Back to the token screener

DeFi Sentinel Index

A rules-based basket built from the screener's own numbers. Three eligibility gates, two sleeves — one weighted on cash flow, one on how little recovery the price already assumes — and a per-name cap. Every input is measured; the only judgement is where you set the dial.

Snapshot 2026-09-26 — same daily feed as the screener. · prices to 2026-09-27

How these numbers are measured

Index performance · 100 since 2022

All three rebased to 100 on the same day, so what is plotted is growth of the same starting capital. Positions are reset to target every 90 days and left to drift in between — the weights listed below are targets, not what the book holds on any given day. The blend is 25% BTC / 25% ETH / 50% index on the same schedule. The vertical axis is logarithmic, so equal vertical distance is an equal percentage move. Lines are a 7-day trailing average, which is why the last point sits slightly off the exact figures in the legend; those are close-to-close and unsmoothed. The dashed marker is where the basket first held every member — to its left it is a subset, with weights renormalized pro-rata over whoever was trading.

  • Sentinel Index +3573%
  • 25/25/50 blend +1103%
  • BTC +247%

This is a backtest, not a track record. This is a backtest, not a track record. It applies TODAY'S weights to past daily returns, so the holdings were chosen already knowing what happened. A point-in-time reconstruction is not possible: both weighting inputs would have to be recomputed from the fundamentals as they stood on each past day, and only one snapshot per token is kept. Read it as the shape of the rule, not as a return anyone earned.

Weights

Bar length is the final holding; the split inside it is where that weight came from. A name that is almost entirely one colour is being held for one reason only.

  • From the cash-flow sleeve
  • From the torque sleeve
HYPE22.5%
AERO12.0%
JUP11.8%
CAKE11.7%
PENDLE11.3%
UNI10.0%
PUMP7.4%
SKY5.6%
LIGHTER4.3%
LDO3.5%

Out of the index: AAVE No claim on revenue · MORPHO No claim on revenue

Portfolio characteristics

Weighted averages across the holdings, each shown against the same names weighted by market cap. Same universe on purpose — the difference is then attributable to the weighting rule alone, not to the gates.

Gross yield
4.38%
cap-weighted 3.35%

Holders revenue over market cap, before dilution. Supply drag is reported separately rather than netted off, so the two are not conflated.

Priced-in recovery
65%
cap-weighted 78%

The share of last cycle's fee peak that the book's current prices already assume. Lower means more room.

Peak-recovery multiple
1.8x
cap-weighted 1.4x

What the book returns if every holding merely regains its own last-cycle fee peak — already net of re-rate and dilution.

Supply drag
1.088
cap-weighted 1.081

Weighted forward-12-month supply growth. 1.10 means a tenth of the return is consumed before the price moves at all.

The rules

Nothing here is discretionary. A token is in or out on three tests, and its size comes from two formulas over figures the screener already publishes. Change a control on the left and every table below recomputes.

Eligibility — all three must pass

It must pay something

Non-accruing tokens are excluded outright. No mechanism routes revenue to them, so there is no cash flow to weight on and no payout ratio to discount. This is a statement about the token, not about the protocol — several of the best businesses in the set fail it.

It must be exitable

30-day average dollar volume below the floor is out. A weight you cannot sell is not a weight, and the floor should rise with fund size — at $500M the names clearing $10M a day take weeks to unwind.

Thin data is sized down

Confidence tier C — more than one data caveat, or no emissions schedule to measure dilution against — takes half weight by default rather than exclusion. The uncertainty is real, but it is uncertainty, not a defect.

Weighting — two sleeves, blended

Sleeve A — cash flow

w ∝ √(holders revenue) × tier

What you own if the cycle never turns. The square root is the load-bearing part: on raw revenue a single name takes three quarters of the book, which is a concentration decision disguised as a weighting rule.

Sleeve B — cycle torque

w ∝ (1 / breakeven − 1) × tier

What you own if it does. Breakeven is the share of a protocol's own last-cycle fee peak that today's price already assumes, so its reciprocal is the return on a mere return to that peak — already net of the multiple re-rate and of forward dilution.

Cap, then blend

w = split × A + (1 − split) × B

The cap is applied inside each sleeve and the spill redistributed pro-rata, so a name capped in one sleeve can still be small in the other and the blend softens on its own. Splitting the capital rather than the score is what keeps both extremes: a single combined score multiplies the two axes, so anything strong on one and weak on the other lands mid-table.

Sleeve split
70 / 30
Liquidity floor
$10M
Per-name cap
30%
Tier-C names
Half weight
Rebalance
Every 90 days

What this does not tell you

The yield is gross. Dilution is carried in the drag column instead of being netted off, because a token can be cheap on one and expensive on the other, and averaging them hides which.

End-to-end capture is unreliable wherever the fee adapter under-reports the supply side — the ratio can approach 100% for reasons that have nothing to do with generosity. It does not corrupt the torque sleeve, which uses fee ratios, where the distortion cancels.

Any token whose payout runs against a stated cap is being weighted on that cap. Where the underlying revenue sits below the programme's own trigger, the true figure may be zero, and the position should be read as an option on the programme activating rather than as cash flow.

Market cap is the denominator of every valuation metric here, so a cap-weighted benchmark is structurally the opposite of what the sleeves select for. It is a reference point, not a target.

A quarterly rebalance matches the 90-day measurement window. Sleeve A moves with revenue and is slow; sleeve B moves with price and therefore trims winners on its own. A change of capture class is an event, not a scheduled item.

This is an illustration of index-construction method using public data. It is not investment advice, not a recommendation, and not a licensed advisory product.