Using the screener: two worked examples that end in a question

The screener is a filter over the same metric table that feeds every asset file, with three controls: a universe, a set of conditions that must all hold, and a sort. It does not score assets and it does not rank them by quality. What it produces is a shortlist whose members share a measurable property, and the useful output of a screen is almost always a question about why a particular asset appears where it does.

What the three controls actually do

The universe is chosen first, usually by sector from the sector map or by a coverage threshold. Conditions are then applied with an implicit AND: an asset survives only if every condition holds. Sorting reorders the survivors on one column and changes nothing else.

One behavior is worth internalizing before running anything. A condition on a metric silently removes every asset for which that metric is empty. Filtering on 30-day fees excludes assets with no fee feed, and those assets are not necessarily assets with low fees — they may be assets the fee feed does not cover. Every screen therefore reports both the number of matches and the number of assets excluded for missing data, and the second number is the one that tells you how much of the market the screen actually looked at. Any result set can be sent to the comparison view, which places the same columns side by side with the definitions attached.

First screen: chains where market value is small relative to measured fees

Set the universe to smart contract platforms. Add a condition requiring 30-day fees to be present, add market capitalization, market capitalization to fees, 7-day fees and 30-day fee growth as columns, and sort ascending on the ratio. The head of that list holds chains whose market value is a smaller multiple of trailing fees than their peers.

Before reading anything into the ordering, note what the ratio is not. It resembles a price-to-earnings multiple and shares none of its content. Fees are gross amounts paid by users to transact; they are not profit, they are not audited, no cost of operating the network has been deducted, and holders of the asset generally have no legal claim on them. Where any claim exists it comes from a mechanism such as a token burn or a fee switch, both of which are design choices that governance can change. Two chains with the same ratio can distribute fee income to completely different parties, which is the distinction between protocol revenue and supply-side revenue.

Then look at the composition of the fee total. Multiply the 7-day figure by roughly four and set it beside the 30-day figure. When the 7-day figure implies a far larger month than the one recorded, activity has accelerated recently; when it implies a much smaller one, the trailing month contains a burst that has since passed. Fee spikes come from identifiable causes — a token launch, a congested market, an airdrop claim window, a liquidation cascade — and a chain whose trailing fees are dominated by one such week is being measured on an event rather than on a baseline. Fees denominated in a token whose price moved sharply produce the same distortion for a different reason, since the dollar total mixes activity with a price path.

The screen has not concluded anything. It has produced one question per row: how much of the trailing fee total came from a single week, and what was happening on that chain during it. Two places on this site answer that directly — the fee chart on the asset's own file, which shows the daily series rather than the total, and the incident log, which records outages, exploits and congestion events with dates. The calendar covers scheduled events such as upgrades that can move fees for a week and then stop.

Second screen: staking yields that differ for structural reasons

Set a condition requiring staking ratio to be present, which restricts the universe to networks using proof of stake. Add nominal staking yield, real staking yield, real inflation rate, validator count and Nakamoto coefficient, then sort descending on the real figure. The spread across the resulting list is usually wide, and almost none of it is explained by anything a reader would recognize from a savings product.

The nominal figure is the rate at which newly issued units, plus any share of fees directed to validators, accrue to staked units. It is not a contractual rate. It moves with participation, because in most designs the same issuance is divided among a larger staked base as the staking ratio rises, so a network with high participation mechanically shows a lower per-unit rate. It can be reduced to nothing or made negative by slashing, the protocol penalty for validator misbehavior such as double-signing. It is earned in the network's own token, so its dollar value depends entirely on that token's price. And it is not liquid: most networks impose an unbonding period during which units are neither staked nor transferable, which is the gap that liquid staking tokens exist to fill, at the cost of adding a smart contract and an issuer between the holder and the stake.

The real figure subtracts supply growth. It answers a narrower question than the nominal one: whether a staked unit's share of total supply grows, shrinks or holds. A network paying a large nominal rate out of large issuance can show a real figure near zero, because the reward and the dilution cancel. That is the single most useful thing this screen shows, and it is the reason both columns are published rather than one.

Again the output is a question rather than a finding: for each row with a wide gap between the two columns, whether the gap comes from the issuance schedule, from the participation level, or from a fee component that would persist even if issuance stopped. The staking section and the supply tab of each asset file carry the inputs needed to separate those three.

Three ways a screen misleads a careful reader

The universe contains assets that still exist. Assets that failed have left it, so any historical pattern found by screening today's universe is subject to survivorship bias — the 2022 failures of Terra, Celsius, Three Arrows Capital and FTX removed a substantial set of entities from every list of this kind, and their absence flatters whatever remains. Screening on a series that has since been corrected introduces look-ahead bias, because the screen uses values nobody could have seen at the time. And ratios built on a small denominator behave badly: a chain with negligible fees will show an extreme fee multiple that reflects arithmetic rather than the market's assessment, which is why the screener suppresses ratios below a coverage floor rather than displaying an unstable number.

A screen is most useful when it is treated as the first step. Send any result set to the comparison view, read the definition of each column in the metric catalog, and check the comparability rules in the next article of this track before drawing a line between two rows.