October 24, 2026

Invasion, decoded — the ladder's balance table

Version 27.0.0 ships a 23-key tuning block for Invasion — the mode where you spend an Invasion Ticket to attack another player's world ("Used to invade another player's world", says the item's own description) and defend yours in return. The tables and the 237-mode locale strings have been in the client for builds; what's new in 27 is a INVASION_BAL_* remote-config family that tunes how its battles resolve. Here's every key, read against the mode's own tables — with the naming-based readings marked as exactly that.

The mode on the ground, from its tables

Invasion is a ranked PvP ladder: six tiers, promotion by rank position or by placing top-percentile, matchmaking and score gain scoped to your own level of investment through costume-rarity bands, win-streak checkpoints, a coin bank, pass missions (plain wins, streaks, HP-remaining, auto-battle) and a revamp shop.

The matchmaking bands are the interesting engineering: instead of a raw power number, the ladder brackets you by what you own — top-range players are matched inside a costume-rarity glass ceiling so a fully built account can't farm novices, while everyone sees their score-update range shrink as ranks get tighter:

The ops keys — what ran before this build

These were already in the remote config at 26.2.0 (shop refresh costs, coin supply, the season-end countdown, the coin bank toggle, CP reward thresholds): the business plumbing. They don't move in 27.0.0 — and none of the new keys touch them.

The new balance block, key by key

Race pacing

Reading (inferred from the names): invasions resolve on a timeline. A 10-second base window (RACE_T_BASE), final clear times clamped between 3 and 25 seconds (FINAL_T_MIN/MAX), and a gap exponent of 1.0 — linear — deciding how differences between the two sides spread across that window. A race frame explains the cap: the system won't simulate a walkthrough-win shorter than 3s nor stretch one past 25s, whatever the stat gap says.

Power realization

• DISPLAY_POWER_WEIGHT=0.8 — 80% of the outcome model leans on the displayed power number, the last 20% presumably on live performance. Never 1.0: the mode always keeps some luck.
• REALIZE_BA=0.65 vs REALIZE_SKILL=1.0 — how fully each damage source "realizes" in the model: skill damage counts at face value, basic attacks at 65%. In a race against a defence, auto-attack DPS chews slower than the paper math says — a classic anti-alt-farm trim.
• DAGGER_BA_MULT=1.2 — the class knob right behind that trim: dagger basic-attacks get a 1.2× inside the model, partially reimbursing the 0.65 trim for the BA-heavy weapon family.
• MIN_NET_DPS_FRACTION=0.5 — the floor: however the trades resolve, each side keeps at least half its nominal DPS — fights can't degenerate to zero.
• MAX_REGEN_RATE=0.08 — healing per tick caps at 8%: no lifesteal-and-wait stall comps in invasion.

Outcome margins

A margin curve shapes how the power ratio between defender and attacker converts into the outcome: centered at a 10× ratio (RATIO_CENTER), extrapolating up to 6× worth of effect (RATIO_MAX_MUL), with the margin's growth at slope 0.5, capped at 75% (with an extended cap variant at 95%), and a flatter 0.4 tail slope for the far end of the curve. Read plainly: overwhelming mismatches still cap out — dominance stops at the cap, and the tail never compounds linearly.

Debuffs and cleansing

Both sides' dispel and self-cleanse uptime sits at 35% with an 8s reference cooldown, and no debuff can ever strip a stat below 10% of its value (MIN_DEBUFF_MULT_FLOOR). The floor is the same anti-softlock idea as the DPS floor: control effects can cripple, never erase.

Schedule jitter

Day endings scatter up to 60s and season ends up to 10s. That's desync-avoidance and anti-sniping plumbing — final seconds of a season no longer gather at one timestamp, server-side.

What this means for players

• The balance is server-side and reversible. All of it lives in dashboards, not APK patches — the numbers here can move under the client without a single byte changing. When they do, we'll diff the config.
• Dagger carriers get an explicit boon (1.2× BA) and the model's race-window (3–25s) rewards burst speed — dash-and-burst kits, the same families the tier table reads out.
• Dominance is capped by design: margin caps at 75–95% and debuffs floors at 10%. Nothing here turns investment into a foregone conclusion; the ladder keeps paying people for playing.

What we can't see

• Which curve the client actually runs. All 23 keys are remote config — the engine is server-side and closed. Every "reading" above is inferred from the key name, its value shape, its neighbours, and the mode's tables, in that order. It's a faithful reading, not a decompilation claim.
• Whether the mode turns on for everyone in 27.0.0. The tables and strings are complete, but rollout flags haven't been seen flipped in any config we hold; treat it as staged, not announced.

Data from the 27.0.0 APK tables (game v27.0.0), diffed against the 26.2.0 extraction. XP Hero is developed by Supercent — XPHero Fans is unofficial and not affiliated with or endorsed by them. Remote-configured balance can change between builds and regions; nothing here is pull advice.