blushneybusiness-glitch/roblox-omniskills/rbx-pvp-balance

rbx-pvp-balance

When to load

Any "overpowered/underpowered/unfair" complaint, before shipping a new weapon or ability into an existing PvP pool, after a meta shift (one kit's pick rate doubles), and as a scheduled pass in rbx-wf-polish for PvP games. Load alongside the combat subskill that owns the numbers being changed.

TTK targets by genre

TTK = time from first hit to kill against a full-health target, assuming ~70% of shots/swings land. Measure it in a test place with fixed loadouts before arguing about feel.

Two derived laws: burst damage never exceeds 60% of max HP in one unavoidable hit (the victim must get one decision before dying), and healing throughput stays below ~50% of incoming DPS or fights stall into stalemates.

Tuning methodology

• One variable per patch. Change damage OR cooldown OR range, never two — a double change

is unreadable in the data and untrustable in playtests.

• Step sizes: damage/healing ±10%, cooldowns ±15–20%, ranges ±2 studs, windup/telegraph

±0.1s. If a kit needs more than a 20% cut, its design is wrong, not its number — redesign per the counterplay audit below.

• Nerf the outlier before buffing the field. Buffing 9 kits to chase 1 inflates TTK-drift

across the whole game; one nerf holds the baseline.

• Read data before feel: instrument kill events with weapon/kit id via

rbx-analytics-events. Healthy pool: per-kit win rate 48–52%, pick rate within 2× of the mean. Win rate 55%+ AND pick rate top-of-chart = nerf now; win rate high but pick rate low = a pocket pick, watch it; pick rate high but win rate ~50% = it's fun, leave it alone.

• Segment by skill. Pull win rates for top-decile players separately (by rating from

rbx-pvp-ranked). A kit that is 50% overall but 60% in top-decile hands is a skill-ceiling problem; nerfing it globally punishes beginners who were already losing with it.

• Ship balance patches server-side as CONFIG values so a hotfix is a value edit that

takes effect on soft shutdown, not a place republish players must rejoin for.

Counterplay audit

Run this table for every attack in the pool; any row with two "no" answers is a design bug regardless of its win rate:

1. Telegraph — is there ≥ 0.3s of visible windup (0.5s+ for attacks dealing > 40% HP)

rendered from the VICTIM's camera, not just the attacker's?

1. Answer — does at least one universal option beat it (block, dodge, interrupt, walk

out of range) that costs less than the attack costs the attacker?

1. Punish — if the answer is played, can the defender act before the attacker recovers

(≥ 0.2s recovery disadvantage on whiff/block)?

1. Escape — can a losing player disengage at all? Chain-stuns that exceed 2.5s of total

lost control, or gap-closers that outrange every escape, fail here.

1. Information — can the victim tell WHICH attack is coming, or do two moves with

different answers share one animation?

"Overpowered" complaints usually name a number; the audit usually finds a missing row 1 or 2. Fix the missing counterplay first — it's cheaper than chasing damage values forever.

Sharp edges

• CRITICAL — Balancing values the client controls: if damage, cooldown, or range live

client-side, exploiters set their own patch notes. Every tuned value is enforced on the server (rbx-combat-hitreg validates range/LOS, cooldowns per rbx-combat-abilities); balance work on unvalidated numbers is decoration.

• HIGH — Tuning from the complainer's clip: one kill-cam of a 100-0 combo triggers a

panic nerf of a 46%-win-rate kit. Symptom: forum outrage, healthy data. Fix: hold every nerf until kill-event data covers 500+ matches or the counterplay audit fails.

• HIGH — Testing balance only in even 1v1s: battlegrounds and BR fights are mostly

uneven (2v1, gear gaps). Symptom: kit fine in duels, oppressive in groups (AoE, chain CC). Fix: audit rows 4–5 specifically in an outnumbered scenario playtest.

• MEDIUM — Patch-note silence: shadow-nerfing a popular kit reads as a bug and doubles

the complaint volume. Fix: publish the change and the reason ("52%→ was 58% win rate") — players tolerate nerfs they can see the math for.

• MEDIUM — Balancing mobility last: +2 studs of dash range warps more matchups than +10%

damage because it changes WHO can be attacked, not how fast they die. Treat mobility steps as half-size (±1 stud, ±0.05s).

Load with

rbx-combat-melee / rbx-combat-ranged / rbx-combat-abilities (owners of the numbers), rbx-analytics-events (kill-event instrumentation), rbx-pvp-ranked (skill-segmented data), rbx-design-difficulty (skill floor/ceiling framing).