"Free, never pay-to-win" — then a livestreamed apology 15 hours after launch

One-liner: NCsoft's Aion 2 launched under a "free to play, never pay-to-win" banner. Players then found two things: supply boxes in the shop selling combat enhancement scrolls and resurrection stones outright, and MMO infrastructure — auction house trading, face-to-face trading, extra Abyss dungeon time — locked behind multiple tiers of paid membership. Fifteen hours after launch, the development team went live to apologise. The problem was not the price. It was that players walked in assuming these things came with the game.

What happened

Aion 2, developed by NCsoft, launched first in Korea and Taiwan last year; the mainland China service is handled by Shanda Games, and the global Steam release is set for 5 October 2026 (a global stress test opened on 17 September, peaking at 73,000 concurrent players on SteamDB).

The Korean launch numbers were excellent: over 1.5 million DAU, and 100 billion-plus KRW within two months. IP affection and Unreal Engine 5 visuals carried the opening.

Then the reception collapsed, in two places.

First, direct sales of stat items. NCsoft promoted "free to play" and "never pay-to-win" heavily before launch. Players then found supply boxes in the shop containing combat enhancement scrolls and resurrection stones — core stat items.

Second — and more importantly — core functionality locked behind multiple subscriptions. Auction house trading, face-to-face trading, and extra Abyss dungeon time sit behind different tiers of paid membership. Two player quotes capture the structure:

"You need the $15/month subscription just to unlock the trading house and auction house. Without an auction house this game is unplayable. Two subscriptions bundled is $30, so $360 a year. This is essentially a high-subscription, pay-to-win game."

"After claiming it was free and absolutely not pay-to-win, we assumed the worst case was a single subscription plus cosmetics. Instead all the substantive content is locked behind multiple subscriptions. It's a trap."

The result: 15 hours after the Korean launch, the development team had to go live and apologise — the same leads who had been beaming before launch, grim-faced on camera.

For contrast, the same reporting notes that Chinese titles going overseas, such as Where Winds Meet, sell cosmetics only and no stats, and drew consistent praise from Western players. NCsoft adjusted and delisted some disputed items during the Steam test, but whether the global launch rebuilds trust will only be clear after 5 October.

Why a free promise makes charging so much more expensive

Two mechanisms stack here, and they should be counted separately.

The first is the reference point. The marketing said "free." So the reference point players form on entry is not "this is a subscription MMO" but "I already have these things." In an MMO, the auction house, trading and dungeon time are infrastructure, not value-added content — plumbing and electricity, not interior design. Take infrastructure away and put a price on it, and the player does not experience "this service is worth $15." They experience "something I already had was taken from me."

Under prospect theory, a loss of a given magnitude lands roughly 1.5–2.5 times harder than an equivalent gain. The same $15 therefore feels entirely different as "a new subscription benefit" versus "ransom to unlock what existed." This is the same mechanism as yesterday's Destiny 2 card, from the other side: reclaiming anything inside a reference point is priced at 1.5–2.5x (cards/2026-09-23-destiny2-content-vault-reversal.md).

The second is commitment and expectation violation. Saying "never pay-to-win" before launch anchors expectations at "payment buys cosmetics only." When the actual configuration contradicts that, the reaction is not "expensive" but "you lied." The two have very different repair costs. Expensive can wait for a sale. Lying has to be bought back with time and action — which is why the apology came within 15 hours (public pressure), while trust rebuilding is measured in years.

What is sourced and what is inference. The source provides the facts: the marketing claims, the shop configuration, the player quotes, the 15-hour apology, the contrast case. "Infrastructure held hostage" and "reference-point deprivation amplifies the felt cost by 1.5–2.5x" are this card's inference — theoretically grounded, with no product data behind that chain.

This structure doesn't hurt whales first — it hurts the ecological base

Locking infrastructure behind subscriptions looks good on the near-term P&L. People who want to play have no choice, so ARPU rises immediately.

The cost lands elsewhere. Revenue concentrates heavily in a small number of high spenders, but overall activity is determined by non-payers; whales need an environment with other people in it, and free players are that environment. When the auction house and trading go behind a paywall, the first to leave are not whales (they can afford it) but the mid-and-light players who treat the MMO as a social space. When they go, market depth, matchmaking responsiveness and guild activity fall together — and what the whale feels as "is this game dying?" is second-hand.

The typical curve for this design is therefore: a strong first month, a felt sense of "where did everyone go" in months two and three, and no visible signal on the P&L — because revenue is concentrated in a few people whose payment stickiness decays far more slowly than activity does.

Scope, stated plainly:

• Does not apply: if the locked features are not essential to the core loop (extra storage tabs, cosmetic emotes, convenience items), the backlash is much smaller.
• Does not apply: if the subscription structure is stated from the marketing phase onward (traditional subscription MMOs charge before entry), the reference point never forms around "free," and the reaction is far milder. Where the reference point sits decides what kind of event this is.
• Weaker: for products aimed purely at whales and not dependent on a free-user ecosystem (some SLG), the cost is much lower.
• It reverses: when players are deeply invested (progress, social ties) and facing a certain loss, some will not leave — they will top up the subscription. That is exactly where the "trap" feeling comes from, and it is why short-term numbers look good. Loss aversion flips to risk-seeking under deep investment plus certain loss; one mechanism, opposite behaviours.

How to use this

Before putting a paywall on core functionality in a "free" product, ask one question: at the moment players enter, do they already assume they have this? If yes — either keep it in the free layer, or state the subscription in the entry conditions during marketing. Never move the reference point after launch.

Transferability:

• Holds across genres (MMO, competitive, social) because the mechanism sits on the player side.
• Most dangerous for utility features: trading, matchmaking, storage, dungeon entries — the preconditions for play, not embellishments.
• Safest for expression features: cosmetics, titles, emotes, display slots. Players have long accepted "pay to look better," so the reference point already sits on the paid side.

What you can do tomorrow

1. List everything in your product that players assume is free but is technically lockable (trading house, matchmaking, storage, dungeon entries, auto-battle). For each, write one line: would locking this make it a hostage? Anything marked yes does not go into a subscription bundle.
2. Check every external claim against the live shop. Pull all absolute phrasing — "free," "never pay-to-win," "no stat sales" — and verify item by item against what is actually on sale. Where they disagree, change the copy or change the configuration. Pick one; do not let it sit — the inconsistency itself is the broken promise.
3. If you are doing subscriptions, classify what you are selling: new value (cosmetics, convenience, extra content) or unlocking existing functionality. The latter must be announced one version earlier, so the reference point forms before the charge lands.

Appendix

Evidence and sources

• 《永恒之塔2》Steam测试爆火… — secondary reporting (Chinese industry media, retrieved via GameLook's full-text feed), 2026-09-24. Provides: Korean DAU and revenue, stat items in supply boxes, the multi-subscription structure, player quotes, the 15-hour apology, 73,000 concurrent during the Steam test, the 5 October release date, and the Where Winds Meet contrast

Evidence and sources (assessment)

• Source type: a single Chinese industry outlet's secondary reporting. Player quotes come from Western communities (Reddit and similar) and platform figures (SteamDB concurrency) are independently checkable, but the revenue and DAU figures carry no stated methodology or source.
• Evidence tier B: the event and structure are clear, with first-hand quotes and platform data; but there is no retention, payer conversion or ARPU data, no control group, and only one source without cross-checking. Under the filtering rules, a single-source conclusion without supporting data caps at B.
• Confidence: medium. The factual layer (marketing claims, shop configuration, timing of the apology) is likely multi-source; "locking features → ecological base erodes → long-term cost" is inference, unvalidated against this product.
• Measurement caveats: (1) "over 1.5 million DAU" and "100bn+ KRW within two months" are Korean launch-period figures with no stated measurement point or definition; (2) "73,000 concurrent" is a Steam stress-test figure, not post-launch performance; (3) "$15/month, $30/month, $360/year" come from a player quote — player arithmetic, not an official price structure; (4) "100 billion KRW" is left unconverted, per the language rules on preserving original figures.
• What is inference: "infrastructure held hostage," "reference-point deprivation amplifies felt cost 1.5–2.5x," and "good numbers first, cost visible in months two and three" are all this card's inference. Theoretically grounded; unvalidated on this product.

Scope and limits

• Genre: player-side mechanism, holds across genres. Most damaging in genres that depend on social and trading systems (MMO, open world), because the locked features are preconditions for play.
• Region: the case is a Korean product launching first in Korea/Taiwan and the resulting global reaction. Western players tolerate subscriptions far more readily than Asian mobile players do; whether "locked features" produces the same backlash in Western markets is unevidenced here — the most important transferability gap.
• Stage: the risk concentrates in one move — acquiring users with a free promise, then adding the paywall after launch. Products that state their subscription structure before launch are not affected.
• Regulation: a mismatch between "free" messaging and the actual monetisation structure can touch consumer protection and false-advertising rules (see kb/OPS-03-合规即设计约束.md). Not legally verified; flagged only.

What would falsify this

1. **If, after the 5 October global launch, Aion 2 keeps its multi-subscription structure and posts better 90-day-plus retention and revenue than cosmetic-only competitors**, this card fails — locking features carries no material trust cost in that genre and region.
2. If a counter-example exists — an MMO that locked core features like the auction house behind a subscription with no notable backlash and unaffected retention — the mechanism fails, or a moderating variable is missing (for instance, a market long accustomed to subscriptions).
3. If the backlash was driven mainly by direct stat sales (supply boxes) rather than by locked functionality, this card's combined attribution is wrong and the two must be separated. This is where it is most likely to be mistaken, since the reporting presents both threads side by side without weighting them.

Weakest link: the chain runs "broken promise + locked features → trust collapse → long-term cost." But Korea's 1.5m DAU and 100bn KRW show short-term revenue was strong, and the 15-hour apology reflects public pressure, not commercial failure. Establishing the long-term cost needs a 6–12 month retention and revenue curve for Korea/Taiwan, which this card does not have. "Long-term cost" is therefore mechanism reasoning, not an observed result.

Unverified

• 6–12 month retention and revenue curves for the Korea/Taiwan service (any third-party data).
• The actual global shop configuration after the 5 October launch: did the "adjustments and delistings" land, and did the subscription tiers change?
• The monetisation detail behind Where Winds Meet's overseas version, needed for a real control group (currently only one sentence of paraphrase).
• Whether the mainland China service (Shanda Games) matches the Korean monetisation structure.
• The relative volume of player complaints about stat sales versus locked features — this determines whether falsification condition 3 triggers.