The 90-Pull Guarantee and the Unfixed Rerun Calendar: The Revenue Engine Esports Cannot Copy

On the last night of a banner cycle, the Seoul studio held nothing but the hum of a fan and the clatter of keys. I was supposed to be finishing a script about the mid-season transfer window, but a second monitor stayed open on a Korean livestream where a host was pulling for characters in Genshin Impact on camera. The chat scrolled too fast to read: four hours left, three hours left, sixty minutes, "nobody sleeps until it drops." When the banner closed, the message rate collapsed in under a minute. That digital room emptied in silence, and nobody had lost a match.

I have covered sport long enough to know that audience emotion usually attaches to an event that can be retold: a stoppage-time goal, a red card, a knockout round. The rhythm I had just watched belonged to a different species. It ran on a clock, and that clock was set by exactly one company. For a week afterwards I replayed the recording, counted the moments the chat changed tempo, and wondered how a sales mechanism had drawn an emotional curve that looked like a tournament bracket.

The calendar behind the emotional curve

Genshin Impact is published by HoYoverse and launched worldwide on 28 September 2026. It is an open-world action role-playing game, single-player or co-op, monetised through gacha. It has no professional tournament circuit, no teams, no player-transfer market and no competitive balance patches. Everything called a "version" here is a PvE content drop.

Its time structure is steady: each version splits into two phases of roughly three weeks, each phase opening one or more banners. A banner is a wish pool where players spend premium currency for a chance at limited characters or weapons.

The material I read for this episode described the schedule for version 7.0 phase two and version 7.1 with its two phases: phase one introducing two new characters simultaneously, phase two consisting of reruns. That is all the calendar we get. No strength data, no kit breakdown, no ranking accompanies it.

On reliability, one thing must be said plainly. Of the 28 information points in that material, 20 carry no source, exactly one cites an official publisher announcement, and three are the writer's opinion. Several names, including Odette, Flins, Ineffa, Vesna and Vodyanitsa, along with version numbers 7.0 and 7.1, cannot be cross-checked against known game state, and the material itself concedes the banner schedule is still awaiting confirmation. I keep that caveat in place throughout this piece, because an analysis built on an unverified schedule is worth exactly as much as a hypothesis.

So the interesting question is not who arrives next, but the commercial architecture behind the calendar. That architecture is real, has run for years, and can be described in concrete numbers. One thing my job has taught me: when a system refuses to disclose performance data, you can still read it through its pricing design.

The 90-pull guarantee: turning chance into a capped-price product

Three rule layers sit behind every pull. First, a player is guaranteed a five-star character within a maximum of 90 pulls. Second, on a limited banner the first five-star has a 50 per cent chance of being the promoted character and a 50 per cent chance of a standard one; if you lose, the next five-star is guaranteed to be the promoted character. Third, accumulated pulls are shared across banners of the same category.

The first layer is a quiet but extraordinarily effective psychological device. Once the maximum cost of an item can be calculated in advance, players stop comparing it to gambling and start comparing it to an order. Every ethical debate about spending gets pushed behind a very technical question: how many more pulls do I need.

The core lies here: the 90-pull guarantee does not remove risk, it merely frames risk as a ceiling price that players can compute themselves. For the same character, a lucky player pays a third, an unlucky one pays the full ceiling, and both leave the banner feeling they bought something rather than gambled on it.

The second layer is where the money is made. The 50/50 mechanic raises variance without touching the listed price. A player who loses the first five-star keeps pulling, because the next one is guaranteed, and that moment of losing is the single highest-propensity moment in the entire life of a banner. No scoreboard records this, yet it is a more important operating metric than anything I have seen on a match statistics sheet.

I am used to reading PPDA, xG, passes into the box in the final fifteen minutes. Applying those measures here would manufacture a false precision. There is no opponent, no tactical error, no match to win. There is only a price structure, engineered so that players convince themselves.

Shared pity: cutting friction to raise spending frequency

Sharing accumulated pulls across banners of the same type sounds like a concession to players. It is in fact a revenue-smoothing mechanism. If the counter reset every time you switched banners, players would be locked into one banner and would deliberate at length before leaving. When the counter carries over, the marginal cost of switching to another banner is close to zero.

The insight most readers miss: when switching costs between a debut banner and a rerun banner fall close to zero, the total number of windows in which players are willing to spend does not shrink, it grows. They do not spend more in one sitting; they spend in more sittings. In cash-flow terms, that is the difference between one explosion and a steady stream.

In football, we call that tempo control. Here tempo is controlled by a status line showing how many pulls remain. It converts an emotional decision into a progress bar. Progress bars are what stop people from walking away midway.

An unfixed rerun calendar and fear, packaged

The next notable feature is the absence of a fixed rerun schedule. Some characters are absent for more than a year; others return within a few versions. The publisher commits to no regular rotation.

That unpredictability is the product, not a flaw. When the calendar cannot be guessed, the only rational response for a player is to hold a reserve at all times. Reserves tend to get spent, and which banner absorbs them is decided by the publisher. An invisible hand, considerably simpler than any theory of free markets.

The Chronicled Wish mechanic, a separate banner type for older characters, adds a secondary revenue lane. It allows dormant inventory to be monetised again without disrupting the cadence of primary banners. This is the inventory-management problem every club commercial department dreams of solving: reselling old assets without diluting the new line.

What is fascinating is that the community does not describe these returns in the language of scheduling. It uses the language of fate. A long-absent character gets retold as a reunion, and that story is stronger than any announcement. A transfer is not real until someone agrees to tell it as a destiny. Here too: a returning banner carries no weight until people narrate it as a homecoming.

Who writes the rules, who collects the money, who speaks

The power structure of this model is compact to the point of being unsettling. The publisher is simultaneously the rule-maker of the pull system, the sole beneficiary of revenue, and the only official voice. Of the 28 information points I read, only one cited an official source, and that source was the publisher itself. There is no independent arbiter and no third party verifying the rates.

The legal landscape around this model is far from empty. Japan tightened rules on "complete gacha" in 2026 after a wave of player backlash. China issued a requirement in 2026, effective in 2026, that games disclose probabilities for randomised items. The Belgian Gaming Commission concluded in 2026 that paid loot boxes constitute gambling, and the Netherlands reached a similar conclusion the same year. Gacha is not classified as gambling in most markets, but it sits adjacent to that boundary, and any regulatory change strikes directly at the price architecture.

The biggest risk to this engine is not a losing season, it is a regulatory document. Gacha revenue does not depend on weather, venues, fixture lists or anyone's form. It depends only on whether the rules of play stay in place.

Why esports struggles to copy this engine

The value of the esports chain comes from indirect sources: sponsorship, broadcast rights, prize money, merchandise, revenue shares from in-game skins. Every stream of money passes through a third party. A tournament organiser needs teams, teams need players, players need audiences, audiences need a stable competitive calendar. The chain is long, and the weakest link always sits beyond anyone's control.

I remember the summer of 2026, when the global sporting calendar stopped. I was making a podcast interviewing supporters about their stadium memories, and I learned something no dataset could have taught me: sporting emotion can survive without a single match being played. A summer without crowds, yet we still rehearsed for an audience that only existed in imagination. But the price was an entire industry living on forecasts about a future nobody could guarantee.

A banner cycle never postponed. It needs no crowd in a stadium, no visas, no renewed sponsorship contracts. It needs a server running and a countdown clock. That resilience comes not from a better model, but from the publisher owning both the supply and the information channel. Esports enjoys no such privilege, and every attempt to copy the model stumbles right there.

Where I could be wrong

I should argue against myself before finishing, because analysis without self-doubt is just advertising.

If the publisher publishes a fixed rerun calendar in the coming versions, my thesis of engineered scarcity weakens considerably. Once the schedule becomes predictable, players shift from permanent reserves to long-term planning, and total spending could fall.

If sharing pull counts across banners of the same type makes players patient rather than spendthrift, my revenue-smoothing argument reverses entirely. I have no behavioural data to settle this, and I would very much like some.

And the biggest risk remains the sourcing. Twenty of twenty-eight information points carry no source. If the version 7.1 schedule changes, every concrete example in this piece loses its illustrative value, even if the structural analysis of pricing holds.

One more thing nags at me as I write this. I was once a laughing stock because of my pronunciation, and now I am the voice people choose every evening. Having stood on both sides of scepticism, I know that an unverified claim does not become true simply because it has been repeated often.

My call

Over the next three months, I expect the spending pressure of this cycle to concentrate in the first phase of the new version, where two new characters arrive together, rather than in the rerun phase. If that holds, first-phase revenue will be heavily skewed towards players who had already accumulated pulls, and the second phase will show markedly lower spending despite a more attractive character list.

The next thing worth watching is not which character returns, but whether the publisher agrees to publish a stable calendar framework. Such a framework would signal that regulatory pressure has grown large enough to justify trading opacity for transparency. If the calendar stays fluid, it tells us the clock remains the single most valuable asset in the game.

The widest stadium is not where the crowds are, but where people agree to listen. For the sports industry, the question is not how to build a sales engine as seamless as this one. The question is whether we are gradually, voluntarily, handing the power to write the rules, collect the money and speak for the record into a single set of hands.