The 90-Pull Architecture: Reading Genshin Impact's Gacha Machine Through a Data Analyst's Eyes

At 3 a.m. in Hai Phong, the transfer market closed long ago, yet another sheet of numbers is still scrolling on screen. A system runs on its own rhythm: 90 pulls for an insurance guarantee, a 50/50 split of risk, and a 21-day cycle repeating as regularly as a group-stage schedule. I do not play Genshin Impact. But I read its structure the way I read a transfer-contract clause, because both are machines that price three identical things: time, scarcity, and expectation. At 3 a.m., the market sleeps. That is when the numbers are most awake.

The document I am reading is an explainer of Genshin Impact's banner schedule — version 7.0 phase two, then 7.1 with two phases. It has no teams, no players, no standings. But it contains a monetization model that anyone doing esports-industry analysis should read carefully, because it is a closed revenue machine running on its own rules. What is worth studying here is the pricing architecture, not the schedule.

---

Context: When a sports analyst reads a gacha sheet

I work in transfer-market administration and esports data analysis. My job is to look at a number and ask: where is this number moving, who benefits, who pays, and what would change its direction. When I picked up the Genshin Impact article, the first thing I did was not read character names. I read the structure.

A player-transfer market has three core variables: contract length, fee, and performance expectation. A gacha system has exactly three corresponding variables: the banner's time window, the pity guarantee threshold, and the featured-character probability. Swap the names, and the two sheets speak the same language.

The source document describes Genshin Impact as an open-world action-RPG published by HoYoverse, operating on a gacha model. It has no professional tournament circuit, no franchised league, no transfer market in the sporting sense. This title's "versions" are PvE content drops, not competitive-balance patches. That is the most important distinction, and also the one a hasty analysis is most likely to blur.

I approach this document by the same principle I use when reading transfer news: evidence first, conclusion later. Of the 28 information points the source provides, only one cites an official HoYoverse source, three are the author's opinion, and the rest are mostly unsourced. That is a signal I never ignore. Before analyzing anything about strategy or value, I must assess the reliability of the very ingredients I am cooking.

An empty stadium taught me I was missing a variable: emotion does not sit in a spreadsheet. But emotion is what decides many of the numbers inside the spreadsheet. And that is exactly the intersection between sports analysis and gacha analysis.

---

The core: Pity architecture — the 90-pull contract

The first anchor of the entire system is the pity threshold. The source states clearly: players are guaranteed a five-star character within 90 pulls. This is the foundation of every spending calculation.

Set the game aside and look at it as a financial structure. A 90-pull guarantee threshold turns a random process into a contract with a risk ceiling. Players do not know whether they will hit on pull 3 or pull 87, but they know that by pull 90 they will have a result. The risk ceiling is what turns fragmented consumer behavior into a predictable spending plan.

In the transfer market, the equivalent structure is a release clause. A club does not know whether a player will explode or get injured, but it knows there is a price to exit the contract. The pity threshold is that exit price, differing only in that it sits on the buyer's side rather than the seller's.

The subtlety is this: the 90 threshold is not the average amount a player spends. It is the ceiling. Most players hit earlier, but that ceiling figure shapes every top-up decision. When I analyzed an import striker's profile, I did not look at the highest price in history. I looked at the floor price of a guaranteed contract. The gacha structure is designed the same way: it sells you reassurance about a ceiling, not a probability.

Every article I write begins by identifying what is raw data and what is interpretation. With the pity threshold, the raw data is very short: a figure of 90 and a probability. But the interpretation opens an entire chain of analysis about consumer behavior over time, and that is where my interest lies.

---

The 50/50 mechanism: splitting risk, doubling expectation

The second layer of the architecture is the 50/50 mechanism. The source describes: on an event banner, the first five-star has a 50% chance of being the featured character and a 50% chance of being a standard character. If the latter occurs, the next five-star is guaranteed to be the featured one.

This is the most analytically interesting design point in the whole system, because it creates a two-tier structure: a short-term risk tier and a long-term insurance tier.

From a psychological angle, this mechanism exploits something very human: the pain of loss is greater than the joy of an equivalent gain. A player invests 80 pulls, sees a golden light, and receives a standard character — that is an emotional shock. But at the same time, the system whispers: next time it is certain. The shock is neutralized instantly by a certain promise. This is the art of managing an emotional investment portfolio: sell the disappointment, bundle in the guarantee.

In esports, a similar model exists in skin programs and battle passes. You do not know whether you will get the item you want on the first opening, but you know that at a certain spending level you certainly will. Both are retention techniques that turn volatility into a journey with a clear ending.

The source provides another important detail: banners of the same type may share pity. That means pulls accumulated on one banner retain their value when you switch to another in the same category. Technically, this lowers the marginal cost of switching between banners, and lower marginal cost usually means higher action frequency. When friction falls, the number of decisions rises.

A night in Hai Phong taught me one thing: people look at the price sheet, I look at the movement sheet. A player does not spend more because one banner has a more attractive offer. They spend more because switching banners becomes cheaper in perception. That is the movement a static price sheet will never show you.

---

The 21-day cycle: the rhythm of cash flow

The source states that each version is split into two phases of roughly 21 days each, and each phase has its own banner. The number is small but carries great weight in revenue-rhythm analysis.

A 21-day cycle creates a recurring decision window. Split a version into two phases, and players get two spending decision points per version; at an online title's release frequency, this creates an almost continuous stream of decisions.

In football, this rhythm equals two matches per week. The injury literature I have followed shows that fixture density is the biggest culprit behind injuries, and no medical team can save you from two matches a week. In gacha, the 21-day rhythm creates no physical injury, but it creates another form of wear: budget wear and wear on the ability to resist temptation. Every time the window opens, the player must decide. Repeated decisions mean repeated errors.

Structurally, 21 days is long enough for players to feel they have time to save, but short enough that they do not grow bored. This is a time design aimed at expectation, not at price.

I have seen the same in the transfer market: the winter window is not when clubs spend the most, but it is when they decide fastest, because a window that opens will close. The gacha design exploits exactly that open-and-close window mechanism.

---

Reruns and controlled scarcity

The next layer is the rerun mechanic — re-offering older characters on limited banners. The source notes a notable feature: the rerun schedule is not fixed. Some characters are absent for more than a year, while others return within a few versions.

This is the intersection of game design and behavioral economics. A non-fixed rerun schedule creates controlled scarcity. Players cannot calculate precisely when the opportunity will return, so the cost of waiting becomes ambiguous. When the cost of waiting is ambiguous, the value of spending now rises.

Scarcity in gacha is unlike scarcity in aviation or agriculture, where supply is limited by nature. Here it is limited by the publisher itself, and that limitation is deliberate.

In sports, a team does not announce exactly when its star will return from injury, and that makes the ticket market swing. But in gacha, the ambiguity of timing is not the consequence of an external variable. It is a design tool.

This leads to a question I always ask when reading any document about a monetization model: who controls the time variable? If the answer is the publisher, then it is not a market — it is a pre-arranged chessboard. Players are not playing against probability. They are playing against a schedule.

---

Chronicled Wish: the legacy monetization lane

The source mentions a mechanism called Chronicled Wish — a separate banner type with its own rules, typically for older characters.

This is the most subtle point in the whole architecture, and the one an industry observer should note. A separate lane for legacy characters solves a basic problem of any content business model: how to monetize old assets without disrupting the new catalog.

In football, the same problem is solved with legends' friendly matches or retro shirt sales. You do not bring a retired star back into the main squad, but you can still sell the memory of him. Chronicled Wish is the video-game version of that strategy.

This matters because it shows the publisher has split two revenue lanes: one for new characters, one for legacy characters. This split reduces the pressure to return old characters to the main banner, while opening an additional revenue channel from players who joined the game late.

At 3 a.m., the market sleeps. That is when the numbers are most awake. And the most awake number here is the one showing a revenue model can exploit the same asset multiple times, across multiple lanes, without creating new assets.

---

The publisher as both referee and beneficiary

The only point in the source cited from an official source is a HoYoverse announcement. This creates a governance paradox worth analyzing.

In the esports ecosystem, the game publisher holds a similarly central role. They decide patches, schedules, transfer rules, and sometimes revenue sharing. But at least in esports there are independent tournament organizers, team associations, and third parties acting as checks and balances.

In the gacha model, the publisher sets the pity rules, publishes the probabilities, operates the banners, and collects the entire revenue. This is a higher degree of power concentration than most sports ecosystems, because no third party independently verifies the probability disclosures.

Random-reward disclosure regulations in many markets require published probabilities, and the source does mention pity rules and probability disclosure. But it cites no regulator. The publisher describes its own rules, and the reader must trust that description.

I once wrote about a case in which an organization published prettier numbers than reality, and it took months of independent data to uncover it. The lesson I drew was not to doubt every disclosure, but to recognize that when one party both sets the rules and benefits, independent verification becomes a mandatory condition, not an optional one.

---

Comparison with the esports monetization model

This is the genuinely transferable part of the document. Placed side by side, the two models reveal two revenue philosophies that differ in nature.

The gacha model relies on direct, recurring, in-game spending. Revenue comes from millions of small, continuous decisions made by the players themselves. It depends on no season, no tournament, and no external cultural event.

The esports model relies on a more complex combination: sponsorship, broadcast rights, skin revenue sharing, and prize pools. It is tightly bound to the schedule, to team performance, and to the appeal of stars.

The core difference lies in the degree of control: the gacha model controls almost the entire value chain, while the esports model depends on many parties and many variables beyond its control.

This makes the gacha model less vulnerable to calendar shocks, but more exposed to changes in random-reward regulation. One side weathers storms but not changes to the rules of the game. The other weathers rule changes but not a bad season.

A graph does not lie, but it does not tell the whole story. I look for the missing part. The missing part in this comparison is the human factor: both models make money from emotion, but only one can measure that emotion through the user's own spending behavior.

---

The contrarian angle: the document's biggest problem is itself

Here I must state plainly what many analyses avoid. The greatest value of the source is not the information it provides, but the way it exposes a reliability problem.

Of 28 information points, 20 are unsourced. Only one comes from an official announcement. Three are the author's opinion. Several character names and version numbers cannot be cross-verified against known game state. The document itself admits the exact banner schedule is still to be confirmed.

This is a positive signal of honesty, but also a confession that the content is provisional. An article that admits uncertainty is more trustworthy than one that claims certainty, but it is still an ingredient to verify, not a conclusion to act on.

The paradox is that the very ambiguity of the schedule is part of the scarcity mechanism I analyzed above. Players read articles like this to find an answer to the question "when," but the answer they get is "possibly." The uncertainty is not the article's fault. It is the system's nature.

I once wrote wrongly about a season because my beautiful possession data did not account for pitch temperature, high pressing, and champion psychology. Germany left the tournament, and my article was mocked for a week. I learned one thing: respect the model, but do not trust it absolutely. That lesson applies verbatim to this document.

The contrarian point is this: an article about gacha full of probability figures and schedules can still lead readers to a wrong decision, because it provides "when" but not "is it worth it." And in a model where everything is designed to optimize spending, the question "is it worth it" never has an objective answer from the seller's side.

---

What esports can learn

There are three transferable lessons I drew after reading this document.

First, insurance architecture. The 90-pull pity threshold is a way to turn random behavior into a plan with a risk ceiling. Any loyalty program in esports could apply this principle: let users know exactly what they will receive at a certain spending level, instead of merely advertising the probability of receiving it.

Second, deliberate ambiguity. Not publishing a fixed rerun schedule is a tool, not a lack of organization. Esports organizations usually try to publish everything transparently, but there are cases where withholding timing information creates higher expected value. This requires a very delicate ethical balance.

Third, the legacy revenue lane. Chronicled Wish shows a model can exploit the same asset across multiple channels without creating new ones. Esports teams with rich historical content libraries can learn to design nostalgic products with a similar structure.

My numbers do not need applause. They need to be right — time is the referee. These three lessons will be verified by how they operate in practice, not by whether they sound reasonable.

---

Risks and what to track

The greatest risk of the source is information reliability, not financial risk. A schedule explainer built on 20 unsourced points can lead readers to plan spending on false information. This is the kind of risk I call "acting on assumption risk."

The second risk is the possibility that the content was auto-generated or rumor-based. The appearance of unverifiable character names and version numbers increases this possibility. Before using any information from the document, it must be cross-checked against official channels.

The third risk, at industry level, is changes in random-reward regulation. If rules on probability disclosure and minor protection tighten, the entire pity and rerun architecture I analyzed may need adjustment. This is a variable every industry analyst should track.

Signals to track are official confirmation of the banner schedule, verification of character names, and any change in probability-disclosure rules in major markets. These signals will confirm or refute the value of the source.

---

A progressive thought

What I carry away after reading this document is not a list of banners. It is an awareness of how pricing machines operate. A gacha system and a player-transfer market share the same nature: both sell expectation, and both collect money from the gap between what the buyer knows and what the buyer wants.

People remember Hai Phong for the noise. I remember it for this: that the success rate came later. And in the story of this gacha machine, what I will remember is not the number 90 or the 50/50 ratio. It is the question the spreadsheet can never answer: when everything can be designed to optimize, what remains that is real?

Perhaps the answer lies in the part the data leaves blank. In the moment a player looks at the screen at 3 a.m. and wonders whether they are buying a character or buying a feeling. A spreadsheet cannot answer that. But precisely because it cannot, it is the most worthwhile question to ask.

The truth is I will still read articles like this, and I will still analyze them with sheets of numbers. But I will always leave one cell blank in the sheet. That blank cell is for the human part that every model, however perfect, can never capture. At 3 a.m., the market sleeps, and that is when I remember why I do this work: not to predict, but to honestly record what is happening, even when it cannot be measured.

Time is the referee. And the referee never blows the whistle early.