design: multi-hit as a first-class dial — hit count IS break rate

[mtklein]

Owner direction (2026-07-29): "Pummel seems like a good candidate to do 2x bludgeon, where Suplex is one big hit. Bum Rush could go either way. And not just Sabin, but also Tools, Quadra Slam should hit 4x, etc. This one's a big one, not necessarily v0.8 but somewhere soon."

Why this is a balance pass and not a flavour pass. In OT6 a landed hit that matches a weakness chips a shield. So hit count is the break-rate dial — a 2× Pummel against a bludgeon-weak enemy chips twice per use, a 4× Quadra Slam chips four times. It multiplies straight against the MP costs just set in #45, and it interacts with the boost economy, which already grants Fight +1 hit per BP.

That makes it the strongest single lever we have over the break loop, which is the whole game. Hence: design first, with measurement, before any data moves.

Phase 1 — audit what is actually true today

• Which abilities multi-hit in the shipped ROM, and how many times? Read it, don't trust the docs: kits.md already annotates Flurry ×4, Tempest ×4, Bum Rush ×8, and those claims have never been verified against attack_prop/the damage path.
• How does a multi-hit ability's chipping actually resolve — per hit, or once per action? Ot6HitJoin (ot6_break.asm) is the join point; establish the real rule with a live measurement, because the entire design rests on it.
• What does vanilla already do for each candidate (Quadra Slam/Slice are 4-hit in vanilla; several Tools are single).

Phase 2 — the design

Per ability, a hit count with a stated reason, in the discipline magicite-tube-six.md used. The owner's seeds: Pummel 2× bludgeon (a fast prober), Suplex one big hit (a committer), Bum Rush open, Quadra Slam 4×. Extend to Tools and the rest of SwdTech.

The shape worth aiming at, stated so it can be argued with: hit count is a character's break-probing identity. Sabin opens gauges fast with cheap multi-hits; Cyan's ladder buys single big commitments that pay off once a gauge is already open; Edgar's Tools sit between. That gives the three ability characters genuinely different jobs against the same enemy, which is the Octopath structure the whole project is chasing.

Phase 3 — measure before shipping

• Cost-per-chip per ability, the same way balance: ability MP costs are noise against mid-game pools — the economy doesn't bite where it's learned #45 measured cost-per-pool. That is the number that says whether a 2× Pummel at 4 MP is broken.
• Re-run check_break_reach.py — hit counts change how fast a party opens a gauge, not which classes it can field, but the bands' feel moves.
• A knife-edge sanity check on a band the owner has played (the Serpent Trench precedent).

Not blocking

Unmilestoned deliberately — the owner said soon, not next. It should land as one coherent pass rather than dribbling out per character, because half a hit-count table is worse than none: it would make two characters' probing rates incomparable.

[mtklein]

The design target, from the owner (2026-07-29):

"In Octopath most characters have at least one multi-hit ability they can lean on for breaking, and a couple have more. That's a nice balance... even a weak attacker can be useful in a pinch if they can multihit with a staff or something."

That is the rule the hit-count table should satisfy, and it is stronger than 'assign counts sensibly' because it decouples damage from break contribution. A character who hits weakly is still pulling their weight if they can open a gauge, which is exactly why Octopath's benched characters stay interesting.

So the audit gains a per-character axis: does every playable character have at least one multi-hit option they can lean on to break, and do one or two have more? Enumerate it, and name the characters who currently have none — those are the gaps the pass exists to fill.

An observation that shapes it: OT6 already grants Fight +1 hit per BP (v0.1, the alternating-hands machinery; Genji Glove doubles the bonus). So every character can already multi-hit by spending boost — the universal fallback exists. What a dedicated multi-hit ability adds is a cheap, repeatable probe that does not spend the bank, which is precisely the Octopath texture: bank on the weak turns, break with the ability, nuke with the bank.

That reframes the per-character question usefully — not 'can they multi-hit at all' (they can, for BP) but 'do they have a way to probe without draining the resource they are supposed to be saving.' A character with no cheap multi-hit is one who must pay BP to do what others do for MP.

Candidates worth checking during the phase-1 audit, since vanilla already varies wildly here: Edgar's AutoCrossbow, Setzer's Fixed Dice, Shadow's Throw, Cyan's Quadra pair, Sabin's Bum Rush — several are already multi-hit in vanilla and may need no change at all.

[mtklein]

Second axis, from the owner (2026-07-29): multi-hit abilities "let you spend MP where you'd have spent BP before, or let you hit with a different weakness (e.g. claws vs. pummel)".

The first half is the resource substitution already noted. The second is a distinct and arguably bigger point: an ability's break class is independent of the equipped weapon's. Fight swings whatever you are holding; a Blitz, Tool or SwdTech carries its own class from Ot6SkillClassTbl. So a multi-hit ability can open an axis the weapon cannot reach — repeatedly, cheaply, with no equipment change.

That gives the hit-count table two dimensions to design against, not one:

1. Rate — how fast this character opens a gauge on the axis they already have.
2. Reach — whether they can open a second axis without re-equipping or spending BP.

A character whose only multi-hit is the same class as their weapon is fast but narrow; one whose multi-hit differs is genuinely two probes. That is a per-character identity question, and it is the sharper version of the 'weak attacker still useful' rule: they are useful because they can reach an axis nobody else in the party is carrying.

Concrete consequence for the audit: phase 1 must record, per ability, both the hit count AND its class relative to that character's equippable weapon classes. tools/check_break_reach.py already computes the second half — it models a party's fieldable classes from weapons plus command abilities — so the data is one query away. What it does not model is rate, which is exactly what this pass adds.

Worth noting a live example the checker already knows: Sabin's bare fists are OT6_BLUDG, so his unarmed Fight and a bludgeon Blitz are the same axis; his claws and AuraBolt's holy chip are not. The interesting rows are the ones where ability and weapon disagree.

[mtklein]

A cost curve the survey must include: status DOTs.

Owner playtest (2026-07-29): Bio Blaster's poison damage plus its status ticks ("little plink poison damage") break a 2-shield poison-weak enemy between them. Filed as #60 to establish whether DOT ticks genuinely reach the chip path.

If they do, it matters here: a DOT is multi-hit spread over turns, paid once. That is a third cost curve alongside the two this issue already has —

• rate (hits per action, paid per action),
• reach (a second class axis, paid by having the ability),
• and duration (hits per turn thereafter, paid once at application).

Edgar already has it, which sharpens his identity in the framing above: he does not merely hit a weakness, he installs one. Any hit-count table that ignores status abilities will misprice him.

[mtklein]

Moving to v0.10: phase 1 (the audit) and phase 2 (docs/design/multi-hit.md) landed in v0.9, but no data moved — Pummel is still ×1 and Bum Rush is still ×1. The build list in §10 is the remaining work, and it is gated on the owner's call between Bum Rush ×4 and ×8.

The audit finding worth repeating here: across all 256 MagicProp + 256 ItemProp records there are exactly three multi-hits in the entire game — Quadra Slam ×4, Quadra Slice ×4, Empowerer ×2. Cyan is the only character with a multi-hit ability. Every ×2/×4/×8 in kits.md's Chip column is intent that was never built, stated there as fact.