How the DeFi Sentinel Index chose and sized its ten tokens for the 2026-10-23 reset: forward cash flow instead of trailing payouts, estimated fee switches, a stated fee-base view per name, a top-ten cut with a watchlist, and a 90-day calendar charged 40 bps per dollar traded.
Until this quarter, the DeFi Sentinel Index recomputed its weights every time the page loaded. The engine was the same each day, but its inputs moved every morning, so the "index" quietly became a slightly different portfolio daily. An index is supposed to be a commitment to one book until the next reset. From 2026-10-23 it is: the weights below were fixed on 2026-10-18, they are published as a record, and the page shows exactly that record until the next scheduled reset on 2027-01-21.
This piece explains how those ten weights were reached, what changed from the book the page showed before, and how the rebalance itself is costed.
| # | Token | Weight | Before | Change | Fee-base view | Forward holders $/yr |
|---|---|---|---|---|---|---|
| 1 | HYPE | 25.2% | 22.4% | +2.8 | 3.0× | $2.24B |
| 2 | PENDLE | 12.2% | 11.1% | +1.1 | 8.1× | $34M |
| 3 | AERO | 10.6% | 11.2% | −0.6 | 2.7× | $277M |
| 4 | UNI | 9.9% | 10.7% | −0.8 | 1.5× | $192M |
| 5 | PUMP | 8.3% | 7.6% | +0.7 | 1.0× | $235M |
| 6 | JUP | 7.9% | 11.1% | −3.2 | 1.5× | $51M |
| 7 | CAKE | 7.3% | 11.3% | −3.9 | 1.2× | $48M |
| 8 | LIGHTER | 7.1% | 4.8% | +2.3 | 1.7× | $53M |
| 9 | SKY | 6.0% | 6.5% | −0.4 | 1.4× | $100M |
| 10 | ENA | 5.5% | — | added | 2.5× | $54M (estimated) |
| — | LDO | — | 3.3% | removed |
"Before" is what the page showed until this reset: the live engine on the 2026-10-15 snapshot. The biggest moves are CAKE and JUP down and HYPE and LIGHTER up. Those come mostly from two changes, the views and the decision to stop halving low-confidence names, which the next sections cover.
The index has two sleeves, and the capital is split between them rather than blending two scores into one:
What changed is the input to sleeve A. It used to be measured trailing holders revenue, under a rule that "a weight may not rest on a modelled payout". The trouble is that trailing revenue describes the last quarter, while the index holds the next one. The new input is forward expected holders revenue:
Several of the most valuable protocols pay holders nothing yet. Before, they entered with zero cash-flow weight. Now, a switch with approved or concrete terms is priced from those terms and flagged:
| Token | Status | Estimate | Basis |
|---|---|---|---|
| ENA | Approved (Sep 2), not triggered | $54M at the view | 5% / 10% / 15% / 20% of gross revenue at $7.5B / $10B / $15B / $20B of USDe, 95% to buybacks |
| MORPHO | Off, no proposal | $15M | Peer capture: the mean of Uniswap (7.6%) and Aave (5.4%) end-to-end, on $230M of fees |
| ETHFI | Proposal, outcome unconfirmed | $16M | The Foundation's weekly buyback proposal ($1.33M a month) |
| AAVE | Suspended since 2026-04-19 | $30M | A restart at the budget authorised in March |
| LDO | One-off live, NEST gated | $12M | The one-off stETH programme; NEST pays $0 below $40M of revenue or $3,000 ETH |
Only ENA made the final ten on an estimated switch. It carries 5.5%, and its payout starts only when USDe's 14-day average reaches $7.5B, against $4.71B on the snapshot date. That is the largest single forecast in the book, and it is disclosed as one.
Forward cash flow still needs an answer to "at what fee level?". Each name gets a fee-base view, a multiple of today's fee base that feeds both sleeves:
The views split into two groups:
A view is the one piece of judgement in the process. It is written down per name, so it can be disputed, and revisited each quarter.
Low-confidence names now count at full weight. The engine used to halve any name with more than one data caveat. Every input here comes from DefiLlama or public disclosures, so the haircut was dropped. That is what lifted LIGHTER and PUMP.
Fifteen tokens were weighted. The ten largest weights make the book, and the engine is re-run on those ten, so the caps still bind; nothing is topped up pro-rata. Cutting from fifteen to ten changed risk very little. Daily returns across the set correlate at about 0.47, and the fifteen-name version showed almost the same volatility and drawdown. What the cut did change is how much of the book rests on payouts that do not exist yet: it fell from 17% to 5.5%.
Names left out stay on a watchlist with an explicit trigger. Each re-enters at the next scheduled reset once the trigger fires, so the index buys a payout when it exists, not when it is forecast:
| Token | Re-entry trigger |
|---|---|
| MORPHO | A protocol fee switch passes governance |
| SYRUP | Maple's MIP-021 buyback is renewed at a higher payout tier |
| ETHFI | Weekly buybacks appear on-chain |
| LDO | NEST starts paying (revenue above $40M/yr and ETH above $3,000) |
| AAVE | The suspended buyback is restarted and executing |
When. Resets fall every 90 days from 2022-08-15: …, 2026-07-25, 2026-10-23, 2027-01-21, 2027-04-21. Ninety days matches the window every input is measured on, so rebalancing faster would be trading on noise. Positions reset to target at that day's close and then drift until the next reset. The listed weights are targets, not what the book holds on any later day.
What it costs. Every reset is charged 40 bps per dollar traded. That is 10 bps of venue fee (a CEX taker fee or a DEX swap fee) plus 30 bps of slippage, sized for a $50M book whose every position exits in under a day at 20% of average daily volume. The blend's BTC and ETH legs are charged 10 bps. The initial purchase is charged too.
| This reset | Typical backtest reset | |
|---|---|---|
| Turnover (both sides of every trade) | 24.8% | 20–40% |
| Cost, share of the book | 0.10% | 0.08–0.16% |
Across the thirteen resets since the basket's backtest begins (Sept 2023), costs total about 2.3% of the book. The index curve is drawn net of them, so it is what the positions would be worth after paying to rebalance.
The curve is also anchored to the real calendar. The backtest used to reset every 90 days from whenever its price data began, which put its resets on days nobody traded. Now the backtest and the published schedule turn over on the same dates.
For the token-level measurement behind every number here, see How to Value a DeFi Token. The live weights, the rebalance table and the watchlist are on the index page.
This article is research and education, not financial advice. The index illustrates an index-construction method using public data; it is not a recommendation or a licensed advisory product. Do your own research before committing capital.
From 2026-10-23 the index holds ten DeFi tokens: HYPE 25.2%, PENDLE 12.2%, AERO 10.6%, UNI 9.9%, PUMP 8.3%, JUP 7.9%, CAKE 7.3%, LIGHTER 7.1%, SKY 6.0% and ENA 5.5%. ENA was added and LDO removed. The weights were fixed on 2026-10-18 and stay the published targets until the next scheduled reset on 2027-01-21.
Capital is split 70/30 between two sleeves. The cash-flow sleeve weights each token by the square root of its forward holders revenue. The torque sleeve weights it by how far its price sits below what a return to its own last-cycle fee peak would justify, net of dilution. Each sleeve caps any name at 30%, then the two are blended and the ten largest weights make the book.
Every 90 days on a fixed calendar counted from 2022-08-15, so resets fall on 2026-07-25, 2026-10-23, 2027-01-21 and 2027-04-21. Ninety days matches the window every weighting input is measured on. Positions reset to target at that day's close and then drift freely until the next reset, so the published weights are targets rather than daily holdings.
Yes. Every reset, including the initial purchase, is charged 40 basis points per dollar traded on the basket: 10 bps of venue fee plus 30 bps of slippage sized for a $50M book. BTC and ETH legs of the blend are charged 10 bps. The October 2026 reset turned over 24.8% of the book for about 0.10%, and costs across thirteen backtest resets total roughly 2.3%.
Ethena's fee switch has approved terms: 5% of gross revenue goes to ENA buybacks once USDe's 14-day average supply reaches $7.5B, rising in tiers to 20% at $20B. The index prices that payout from the approved terms at a stated USDe view and flags it as estimated. It carries 5.5% of the book, and until USDe reaches the trigger that weight should be read as a forecast.
They ranked eleventh to fifteenth by weight, and most of them were sized on payouts that do not exist yet. Cutting them reduced the book's reliance on estimated fee switches from 17% to 5.5% while barely changing volatility. Each sits on a watchlist with a named trigger, such as a MORPHO fee switch passing or AAVE restarting buybacks, and re-enters at the next scheduled reset once it fires.
Practitioner turned analyst tracking how incentives, liquidity, and capital flows shape DeFi protocols.
