Comparing assets without comparing apples to oranges
What has to match before two assets can be lined up: the layer, the supply denominator, the fee definition, the unit of user.
Two assets can be compared on a metric when the metric is constructed the same way for both. That sounds trivial and is the hardest part of digital-asset research, because the same word is used for measurements taken at different layers of a stack, with different denominators, over different definitions of a user and different definitions of revenue. The comparison view shows the definition next to each column for exactly this reason.
The layer has to match
A layer 1 settles its own transactions and prices its own block space. A layer 2 executes transactions elsewhere and posts data or proofs back to a layer 1, paying that layer for the privilege. When a rollup collects a fee from a user, part of that fee is a cost it will pay onward for data availability — on Ethereum, through blob space introduced for that purpose. Adding a rollup's fees to its settlement layer's fees counts the same economic activity twice, and the two figures also answer different questions: the rollup's fee is a price charged to users by a sequencer, while the layer 1's fee is a price charged for permanent data.
The same problem appears one level up. An application built on a chain and the chain itself both generate fees, and they are not additive. A decentralized exchange collects a trading fee that mostly goes to liquidity providers, while the chain collects gas for including the trade. Both are real; neither is a component of the other.
The denominator has to match
Any ratio with market value on top inherits a choice about supply. Market capitalization uses circulating supply; fully diluted valuation uses the eventual total. Comparing one asset on its circulating figure with another on its diluted figure produces a difference that is entirely an artifact of the choice. The ratio between the two makes the size of that choice visible: an asset with nearly all supply issued shows a value close to one, while an asset with years of vesting ahead shows a much larger one. The issued share of supply is the cleanest single column for putting two assets on comparable footing before any ratio is read.
The word revenue does at least four jobs
Fee figures are the most frequently mismatched numbers in this field, because four distinct quantities travel under similar names. Gross protocol fees are everything users paid. Supply-side revenue is the portion passed to the parties providing the service — liquidity providers, validators, miners. Protocol revenue is the portion retained by the protocol itself, often accruing to a treasury. Holder revenue is the smaller portion that reaches token holders, through a burn or a distribution, and for many assets it is zero. Take rate expresses the retained share as a fraction of the gross, which is the column that shows how much of the headline fee number is captured rather than passed through.
Because of this, two assets showing similar 30-day fees may deliver very different amounts to holders, and a comparison built on gross fees says nothing about that difference. The site publishes holder revenue yield separately for that reason, and states where it is undefined rather than assuming zero.
A user is not an address
Activity counts are the second common trap. An active address is an address that transacted. Chains built on an account model reuse a single address indefinitely, while chains built on UTXO accounting generate a fresh address for change on nearly every transaction, so the same behavior produces different address counts by design. Exchange addresses act on behalf of enormous numbers of customers, contract addresses transact without any human involved, and address clustering that attempts to correct for this is heuristic rather than exact. Comparing active address counts across two chains of different design measures the designs as much as the usage; comparing a chain against its own history is far more defensible.
Transaction counts carry a related problem. Networks differ in what a transaction is: some batch many transfers into one, some charge nothing for failed attempts and some record them, and low-cost block space invites automated traffic that would not exist at a higher price. Average transaction fee alongside the count usually explains more than either alone.
| Comparison | Verdict | What has to match first |
|---|---|---|
| Two assets on 30-day return | Sound | Same window and same quote currency |
| Two proof-of-stake chains on staking ratio | With care | Whether liquid staking and custodial stake are counted |
| Two chains on active addresses | With care | Account model versus UTXO accounting |
| Two assets on market cap to fees | With care | Fee definition and supply denominator |
| A rollup and its settlement layer on fees | Unsound | Rollup fees include costs paid onward to the layer 1 |
| A lending protocol and a decentralized exchange on total value locked | Unsound | Deposits and pool inventory are different quantities |
Total value locked is a count of deposits, and it double counts by construction
Total value locked is the dollar value of assets sitting in contracts. It is not capital raised, not assets under management and not revenue. It moves when prices move even if no user acts. It counts the same underlying units more than once whenever a deposit receipt is itself deposited elsewhere — a liquid staking token used as collateral in a lending protocol appears in both places — and a wrapped token counted on a destination chain may also be counted where the original is held. The site states its double-count rules on the methodology page, and figures built on it, including market cap to total value locked, inherit whatever those rules decide.
Correlation columns compare behavior rather than construction
Where two assets resist direct comparison on fundamentals, their price behavior can still be lined up. Correlation to bitcoin and beta to bitcoin are computed over a stated window on the same return frequency for every asset, so they are among the few columns that are comparable across otherwise unlike assets. They describe how returns have moved together in the past and carry no implication about the future; a correlation computed through a period containing a single market-wide shock will describe that shock more than anything structural.
Before lining up any two rows, open both in the comparison view and read the definition shown with each column; where a definition involves a judgment call, it is written out on the methodology page. The next article explains the figures this site declines to publish, several of which are declined precisely because they cannot be made comparable.