What loot boxes and gacha actually cost players

The $2 price is the wrong number

Suppose a game sells a mystery pull for $2 and the thing you actually want has a 1% chance of appearing.

The store can truthfully say the pull costs $2.

That does not mean the item costs $2.

If every pull is independent and there is no pity system, a 1% target takes 100 pulls on average to appear for the first time. At $2 per attempt, the expected acquisition cost is $200.

Even that number is easy to misunderstand. Spend $200 on 100 pulls and you are not guaranteed the item. The chance of missing a 1% target 100 times in a row is about 36.6%.

The random purchase has two prices:

• the fixed price of one attempt;
• the uncertain cost of reaching the outcome you care about.

Loot boxes, gacha banners, player packs, card packs, prize wheels and mystery chests dress that structure differently. The ESRB groups all of them under randomized in-game purchases when real money or purchasable virtual currency can lead to an unknown reward.

The economics are the same: you are buying a probability distribution, not a known item.

A 1% chance gets expensive faster than it looks

For an independent drop with probability p, the chance of getting at least one success after n attempts is:

1 − (1 − p)^n

For a 1% target:

Attempts	Chance of getting at least one	Chance of still missing	Cost at $2 per attempt
10	9.6%	90.4%	$20
50	39.5%	60.5%	$100
69	50.0%	50.0%	$138
100	63.4%	36.6%	$200
230	90.1%	9.9%	$460
299	95.0%	5.0%	$598
459	99.0%	1.0%	$918

This is the part a rarity label does not communicate intuitively.

“1%” sounds small, but a player often thinks about it as a sequence: surely I will get it after enough tries. The formula agrees that the chance rises with more independent attempts. It also shows how much money “enough” can mean.

The average wait for a 1% event is 100 attempts. The 50% point arrives earlier, at 69. The 95% point arrives much later, at 299.

Expected cost is therefore not a promise and not a safety ceiling. It is a long-run average.

That distinction matters whenever someone says a rare item “should cost around $200.” If the system is independent, one player may get it on pull three. Another can spend $600 and still miss.

The displayed rarity rate may not be the rate for your target

A second trap appears when the game publishes an honest probability — but not for the exact outcome the player is chasing.

EA’s current Football Ultimate Team probability explanation says store packs display the minimum probability of receiving one or more items within a rating range or category. EA also makes a separate point that matters much more to a player hunting one card: the probability of a particular individual item can be significantly lower than the displayed category probability and is often below 1%.

A label such as “84+ player” answers:

What are my chances of getting something from this category?

It does not necessarily answer:

What are my chances of getting this exact player?

Those are different products from the player’s perspective.

EA also says its pack probabilities are recalculated as live content changes and that pack openings are independent unless a particular offer says otherwise. Opening a string of bad packs does not make the next independent pack “due.”

This is where the gambler’s-fallacy version of loot-box math falls apart. If a 1% target is truly independent on every pull, missing 99 times does not make pull 100 a 100% event. It is still 1%.

A pity system is different because the rules themselves change or guarantee the outcome after a threshold.

Pity systems replace an infinite tail with a cap

Gacha games often soften low base odds with a guarantee.

HoYoverse’s current Genshin Impact Wish guarantee documentation says a Character Event Wish guarantees a 5-star character within 90 wishes. When that 5-star arrives, there is a 50% chance it is the featured promotional character. If it is not, the next 5-star obtained from that banner type is guaranteed to be the featured character.

That creates three different numbers a player should care about:

1. chance per attempt;
2. guarantee for any high-rarity result;
3. guarantee for the specific featured result.

From zero pity, the published rules imply that the featured character can require up to 180 wishes: 90 to reach a first 5-star, then another 90 if the first 5-star loses the 50/50 and the second cycle reaches its maximum.

That does not mean every featured character costs 180 paid wishes. Players can pull a 5-star early, enter with pity already accumulated, carry guarantee state between compatible banners and use currency earned through play.

It does mean the mechanic has a definable worst-case number of attempts under those rules.

That is more useful than saying “the drop rate is low.”

A pity system can make a randomized purchase easier to budget because the tail is capped. It can also make quitting feel harder when the interface tells you that you are already 60 or 70 attempts into the guarantee.

Mathematically, previous failures have value in a pity system if they move the counter. Economically, money already spent is still gone.

Those two facts can coexist.

Virtual currency makes the pull cheaper to feel than to price

Most randomized purchases do not put a credit-card charge beside every spin.

The path is more likely to look like:

real money → currency bundle → premium currency → ticket or points → randomized pull

That structure is convenient for games. It is terrible for mental arithmetic.

If $9.99 buys an awkward number of gems, a ten-pull costs another number of gems, a bundle includes a bonus and leftover currency remains afterward, the player has to perform several conversions to answer a basic question:

What did this attempt cost me in dollars?

The issue became explicit in the United States in 2025. In announcing its proposed settlement with HoYoverse, the Federal Trade Commission alleged that Genshin Impact’s multi-tier virtual-currency system obscured how much players were spending and how much they might need to spend to obtain rare prizes. HoYoverse agreed to a proposed $20 million settlement and other changes; the FTC announcement described an order that would require, among other things, clearer odds and exchange-rate disclosures and a direct real-money purchase option for loot boxes in the US.

The useful lesson is broader than one case.

Before doing probability math, convert the mechanic back into cash per attempt.

If a bundle gives 6,600 units for $99 and a pull costs 160 units, the pull does not “cost 160.” It costs roughly the real-money value of 160 units under the bundle you actually buy, adjusted for any bonus or free currency you are counting.

Do that conversion first. Otherwise the probability calculation is precise arithmetic on top of a fuzzy price.

“Expected value” is not always the useful metric

Traditional gambling analysis often asks about expected value: multiply each possible payout by its probability, add the results and compare that expected return with the stake.

Loot boxes complicate that because many rewards do not have a clean cash value.

A cosmetic skin that cannot be sold for money might be worth $40 to one player and nothing to another. A powerful character may save time or improve performance without having a legitimate resale price. Duplicates may convert into another in-game currency. A tradeable item may sell only into a platform wallet rather than a bank account.

For many game purchases, a fake precision such as “this box has an EV of $1.73” is less useful than three simpler numbers:

• expected attempts to the target;
• probability of getting the target within your budget;
• worst-case attempts if a guarantee exists.

If an item can be sold through an official marketplace, the resale side still needs careful treatment. Steam’s Community Market, for example, settles transactions in Steam Wallet funds rather than ordinary withdrawable cash; our guide to the Steam Community Market covers that boundary.

Price the outcome according to what you can actually do with it, not what a rarity colour implies.

Odds disclosure helps — but it does not do the math for you

The industry is much more transparent about randomized purchases than it was a decade ago.

Apple’s App Review Guidelines require apps offering loot boxes or other paid randomized virtual-item mechanics to disclose the odds of receiving each type of item before purchase.

The ESRB’s “In-Game Purchases (Includes Random Items)” label covers games with paid randomized rewards.

In the UK, the Advertising Standards Authority issued a 2026 enforcement notice requiring covered app-store listings to make the existence of purchasable loot boxes clear; active monitoring began on 26 May 2026.

Those are useful protections. They still leave work for the buyer.

A disclosed “5% Epic” rate does not automatically tell you:

• the odds of one specific Epic;
• whether rewards are sampled independently;
• whether duplicates are possible;
• whether the pool changes during the event;
• how pity works;
• whether the guarantee carries between banners;
• the cash cost of one pull after currency conversion;
• the chance of success within the amount you are actually willing to spend.

Probability disclosure is an input.

It is not a price tag for the target.

The spending is highly uneven

Most players are not spending thousands of dollars a month on loot boxes.

The risk sits in the distribution.

A 2021 secondary analysis pooled open survey datasets covering 7,767 loot-box purchasers. The top 5% of reported spenders — those spending more than $100 a month — accounted for half of reported loot-box expenditure. The same analysis found loot-box spend was associated with problem-gambling scores but not with self-reported earnings.

That result should not be inflated into “half of all industry revenue comes from problem gamblers.” The study used self-reported survey data, not transaction records from every game company.

It does show why averages are weak descriptions of player cost.

If nine people spend very little and one person spends heavily, “average spend” can make both groups look more normal than they are.

The broader research signal is also consistent but needs the same restraint. A systematic review and meta-synthesis found a positive association between loot-box engagement and problem gambling across most included studies, with a moderate weighted association. The authors also stressed that the evidence was overwhelmingly cross-sectional, which means it cannot cleanly establish whether loot boxes cause gambling problems, people with gambling problems are more attracted to loot boxes, or other factors contribute to both.

The UK government’s loot-box review reached a similar position: the association is well established enough to take seriously, while causal direction remains unresolved.

That is a much stronger claim than pretending every random item purchase is equivalent to casino gambling.

How to calculate the cost before you pull

You do not need a statistics degree.

Write down six things.

1. Real-money cost per attempt

Convert the premium currency back into the cash price you actually pay.

If bundles have different exchange rates, use the bundle you would realistically buy.

2. Probability of the exact target

Do not substitute “chance of a Legendary” for “chance of the Legendary I want.”

If the game discloses only a rarity bucket, you may not have enough information to calculate the target cost precisely.

3. Whether attempts are independent

If each pull is independent, previous misses do not improve the next base chance.

If there is pity, soft pity, a guarantee or a changing pool, use those rules instead of the simple independent formula.

4. Your probability at your budget

For an independent target:

chance of success = 1 − (1 − p)^n

where p is the per-pull target probability and n is the number of attempts your budget buys.

This is usually the most useful number on the page.

Not “How much does one pull cost?”

“If I spend my whole budget, what are my odds of still not getting it?”

5. The hard ceiling, if one exists

If the game has a published pity or guarantee, calculate the maximum attempts from your current state.

Do not assume a guarantee transfers between banners unless the rules say it does.

6. The stop number before the first purchase

Random rewards make the next attempt feel unusually important because it might resolve every previous miss.

That is exactly why the spending limit should be chosen before the first pull rather than after pull 47.

A limit made in advance is a budget.

A limit repeatedly renegotiated after losses is a chase.

The real cost is a distribution, not a sticker price

A direct-purchase cosmetic has a simple price: $20 buys the cosmetic.

A randomized cosmetic can display “$2 per pull” while the cost of the actual target stretches across a probability curve.

At 1% independent odds, the expected first hit is 100 pulls. A 95% chance requires 299. A 99% chance requires 459. Pity can cap the tail. Category-level disclosures can make the visible odds look better than the odds of one exact item. Virtual currencies can make the real-money cost harder to feel.

None of this means every loot box is fraudulent, every gacha player is being harmed or every randomized reward is legally gambling.

It means the small number beside the button is only the price of another chance.

If you want to know what the mechanic costs, price the thing you are actually trying to obtain.