Why we rank packs on the typical rip, not the average
One card can carry an average. It cannot carry a median. That difference decides which pack sits at the top of our board — and it changes the answer completely.
Every card-rip site publishes an expected value. It is almost always an average — add up what all the cards are worth, divide by how many packs it took. That is a real number and it is not a lie. It is also, for this particular product, close to useless for deciding what to open.
The reason is the shape of the thing. Rip packs are built around chase cards: most pulls are ordinary and a handful are enormous. When a distribution has a long tail like that, the average stops describing anything that actually happens to you. It describes a blend of the common outcome and the rare one, weighted so heavily by the rare one that it can land somewhere no single pull ever lands.
The same pack, two honest numbers
Take Tilt Rips's Pokemon Mini Pack. We have sampled 55,530 real reveals from it — not a simulation, not the operator's own estimate, but cards that actually came out of that pack on their public feed, each valued at the operator's own published price.
Both numbers are correct. They disagree by 25 percentage points, and the gap is entirely the tail doing its job. Here is where those 55,530 pulls actually landed, measured against what the card cost:
Card value as a percentage of the $10.00 it cost. 30.7% came back worth at least their cost.
The median sits where the middle pull sits. The average is dragged upward by the 390 pulls that came back at more than double their cost — including that top card, which on its own is worth roughly 13 packs. That is not a flaw in the pack. It is the pack. Chase cards are the entire appeal, and a product designed around a long tail will always have an average that outruns its typical result.
Why this decides the ranking
If you rank a board of packs by their average, you rank them by how extreme their tails are. The pack with the single biggest grail floats to the top even when the ordinary experience of opening it is worse than everything below it. We had exactly that: for a while our own board was led by a pack whose average was over 200% and whose typical rip returned barely half the price.
Ranking on the median fixes it. Half of real pulls came back above the number and half below, so a single spectacular card cannot move it. The pack that leads the board is the pack that most often treats you best — which is the question a ranking is supposed to answer.
The average sits beside the median in every table on the site, because the gap between them is itself information. A pack where they are close is steady. A pack where the average towers over the median is a lottery ticket with a card attached — worth knowing before you open it, not after.
When the median is higher than the average
It sounds backwards, and it is rare, but it happens — and when it does it tells you something specific. It means the distribution leans the other way: a floor that most pulls land on, with a tail of duds below it rather than grails above.
We have seen it in packs with a guaranteed minimum value, where a large share of pulls come back at exactly the guaranteed figure. The median parks on that guarantee. The average gets pulled below it by the pulls that missed. Same arithmetic, mirrored — and the same reason to trust the median: it describes what usually happens, whichever direction the tail runs.
How big a sample the median needs
A median is only as steady as the sample under it. We tested this rather than guessing: we took every pack with more than 800 sampled pulls, repeatedly drew smaller samples from each, and measured how far the smaller sample's median landed from the full one.
| Sample size | Typical error | Worst 10% of cases | Within ±10 points |
|---|---|---|---|
| 25 pulls | 4.6 pts | 16.0 pts | 79% |
| 50 pulls | 3.4 pts | 11.0 pts | 88% |
| 100 pulls | 2.3 pts | 8.0 pts | 95% |
| 200 pulls | 1.7 pts | 5.8 pts | 98% |
A hundred pulls is the knee of that curve: 95% of medians land within ten points of where they eventually settle, and doubling the sample buys well under a point more precision while ruling out a quarter of the board. So a hundred attributed pulls is where a pack becomes rankable. Below it we still show the number — greyed, marked provisional, and sorted beneath every pack that cleared the bar, so a thin sample can never outrank a measured one.
The measurement cuts both ways
Measuring real pulls is not only a way to catch packs that underdeliver. Courtyard's Sealed Pokémon Booster runs meaningfully hotter than its own published odds imply — a typical rip of 130.0% against a published expected value of 99.5%, across 2,136 sampled pulls. The operator's advertised figure understates it.
See the full measured record for this packCourtyard · measured from the live feed→We would never have found that by reading the odds table. It only shows up if you count what actually comes out — which is the whole reason we sample the feeds in the first place.
Every figure here is gross — the market value of the card that came out, against what it cost. What you keep depends on how you cash out, and that varies enormously between sites: instant buybacks take a cut, store credit can only be spent on-site, and shipping the card means selling it yourself. We publish that separately on every row rather than blending it in, because mixing a number we measured with a number we read off a terms page would hide which is which.
Common questions
What is the difference between the median and the average return on a card pack?+
The average is total card value divided by total spent, so a single rare high-value pull can lift it substantially. The median is the middle result — half of real pulls came back above it, half below. On jackpot-driven products like rip packs the average is usually much higher than the median, and the median is the better description of what typically happens when you open one.
Why does RipRanks rank packs by median instead of expected value?+
Expected value is an average, so ranking on it sorts packs by how extreme their rare outcomes are rather than by how they usually perform. We rank on the observed median from real sampled pulls, and show the average and the published expected value alongside it.
How many pulls does a pack need before you rank it?+
One hundred attributed pulls. We measured the trade-off by resampling packs with 800+ pulls: at 100 pulls, 95% of medians land within ten percentage points of where they eventually settle. Packs with between 10 and 99 pulls show a greyed provisional figure and sort below every ranked pack.
Can a pack's median return be higher than its average?+
Yes, though it is uncommon. It happens when a pack has a guaranteed minimum value that many pulls land on exactly, with a tail of weaker results below it — the mirror image of the usual jackpot shape. The average gets dragged below the median instead of above it.