Methodology

Rip sites publish a stream of what just came out of their packs. We read that stream every three minutes and write each opening into our own database: which pack, what it cost, which card, what the card was worth at that moment. Each opening is stored under the identifier the operator gave it, so re-reading the same window adds nothing twice. Free promotional openings and anything with no price attached are excluded, because neither tells you what a paying customer gets.

Three kinds of facts, kept apart and labelled everywhere they appear. Sampled means we watched it happen and counted it ourselves. Published means the operator publishes the figure and we read it, such as the odds printed on a pack page. Stated means someone said so, typically the payout terms in their help pages, and we quote it as their claim rather than our finding. A number never quietly upgrades from stated to sampled.

Below ten openings a pack carries no ranking at all: it is greyed out and marked as too thin to judge. From ten it enters the boards, but under 100 it stays flagged as a small sample, because a handful of openings tells you about those openings and not about the pack. An operator gets an overall rating as soon as we measure it at all, so a freshly connected site is not left blank for weeks; but until it reaches 500 sampled openings that grade is marked provisional, because the pull-based criterion can still move. Right now 21 of 36 listed sites carry a grade, 20 of them firm rather than provisional. Sites we only read a catalogue for carry no grade rather than a guessed one.

Boards sort on the typical rip: the median return per dollar spent. The reason is arithmetic. A single card worth a hundred times the pack price drags an average upward for thousands of openings afterwards, and someone reading that average would conclude the pack pays far better than it does for almost everyone. A median cannot be moved that way. The average sits next to it on purpose: when the average towers over the median, the pack is carried by rare hits, and that gap is itself the useful information.

Green, amber and red rank a figure against what this market actually returns, not against getting your money back. No site can pay out a median of 100%: that would mean the operator runs at a loss, and it has never happened in anything we have measured. Judged against 100%, nine of every ten packs we measure would be red, which tells a reader nothing about the difference between 88% and 45%. So on comparative figures — typical rip, gross return, take-home — green starts at 90%, amber runs 80 to 89, and red is everything below. Most packs are still red, because most packs do return less than they cost. One place keeps the strict scale: a single opening, where green means that card really was worth more than the pack, because there the question is about your money and the answer is checkable.

The card value a pack returns is not the money you end up with. Most sites offer to buy the card back instantly at a fraction of its value, and that fraction is what decides your outcome unless you ship every card home. So we show both figures on the same line: the gross return, and what remains at the operator's own published buyback rate. Where a site publishes no rate, that column stays empty. An estimated payout rate would be the single easiest number on this site to get wrong.

There is no overall score that merges returns with terms, and there never will be. Cards shipped to your door, credit inside a closed system, and crypto in your own wallet are three different things, and a figure that averaged them would be arithmetic dressed as insight. The same applies within returns: a site paying out in site credit is not compared against one paying cash as though the units matched. Terms are shown side by side and left for you to weigh.

Each number states when we last checked it, and sampled figures also state how many openings they rest on. When an operator changes their prices or their terms, our figure is old until the next run, and the date is there so you can see exactly how old. If a source stops responding, the figures stay where they were, still stamped with the date they were true, rather than silently drifting into the present tense.

Not because we describe ourselves as neutral, but because of how the site is built. Every operator runs through one pipeline: the same collection, the same thresholds, the same formulas, and no editorial adjustment at the end. There is no field anywhere in this system for nudging a rating. This page describes the method in full, so any figure can be recomputed from the operator's own public data, and we would rather be corrected than be wrong. Affiliate links are marked where they appear and are not an input to any calculation. Last change to this page: 19 August 2026.