Methodology

Show your working, or don’t publish the number.

Every figure on this site should be reproducible by someone who distrusts us. This page defines the math, names where each input comes from, states what we do when a platform won’t publish its odds, and discloses exactly how we get paid — including the part most affiliate sites leave out, which is what our commissions are actually worth.

01How we compute EV

A pack is a probability distribution over outcomes. For a pack costing price, where each outcome i has probability p_i and value v_i:

Gross EV%

Σ(p_i × v_i) ÷ price — the average value returned per dollar spent, before any cost of getting that value out.

Net EV%

The same sum computed on realizablevalues: what you could actually sell each outcome for, after the platform’s sell fee — or the buyback rate, where that is the exit you would really use. Net EV is the number our rankings default to, because it is the one that describes your outcome rather than the platform’s shop-window price.

P(profit)

The probability that a single pack returns more than it cost. This is the number EV alone hides: a pack can average 90¢ on the dollar and still lose money more than nine times in ten, because one rare outcome carries the whole average.

P10 / P50 / P90

The 10th, 50th, and 90th percentile outcomes. P50 — the median — is usually a far more honest answer to “what will happen to me?” than the mean.

Where a platform offers instant buyback, we record the buyback rate and compute a second net EV using it as the exit, kept separate from the market-sale figure rather than collapsed into one ambiguous number. The gap between those two is usually where the house edge actually lives.

Four conventions, stated so you can check our arithmetic

1. EV is value per dollar, so 1.00 is break-even.A pack at 0.87 returns 87¢ for every dollar you put in. Some tools instead quote EV as a surplus over price, where the same pack reads −13%. Identical information, different zero point — worth knowing before comparing our figures to anyone else’s.
2. We never renormalize published odds. Platforms round their percentages, so they rarely sum to exactly 100%. We compute on the odds as published and reject anything that misses by more than half a percentage point, rather than quietly scaling the numbers into a tidier total the platform never stated.
3. Percentiles are real outcomes, not interpolations. A pack cannot pay out a value between two of its own outcomes, so we never invent one. P50 is always a result the pack can actually produce.
4. Breaking even is not profit. P(profit) counts only outcomes worth strictly more than the pack price after fees. A pull that returns exactly what you paid is not counted as a win.

02Where card values come from

An EV calculation is only as good as the values fed into it, so each outcome is valued at what you could realistically get for it, in this order of preference: the platform’s own recent sales or floor price where that data exists; otherwise market comparables from established secondary markets; in both cases minus the platform’s sell fee.

When a value is genuinely uncertain, we use the conservative bound and flag it rather than picking a flattering midpoint. The value source for every platform is documented per-platform, so a disagreement with our numbers can be traced to a specific input rather than argued in the abstract.

03Published odds vs. estimated odds

Some platforms publish their pull rates. Some don’t. We never mix the two, and we never present an estimate as though it were a disclosure.

• ✓Published oddsare taken from the platform’s own odds page or documentation, with the source linked and the date recorded. Where the mechanism can be verified on-chain, we say so.
• !Estimated odds carry a visible estimated badge, and the assumptions behind the estimate are stated on the page. An EV built on estimated odds is an estimated EV, and it is labeled that way everywhere it appears — including in our rankings.

Whether a platform publishes verifiable odds is itself the thing we grade in the Odds-Transparency Index. A platform is never penalised in an EV ranking for our own uncertainty — it is penalised in the transparency grade for causing it.

04Timestamps and staleness

Prices, odds, and card values in this niche move weekly and sometimes overnight. Every EV figure on this site carries the timestamp of the data it was computed from. When a snapshot goes stale, the page says so with a visible banner rather than quietly serving an old number as though it were current. Snapshots are committed to version control as they are produced, so the history of what we published, and when, is auditable rather than assertable.

05How we get paid

We earn referral commissions from platforms we cover. Affiliate links are marked as sponsored, and the commission never depends on what we conclude — but you should not have to take that on trust, so here are the actual terms.

Every one of these programs pays us in platform points or credits, not cash. Those points buy packs. Packs, as this entire site exists to document, return less than they cost on average. So the face value of our commissions overstates what we actually earn — and quoting that face value at you would be exactly the sleight of hand we criticise platforms for.

Terms verified from our own accounts on 2026-08-03. Conversion rates are quoted in pack face value — the sticker price of the packs these points buy, not what those packs return.

What our points are actually worth

Take the clearest case. Collector Crypt redeems 100,000 points for one $50 pack, so a point has a face value of $0.0005. But the thing that point buys is a pack, and a pack does not return its own price. The honest valuation of a points balance is therefore:

points → packs those points buy → measured net EV of those packs → minus the fee to sell what comes out

We are not publishing that figure yet, because we do not have it. It requires a measured net EV for the specific packs involved, which our engine produces and our data pipeline has not yet filled in. Publishing a plausible-sounding estimate here would violate the same rule we hold platforms to, so the number is absent until it is real. When it exists, it goes here.

One consequence worth stating plainly, because it cuts against our own interest: the platform that pays us the least is not the platform we rank the lowest, and the platform that pays us the most is not the one we rank the highest. Rankings are ordered by measured net EV and transparency grade. Where a platform pays us well and grades badly, the grade is what you will see at the top of the page.

06Corrections

We will get things wrong. When we do, the correction is made on the page itself with the date and what changed, not silently edited away. If a number here looks wrong to you, the underlying inputs and their sources are published precisely so you can check — and a specific, sourced challenge will always get a faster response than a general complaint.