Alfa

A valuation screen for blockchains. It answers one question:

Which chain is cheap relative to how well it is actually performing?

Every major chain is scored twice — once on the size and growth of its real economy, once on what the market pays for it — and ranked by the distance between the two.

---

The model

Crypto metrics span six orders of magnitude, so nothing is compared on raw values. Every input becomes a percentile against the peer universe, which is robust to outliers and directly interpretable: 82 means "ahead of 82% of chains".

Three scores are built independently, then combined.

Fundamentals — how large the chain's real economy is.

Input	Weight
Chain fees, 30d	20%
Stablecoin float	18%
Total value locked	17%
Chain revenue, 30d	12%
DEX volume, 30d	12%
Real-world assets	11%
Bridged volume, 30d	6%
Protocols deployed	4%

Momentum — whether that economy is growing or decaying. 30-day fee growth (27%), stablecoin growth (22%), TVL growth (20%), DEX volume growth (16%), bridged volume growth (9%), real-world asset growth (6%).

Active addresses and transaction counts are deliberately absent. Both are trivially inflated by airdrop farming and bot traffic, and neither says much about whether a chain earns anything. They were in the model and were removed. Stablecoin float took most of the freed weight: it is the least gameable measure of whether anyone is actually using a chain, because it is money someone chose to settle there rather than incentivised deposits chasing a yield that leaves when the yield does. Both its level and its 30-day direction are scored.

Cheapness — the valuation ratios, inverted so high means cheap. Market cap over annualised fees (28%), over TVL (23%), over annualised revenue (18%), over stablecoin float (14%), over annualised DEX volume (10%), over real-world assets (7%). A chain needs at least two of these present or it is left unrated — one odd denominator is enough to put a chain wrongly at the top of a "cheapest" list. When that happens the interface says so rather than quietly renormalising the remaining weights.

The headline number

valueGap runs from −100 to +100. Positive means the market prices the chain below what its activity supports.

valueGap = 0.43 × (fundamental percentile − market-cap percentile)
         + 0.32 × (cheapness, recentred to ±100)
         + 0.25 × (momentum,  recentred to ±100)

Momentum earns a quarter of the weight deliberately. A cheap multiple on a shrinking fee stream is a value trap, not an opportunity — those chains are flagged separately and can be filtered out.

Two leaders, not one headline

The screen opens on the widest value gap in each layer rather than one winner overall. An L1 and a rollup are not competing for the same thing, and a single headline hid that: the widest gap anywhere is Monad on circulating supply and OP Mainnet on fully diluted, so the hero changed identity when the basis toggle moved, for a reason that had nothing to do with either chain. Split by division, both leaders hold on either basis and what changes is the margin — Monad is 28 points clear of the best rollup on circulating and 8 behind it on fully diluted.

A leader has to clear the same bar the old hero used: a real token, undervalued, not a value-trap shape, and confidence at 60% or better. Mezo posts a wider L1 gap than the runner-up on decaying activity, and Chainflip beats both leaders on 42% confidence; neither is a finding worth leading with.

Beside each figure is that division's field on the value scale — one tick per rated chain, the leader marked — because "56 points undervalued" says nothing about whether the winner is clear of the pack or a nose ahead. It runs vertically with undervalued downward, the same geometry as the alpha map and the chain page's scale, so the screen keeps one mental model for the quantity.

CoinGecko classifies 41 chains as L1 and 24 as L2, which leaves 20 that are neither — appchains, sidechains, validiums. They are not guessed into a division to tidy the layout: a footnote says how many sit outside, and names one when its gap would have beaten both leaders.

Confidence

confidence (0–1) blends how many model inputs the chain actually supplied with a size gate, because a $2M-TVL chain can post a spectacular multiple purely because its denominator is noise. The screen defaults to hiding anything below 35%.

Only metrics whose absence is a genuine data gap count toward coverage. Real-world assets, bridged volume and Mayan routing are all excluded: the first two come from complete global scans, so a chain missing from them holds a real zero, and the third reaches about a dozen chains by design. Counting any of them would dock a chain's confidence for a fact about the world rather than a fact about the data.

The peer trend line

Market cap is regressed on a log-scale index of economic size across the chains with complete data. The scatter plot draws that fit with a ±1σ band, and each chain reports its distance from the line in standard deviations.

The fit typically explains around half of what chains are worth, leaving a residual spread of several times. That wide band is the honest finding, not a defect — and it is why the distance from the line is reported as a σ figure rather than dressed up as a price target.

---

Where the data comes from

DefiLlama — capital, revenue, real-world assets and bridge flow

Free, unauthenticated. TVL and 90-day TVL history, stablecoin float, DEX volume, bridged volume, tokenised real-world assets, protocol counts, market caps and 30-day price history.

Two things worth knowing about how this adapter reads DefiLlama:

This is the beginning of the README. Read the rest at the original. Read the full README ↗