Short answer: your odds of getting the featured 5-star character in Genshin Impact’s 50/50 system are exactly 50% on your first 5-star pull of a banner, climbing to a guaranteed 100% by your second. The base pull rate is 0.6% per wish, soft pity kicks in hard around wish 74, and hard pity locks in a 5-star by wish 90. Since version 5.0 (live August 28, 2025), a mechanic called Capturing Radiance nudges that first-roll win rate from 50% up to an effective 55%. Priced out, a single guaranteed pull run costs somewhere between $178 and $356 depending on how the coin flip lands. This piece walks through where every one of those numbers comes from, and what they mean for your wallet.
What “50/50” actually means in Genshin Impact
The term “50/50” gets thrown around loosely, so it’s worth pinning down. It refers to a single moment: the instant you pull a 5-star item on a Character Event banner. At that moment, the game rolls a coin flip between two outcomes. Half the time you get the featured limited character advertised on the banner. The other half you get a random 5-star character from the game’s permanent standard pool instead, a mechanic broken down in detail by GenshinTactics’ 2026 pity system guide.
Losing that flip isn’t a dead end. HoYoverse built in a guarantee: if your 5-star turns out to be a standard-pool character, your very next 5-star pull on a Character Event banner is locked to the featured unit, no coin flip involved. That single rule is why players talk about pity in terms of “cycles” rather than raw pull counts. One cycle can end in a win. Two cycles, worst case, always end in a win.
Compare this to Wuthering Waves, where the base 5-star rate sits at roughly 0.8% with its own separate 50/50 structure, or Zenless Zone Zero’s system, where the published base rate runs closer to 0.6% with a different pity cap. Every gacha shooter and RPG on the market has copied some version of this coin-flip-plus-guarantee shape, but the exact percentages and pull caps differ enough that porting intuition from one game to another gets you burned.
The published base rate, soft pity, and hard pity numbers
HoYoverse discloses two hard numbers for Character Event and Standard banners. The base rate for pulling any 5-star item is 0.6% on every wish from pull 1 through roughly pull 73, per the game’s own in-client probability disclosure and the community wiki’s Wish page. Hard pity, the point at which a 5-star is mathematically guaranteed regardless of luck, sits at wish 90.
Soft pity is the part HoYoverse doesn’t publish an exact curve for. What’s confirmed through years of aggregated pull data (and cited across guide sites like GenshinTactics) is that the probability starts climbing sharply around wish 74, and by the high 80s it’s approaching certainty. Because the exact per-pull curve is reverse-engineered rather than officially stated, treat any table of soft-pity percentages as a well-supported estimate rather than a HoYoverse-published fact. The two numbers you can bank on without caveats are the 0.6% floor and the wish-90 ceiling.
There’s a second, smaller pity layer too. Every 10 wishes without a 4-star or better guarantees you one, a rule that’s been stable since launch and applies across both Character Event and Standard banners.
| Mechanic | Rate / Threshold | Source |
|---|---|---|
| Base 5-star rate (wish 1-73) | 0.6% per wish | Genshin Impact community wiki |
| Soft pity start | ~Wish 74, rate climbs sharply | GenshinTactics 2026 guide |
| Hard pity (guaranteed 5-star) | Wish 90 | Genshin Impact community wiki |
| 50/50 win rate (base, no prior loss) | 50% featured / 50% standard | GenshinTactics, HoYoLAB |
| Guarantee after a loss | 100% featured on next 5-star | Genshin Impact community wiki |
| Capturing Radiance conversion (1st loss) | 33.333% chance next loss converts to featured | BitTopup, HoYoLAB Q&A |
| Capturing Radiance conversion (2nd loss) | 66.667% | BitTopup |
| Capturing Radiance conversion (3rd loss) | 100% (guaranteed) | BitTopup |
| 4-star pity | Guaranteed every 10 wishes | Genshin Impact community wiki |
Capturing Radiance: the mechanic that changed the math in 2025
For four years the 50/50 was a flat coin flip with a one-loss safety net, full stop. That changed on August 28, 2025, when version 5.0 introduced Capturing Radiance, confirmed directly by HoYoverse staff in a HoYoLAB community Q&A post. The mechanic only activates on a pull that would otherwise be a loss. When it triggers, that losing roll gets converted into the featured character on the spot, instead of forcing you into a second pity cycle to earn the guarantee.
The conversion chance escalates with consecutive losses, based on data compiled by BitTopup and cross-referenced against player-tracked pull logs. Your first loss carries a 33.333% chance the system converts a subsequent loss into a win. A second consecutive loss raises that to 66.667%. By a third consecutive loss, conversion hits 100%, meaning the system will not let a player lose four 50/50 rolls back to back. Averaged across the full distribution of outcomes, this pushes the effective long-run win rate from a flat 50% up to roughly 55%, a figure echoed by multiple guide sites tracking the mechanic since launch, including BuffGet’s version 6.0 Capturing Radiance breakdown.
What Capturing Radiance does not do is touch the base pull rate, soft pity curve, or hard pity ceiling. Those three numbers are exactly what they were before version 5.0. It’s purely a loss-mitigation layer sitting on top of the existing structure, and it only ever helps you, never hurts.
The pity-cost table: what a guaranteed featured character actually costs
Here’s where the abstract percentages turn into a dollar figure. One wish (an Intertwined Fate) costs 160 Primogems, a rate that’s held steady since the game’s 2020 launch. Genesis Crystals convert to Primogems 1:1 once purchased. The most efficient large cash pack currently listed is $99.99 for 8,080 Genesis Crystals, according to BuffGet’s 2026 Genesis Crystal top-up guide, which works out to 50.5 wishes and an effective rate of $1.98 per wish. That per-wish rate is what the table below uses to price out every pity scenario.
| Scenario | Wishes needed | Primogems | Cost at $1.98/wish |
|---|---|---|---|
| Best case: 5-star lands early and wins 50/50 | 1 wish (theoretical floor) | 160 | $1.98 |
| Typical case: any 5-star via soft/hard pity | 62.5 wishes (published community average) | 10,000 | $123.75 |
| Average cost to guarantee featured (classic 50/50, no Capturing Radiance) | 93.75 wishes | 15,000 | $185.63 |
| Average cost to guarantee featured (with Capturing Radiance, 55/45 effective) | 90.6 wishes | 14,500 | $179.44 |
| Worst case: lose 50/50, no Capturing Radiance conversion, need full second cycle | 180 wishes | 28,800 | $356.40 |
The 180-wish worst-case figure lines up with a separate calculation published by BuffGet’s own pity-system breakdown, which independently arrives at 28,800 Primogems as the ceiling cost for a guaranteed featured pull across two full hard-pity cycles. Two independent sources landing on the same number is about as close to confirmation as you get in a system HoYoverse doesn’t fully document.
Step-by-step expected-value calculation
Here’s how the 93.75-wish average figure gets built, using only the published rates above. No shortcuts, no invented numbers.
- Start with the average pulls per 5-star. Community-aggregated pull data (cited by Sportskeeda and matching HoYoverse’s own disclosed rate table) puts the effective average rate at 1.6% per wish once soft pity is factored in across the full 1-90 range. That means 1 divided by 0.016 equals 62.5 wishes on average to land any 5-star, featured or not.
- Apply the 50/50 split. Half the time (50%) that 5-star is the featured character on the first try, costing exactly 62.5 wishes.
- Model the loss branch. The other half (50%) of the time, you get a standard-pool character and enter the guarantee. Your next 5-star, another 62.5-wish average cycle, is 100% the featured unit. That branch costs 62.5 + 62.5 = 125 wishes.
- Weight the two branches. Expected wishes = (0.50 x 62.5) + (0.50 x 125) = 31.25 + 62.5 = 93.75 wishes. That’s the average cost to guarantee the featured character under the classic (pre-Capturing Radiance) 50/50 rule.
- Adjust for Capturing Radiance. With the mechanic’s effective 55/45 split (per BitTopup’s tracked conversion data), the same formula becomes (0.55 x 62.5) + (0.45 x 125) = 34.375 + 56.25 = 90.625 wishes.
- Convert to spend. 90.625 wishes times 160 Primogems equals 14,500 Primogems. At the $99.99-per-8,080-Crystal rate, that’s 14,500 divided by 8,080, times $99.99, which is roughly $179.44.
Every input in that chain traces back to a cited rate: the 1.6% average pull rate, the 50/50 (or 55/45 with Capturing Radiance) split, the 160-Primogem wish cost, and the $99.99/8,080-Crystal pack price. Nothing here is a HoYoverse-published “expected value” figure, because the company doesn’t publish one. It’s arithmetic built entirely on their own disclosed rate floor and pity ceiling.
Five worked examples with real numbers
Example 1: The lucky roll. A player hits their 5-star at wish 90 (hard pity, worst timing but still a win) and takes the 50/50. That’s 90 wishes, 14,400 Primogems, roughly $178.20 at the $1.98/wish blended rate.
Example 2: Capturing Radiance saves the run. A player loses their first 50/50 at wish 85. Under the 33.333% first-loss conversion rate, there’s roughly a one-in-three chance that very pull flips to the featured character instead of costing a second cycle. When it hits, the total spend stays at 85 wishes, about $168.30, with no second pity cycle required.
Example 3: The full double-pity worst case. A player loses the 50/50 at wish 90, Capturing Radiance doesn’t convert, and the guarantee forces a second full cycle to hard pity. Total: 180 wishes, 28,800 Primogems, $356.40. This is the absolute ceiling. It cannot get more expensive than this for one guaranteed featured character.
Example 4: Paying with Welkin Moon instead of cash packs. Blessing of the Welkin Moon costs $4.99 for 30 days, delivering 300 Primogems immediately plus 90 per day for 30 days, for 3,000 Primogems total, a figure confirmed by BitTopup’s Welkin Moon value guide. That’s 3,000 divided by 160, or 18.75 wishes per $4.99, about $0.266 per wish, roughly 7.4 times cheaper per wish than a one-off Genesis Crystal pack. To reach the 90.625-wish average needed for a guaranteed featured pull, a player would need about 4.8 months of Welkin Moon, or roughly $24 total, assuming every Primogem earned from dailies also went toward the same goal.
Example 5: Chasing two featured characters in one patch. Double-banner patches are common, and going for two separate featured units means running the pity math twice. Average case: 2 times 90.625 equals 181.25 wishes, about $358.88. Worst case: 2 times 180 equals 360 wishes, about $712.80. That worst-case number is functionally what it costs to guarantee both units with zero luck involved, which is why most veteran players pick one banner per patch rather than chasing both.
Genesis Crystals vs. Welkin Moon: the cheapest way to buy wishes
If the goal is minimizing cost per wish rather than maximizing speed, Genesis Crystal cash packs are the wrong tool. The math above already shows Welkin Moon running at roughly $0.27 per wish against $1.98 for the largest one-time pack, a gap that holds regardless of how much a player is spending in a given month. The catch is time: Welkin Moon pays out over 30 days and requires a daily login, so it can’t compress a guarantee into a single afternoon the way a cash pack can.
| Purchase option | Price | Primogems delivered | Effective $/wish |
|---|---|---|---|
| Genesis Crystals, smallest pack | $0.99 | 60 | ~$2.64 |
| Genesis Crystals, largest pack | $99.99 | 8,080 | ~$1.98 |
| Blessing of the Welkin Moon (30 days) | $4.99 | 3,000 | ~$0.27 |
Most experienced spenders stack Welkin Moon as their baseline income and use Genesis Crystal packs only to top off the last few wishes before a pity threshold, which keeps the blended per-wish rate well under the $1.98 headline figure.
The weapon banner runs a different system entirely
Everything above describes Character Event banners, where characters are pulled. The weapon banner works on a separate mechanic called the Epitomized Path, and it doesn’t use a 50/50 at all. Instead, players pre-select one of the two featured 5-star weapons as their “fated” pick before pulling. Every time a 5-star weapon drops that isn’t the fated choice, a counter fills. Once that counter hits 2, the next 5-star weapon pull is guaranteed to be the fated weapon. Hard pity on the weapon banner also lands earlier than the character banner, at wish 80 rather than wish 90, based on the same community wiki disclosure used for the character-side numbers above. Because there’s no coin flip involved, players who want a specific weapon face a more predictable, if still not cheap, path to it.
Does pity carry over between banners?
Yes, with a specific scope. Pull count and the guarantee flag (whether your next 5-star is locked to featured) both carry over between Character Event banners that share the same pity pool, which HoYoverse has run as a shared system across the main Character Event banner and its secondary rotating slot for several years now. What doesn’t carry over is anything from the Standard (permanent) banner, which tracks pity entirely separately. Switching from a Character Event banner to the Standard banner mid-pity resets nothing on the Standard side, but it also doesn’t burn accumulated Character Event pity, since that count simply waits until a player returns to it.
Is this actually “provably fair”?
Worth separating two different meanings of “fair” here. In the cryptographic sense used by crypto casinos, provably fair means a player can independently verify, using a published seed and hash, that a specific outcome wasn’t tampered with after the fact. Genshin Impact’s wish system is not that. There’s no client-side seed verification, no published algorithm, and no way for a player to audit an individual pull. What HoYoverse does instead is disclose a base rate and a hard pity ceiling in the game’s own probability panel, then let players reverse-engineer the soft-pity curve through years of aggregated pull logging.
That’s a meaningfully different, weaker standard than cryptographic provable fairness, but it is stronger than a fully opaque loot box with no disclosed rates at all. The 0.6% and wish-90 numbers are real, checkable against the in-game disclosure, and consistent across years of tracked pulls. Whether that counts as “fair” is a judgment call, not a math problem, and it’s a different claim than a blockchain-verifiable RNG.
Myths vs. facts
- Myth: Losing 50/50 twice in a row means you’re “due” for a win next time. Fact: each individual roll is an independent 50% event. The gambler’s fallacy doesn’t apply, though Capturing Radiance does provide a real, non-superstitious escalating boost after tracked consecutive losses.
- Myth: Soft pity at wish 74 guarantees a 5-star. Fact: wish 74 is where the odds start climbing sharply, not where they hit 100%. The actual guarantee doesn’t land until wish 90.
- Myth: Capturing Radiance means the 50/50 is never lost anymore. Fact: it caps the run at three consecutive losses before forcing a conversion, but the first roll is still a flat 50/50. It reduces the tail risk without eliminating the coin flip.
- Myth: The biggest Genesis Crystal pack is always the best value. Fact: Blessing of the Welkin Moon runs at roughly a quarter of the per-wish cost of any cash pack, provided a player can wait out the 30-day payout window.
- Myth: Genshin’s gacha system is cryptographically provably fair like a crypto casino. Fact: it’s a disclosed-rate system, not a verifiable one. Players trust HoYoverse’s published floor and ceiling rather than auditing individual pulls.
For a sense of how these mechanics compare across the genre, Honkai Star Rail runs its own 90-pull cap with a similarly disclosed 0.6% base rate, while NIKKE’s system layers in a 200-pull mileage cap on top of a roughly 2% base rate, a structurally different approach to guaranteeing value. Readers who want to run these numbers against their own pull history can follow a separate walkthrough on building a pity and expected-value calculator using the same rates cited here. Broader background on how these disclosed-rate systems compare to true verifiable RNG lives on the provably fair hub.
Frequently asked questions
What are the exact odds of winning the 50/50 in Genshin Impact?
50% on the first 5-star pull of a Character Event banner with no active guarantee. Since Capturing Radiance launched in version 5.0, the effective long-run rate rises to roughly 55% once escalating conversion chances on repeated losses are factored in.
How many wishes are guaranteed to get any 5-star?
90 wishes, full stop, on Character Event and Standard banners. That’s hard pity. Most players get there sooner, since soft pity sharply raises the odds starting around wish 74.
Is wish 74 or wish 90 the “real” pity number?
Both matter, for different reasons. Wish 74 is when soft pity starts meaningfully raising the odds above the 0.6% floor. Wish 90 is the hard guarantee where a 5-star is mathematically certain. Neither number replaces the other.
How much does it cost in USD to guarantee a featured character?
Averaged across outcomes with Capturing Radiance factored in, about $179 at current Genesis Crystal pricing. Worst case, if the 50/50 is lost and Capturing Radiance doesn’t convert, the ceiling is $356.40 for two full pity cycles.
Does pity or the guarantee carry over between banners?
Yes, between Character Event banners sharing the same pity pool. The Standard (permanent) banner tracks its own separate pity and doesn’t interact with Character Event pity in either direction.
What’s the cheapest way to buy Primogems for wishing?
Blessing of the Welkin Moon at $4.99 for 30 days, delivering 3,000 Primogems total for an effective rate near $0.27 per wish, well below any one-time Genesis Crystal cash pack.
Is Genshin Impact’s gacha system provably fair in the technical sense?
No, not in the cryptographic, player-verifiable sense used by crypto casinos. It’s a disclosed-rate system: HoYoverse publishes a base rate and hard-pity ceiling, but individual pulls can’t be independently audited the way a seed-based provably fair system allows.
Does Capturing Radiance apply to weapon banners too?
Capturing Radiance was introduced specifically for Character Event banners. The weapon banner uses a separate mechanic, the Epitomized Path, which lets players choose a specific featured weapon to fate-lock pity toward, and it isn’t the focus of the conversion-chance system described here.
