The Pity System and the Pricing of Probability: A Sports Betting Analyst's View

"Numbers do not lie; only the people who read them do."

I still remember a July afternoon in Chicago, when my model confidently predicted Germany would beat South Korea on the back of 74% possession. The match ended 0-2. That was the first time I understood that intuition is the enemy, and data is the only friend that never flatters. Years later, reading an analysis of the character-wish system in Genshin Impact — a title with no tournaments, no teams, no players — I realized there was something the sports betting industry should study: how a probability engine is designed to sell belief, and how it quietly admits that most of what it offers cannot be verified.

Esports has no ball, but it still has rhythm and probability to measure. This wish system does not even have a rhythm of competition — it only has a rhythm of spending. And spending rhythm, if you look closely, is a variant of betting rhythm.

Context: a game detached from every definition of sport

Before diving into numbers, one thing any data practitioner must acknowledge. Genshin Impact, published by HoYoverse, is a single-player or co-op open-world action RPG operating on a gacha model — players spend premium currency for a chance at random characters or weapons. It has no official professional circuit, no Worlds or TI equivalent, no club system, no transfer market in the sporting sense. Its "versions" are PvE content drops, not competitive balance patches.

That means the nine-dimension framework I usually apply to esports simply does not fit. There are no rosters to grade on paper strength. No player form curves to draw. No injuries, no team chemistry, no upset risk. Instead, the only thing worth analyzing is a probability-pricing system — and that system is engineered far more subtly than outsiders usually assume.

The original analysis I read was long, carrying 28 information points. Notably, 20 of them carried no source, only one cited an official HoYoverse announcement, and three were explicitly the author's opinion. Methodologically, that ratio is alarming. But in substance, a few numbers were concrete enough to put on the scale.

Core: dissecting the probability engine

The wish system runs on two axes. The first is the pity mechanic — a soft guarantee. Every 90 pulls, a player is guaranteed a five-star character. This is an "insurance floor" that turns a random event into an implicit time-bound contract. The second axis is the 50/50 system: on the first five-star pull on an event banner, the chance of getting the featured character is 50%, and the other 50% falls to a standard character. If it lands on a standard character, the next five-star is guaranteed to be featured.

Together, these two mechanics produce a structure familiar enough to chill me: it turns expectation into a predictable number while preserving the thrill of uncertainty. Designers do not sell you the chance to win. They sell you the gap between the feeling of control and the reality of randomness.

Another key detail is shared pity across banners of the same type. When pity is shared between a debut banner and a rerun banner, the marginal cost of switching from "pulling for a new character" to "pulling for an old one" falls. This is a revenue-smoothing design — it keeps money flowing across both the launch window and the rerun window, rather than letting players concentrate spending in one window and leave.

On pacing, each version cycle splits into two phases of roughly 21 days, each with its own banners. I call this "time-boxed revenue rhythm" — a structure creating recurring, predictable purchase windows. It is fundamentally different from the rhythm of sporting competition, which is set by fixtures and on-field results, not by a content-release calendar.

The most behaviorally notable point is the absence of a fixed rerun schedule. Some characters are absent for over a year; others return within a few versions. This uncertainty, read through market logic, is a deliberate scarcity mechanism — the equivalent of "limited-time events" in retail. It turns waiting into a psychological cost, and turns the decision to pull into one pressured by fear of missing out.

On top of that is the Chronicled Wish, a separate banner type for older characters with its own rules. In the financial language I am used to, this is a secondary revenue lane for inventory assets — it lets the publisher re-monetize characters that no longer generate income on primary banners, without disrupting the cadence of debut banners. Transfer summer is where emotion is most expensive but data is cheapest; in this system, the rerun window is a miniature transfer market, where player emotion is re-priced every 21 days.

What I want to stress is the mathematical precision of the structure. A 90-pull insurance floor plus a 50/50 chance is not an arbitrary arrangement. It is a balancing act between two opposing goals: making the reward feel within reach, while maximizing spending variance. A light spender can still win early. A heavy spender can still lose repeatedly. That variance is the core revenue driver, not the drop rate itself.

Contrarian: when the source is weaker than the numbers

Here I must say plainly what many analyses in gaming and betting avoid. The biggest problem with the original piece is not the wish system — that system runs clearly and can be verified through official announcements. The problem is the article itself.

Twenty of 28 information points carry no source. Some entities named — Odette, Flins, Ineffa, Vesna, Vodyanitsa, or versions 7.0 and 7.1 — cannot be cross-checked against known game state. This is a signal I learned from betting: when a source presents many specifics that cannot be traced, the biggest risk is not false information, but fluently fabricated information. I don't trust intuition; I trust a long enough data series — and here, the series is too short to trust.

Interestingly, the piece admits its own uncertainty. It states that the exact banner schedule is still to be confirmed. That is a positive reliability signal, but it also concedes that much of the content is provisional. When an article both promises a future version and admits doubt, its usefulness drops sharply — unless readers treat it as a list of questions to verify, not a report to act on.

There is a temptation I once fell into and now try to avoid: forcing data into a pre-written story. The systems architect in me always wants things tidy, structured, with clear conclusions. But data does not serve structure. If I assign the role of "players" to characters, the role of "tournaments" to banner phases, I would create a false equivalence — and that is the kind of error that collapses betting models.

Here, the only honest comparison between gacha and sport lies at the pricing layer, not the competition layer. Esports lives on sponsorship, broadcast rights, skin revenue sharing and prize pools — a decentralized ecosystem dependent on many parties. Gacha lives on direct, recurring, in-game spending, controlled by a publisher that holds almost all supply and information. The publisher is simultaneously the rule-maker, the operator and the announcer. No independent arbiter verifies those numbers.

This concentration of power makes gacha less exposed to fixture shocks than esports — but more exposed to regulatory shifts on probability transparency and consumer protection. Every time the market panics, I reopen old data and find what others forgot; here, what is forgotten is regulatory risk, a variable that schedule-type articles rarely quantify.

It is worth adding that gacha spending, under most current legal frameworks, is not gambling. But it sits at the edge of loot-box and minor-protection debates. For a betting analyst, that proximity matters: it reminds us that the line between a consumer product and a risk product is drawn by regulation, not by nature. And regulation can always change.

Takeaway: implications and what to track next

There are three signals I will keep on my watchboard, and I advise readers to do the same before drawing any conclusion.

First is official confirmation of the next version's banners. Only when HoYoverse's official channels name the characters in each phase will we know whether the original analysis was right or wrong. Second is verification of the entities named. If characters like Odette, Flins, Ineffa, Vesna or Vodyanitsa appear in official materials, the article's credibility is restored; if not, we are entitled to doubt the entire chain. Third is regulatory developments on probability transparency in major markets, since that is the only variable that could restructure the economics of the whole system.

What I take away after years of reading football data, esports data and gacha data alike is a humble realization. Data cannot fully capture sudden breakthroughs, whether it is a sixteen-year-old talent at the Euros or a scarcity mechanism designed to touch the crowd's psychology. Every model has boundary conditions, and the boundary condition of this model is the reader's trust — the thing most easily misled by a fluent article with 20 unsourced data points.

As a sports betting analyst, I do not see a match to predict here. I see a lesson in how probability is packaged and sold. If you want to read this system as a model, start from the number 90 — the insurance floor — then ask yourself: if every outcome is guaranteed to happen at latest after 90 pulls, is someone selling you a chance, or selling you waiting time packaged as a ticket?

When football pauses, PPDA keeps showing me who is really pressing. And this time, when the pitch pauses, the numbers keep showing me who is really spending — not because they want to, but because a system is designed to make them feel they are about to win. That is what every probability model, whether branded as sport or entertainment, exploits. Don't ask who plays well; ask who should have won — and in this game, the designed winner has never been the one pulling.