recovery	Varies	Risky filler; telegraphs setplay unless canceled early

Meaty frame 1 vs frame trap frame 3

Frame kills can target different wakeup frames. A frame-1 meaty catches mash and many reversals. A frame-3 frame trap intentionally leaves a 2-frame gap so a delayed button press loses to a faster follow-up. Document which routes aim for which wakeup frame — players who only practice frame-1 meaties lose to delay tech unless the setplay tree includes a branch.

Knockdown route trees and branching

Strong setplay is a tree, not a single sequence. Each node is a situation; edges are attacker choices; leaves are expected payoffs. Example sweep knockdown (+32) tree:

Root: sweep KD, midscreen spacing band B2.
Branch A (frame kill dash-stop): low meaty frame 1 → if blocked, +2 into strike-throw shimmy; if hit, combo to corner.
Branch B (backdash frame kill): overhead meaty frame 1 → beats crouch block habit from Branch A conditioning.
Branch C (empty jump safe): lands before 3f reversal → baits mash into full punish; loses to delay tech into throw.
Branch D (delay meaty): frame-5 meaty after whiffed jab filler → catches delay tech from Branch C.

Trees should share conditioning logic with bait-and-punish systems: early branches establish respect; later branches exploit respect. If branches are unrelated random mixups, defenders learn one answer and collapse the tree.

Corner vs midscreen route divergence

Pushback on knockdown changes which frame kills remain valid. Harbor Siege tags each knockdown ender with spacing class (mid, corner, wall-splat) and ships separate route tables per class. A dash-stop frame kill that works midscreen whiffs in the corner because pushback removes two steps of travel.

Neutral setplay off plus-on-block strings

Setplay is not only knockdown-based. A plus-on-block string ender (+3) can frame-kill into the next string's active frames if the defender respects and blocks. Example: jab → stand medium (+2) → whiffed crouch light (4f filler) → stand heavy catches button press on frame 3. This is reset setplay: no knockdown, but the same timed alignment idea.

Reset setplay pairs with blockstring design: string endings must include both a high-damage route and a plus frame that enables a documented reset. Without published reset routes, plus frames become trivia only experts exploit.

Auto-timed setups and training tools

Modern fighters increasingly expose setplay to players who will not lab for hours:

Recording slots — save knockdown → route sequences in training mode; one-button playback against dummy.
Frame-kill assist — highlight when filler total matches a registered target advantage (green = meaty aligned).
Route browser — in-game list per character per knockdown class; reduces external spreadsheet dependency.
Replay tags — mark tournament moments where a frame kill succeeded or failed; feeds balance patches.

Assist tools should teach timing, not replace it entirely — ranked modes can disable auto-input while still showing overlay hints in practice. Harbor Siege ships assist in training only; ranked requires manual dash-stop execution but displays a post-round “oki efficiency” stat so players see whether their filler matched the documented route.

Harbor Siege oki refactor

The Siege team audited the top twelve knockdown enders per character and found only three had community-documented frame kills; the rest were manual-guess meaties. Changes:

Normalized advantage tiers — sweep +32, throw +28, launcher +38, wall-splat +42; each tier gets at least two frame-kill routes in the official route browser.
Filler standardization — dash stop and backdash lengths rounded to 2-frame increments so frame kills compose across characters without per-matchup bespoke math.
Branch labels — routes tagged MEATY-1, TRAP-3, SAFE-JUMP, SHIMMY so players know which defender habit each branch targets.
Pushback regression suite — automated tests verify frame kills after patch changes to knockback or wakeup animation length.

Outcomes after one ranked season: oki damage share 14% → 36%; reversal success on guess meaties 58% → 22% when attackers used documented routes; average match length +8 seconds (more decisions per knockdown, not longer combos). Characters without historically strong oki gained routes from shared filler standards instead of bespoke buffs.

Technique decision table

Approach	Best when	Weak when
Manual meaty only	Short advantage, simple knockdowns, low route count	Online input variance; players whiff and blame netcode
Documented frame-kill routes	Competitive scene, repeatable oki, spectator clarity	Routes stale if balance patches change advantage without retest
Random mixup oki	Casual modes, low execution barrier	Defenders memorize one answer; oki becomes coin flip
Auto-timed assist (training)	Onboarding, teaching frame-kill concept	Ranked if assist inputs routes for player
Reset setplay (no KD)	Characters with stable plus strings	Defender holds block and waits; low damage if no strike-throw layer
Damage-only tuning	Single-player power fantasy	Does not fix neutral resets after knockdown

Common pitfalls

Routes without branches. One frame kill per knockdown is solved after three repetitions; defenders delay tech once and win.
Ignoring pushback variance. Frame kills that work midscreen whiff cornered; players call it a bug.
Filler that is punishable on sight. Slow whiffed heavies as filler lose to early wakeup buttons before the meaty arrives.
Undocumented advantage values. Internal +32 becomes +30 after a patch; every published route breaks silently.
Character-exclusive math. Only one roster slot can lab frame kills; others feel oki-starved.
Safe jumps omitted from trees. Frame-1 meaties only; defenders mash reversal and succeed at high rates.
No training overlay. Players cannot tell whether a failed meaty was execution error or wrong route for spacing class.
Setplay without strike-throw. Blocked meaties end in respect with no shimmy or throw branch; oki damage plateaus.

Production checklist

Tag every knockdown ender with advantage tier and spacing class.
Publish at least two frame-kill routes per tier per archetype in route browser.
Verify frame-1, frame-3 trap, and safe-jump branches per major knockdown.
Regression-test routes after knockback, wakeup, or filler animation changes.
Document corner vs midscreen pushback deltas for each route.
Pair meaty routes with strike-throw and shimmy branches on block.
Expose frame-kill assist overlay in training mode (not ranked auto-input).
Log oki damage share and reversal success rate per knockdown class in telemetry.
Online QA frame-kill edges at +2 and +4 input delay.
Balance patch notes list which route IDs changed when advantage values shift.

Key takeaways

Setplay is rehearsed offense from known situations; frame kills spend filler frames so the next attack aligns with wakeup timing.
Route trees with branches (meaty, trap, safe jump, delay) beat single-sequence oki that defenders solve in one session.
Plus-on-block strings enable reset setplay without knockdowns — the same alignment math applies in neutral.
Harbor Siege raised oki damage share 14% → 36% by standardizing filler, publishing routes, and regression-testing pushback — not by buffing damage.
Training overlays and route browsers teach frame kills; ranked should still reward manual execution of documented sequences.