2. What RAVA Actually Underwrites
Most curators and risk analysts focus on intrinsic risk: is the underlying portfolio sound? Are the borrowers creditworthy? Is the strategy well constructed? Those are important questions. But they are not what RAVA computes.
RAVA underwrites illiquidity risk. The cost of getting out. When a DeFi vault holds tokenized private credit as collateral, the question is not "is this fund a good investment?" It is: "if we need to buy this asset right now, what discount to NAV reflects the true cost of the liquidity we are providing?"
For a liquid token like ETH, the answer is straightforward: check the order book, compute slippage (the price impact of selling), done. For private credit, the answer depends on redemption queues, reserve balances, fund level gates (rules that limit how much investors can withdraw at once), and the gap between stated NAV and realizable exit price.
There is no daily market price. NAV updates weekly or monthly, smoothed by fund administrators. The price appears stable between updates.
RAVA's approach: decompose the liquidity cost into two layers. A market risk floor (captured through liquid proxies and CVaR) and a fund specific liquidity premium (captured through onchain and offchain operational signals, such as redemption buffer balances and queue sizes, that track how hard it actually is to exit).
3. Building a Proxy Basket
To measure market risk, we construct a synthetic private credit portfolio from liquid instruments that share exposure characteristics with F ONE's underlying strategy. The goal is not to replicate the fund. It is to build a volatility reference from instruments that actually have daily prices.
Senior Loans ETF (exchange traded fund). Leveraged loan exposure, interest rates that adjust with the market.
High Yield Corp Bonds. Tracks how riskier corporate debt moves when market conditions change.
Ares Capital BDC (business development company). Direct lending to middle market companies.
Main Street Capital BDC. Lower middle market lending, equity coinvestments alongside borrowers.
| Proxy | Weight | What it captures |
|---|
| BKLN | 35% | Senior Loans ETF (exchange traded fund). Leveraged loan exposure, interest rates that adjust with the market. |
| HYG | 20% | High Yield Corp Bonds. Tracks how riskier corporate debt moves when market conditions change. |
| ARCC | 25% | Ares Capital BDC (business development company). Direct lending to middle market companies. |
| MAIN | 20% | Main Street Capital BDC. Lower middle market lending, equity coinvestments alongside borrowers. |
A 0.5x dampening factor is applied to the raw proxy returns. Private credit is structurally less volatile than public markets because NAVs are appraised periodically, not repriced by live trading every day. The dampening prevents overstating volatility from liquid proxies.
A structural illiquidity premium is added to the base discount (2.8% in the case of mF-ONE, broken down in the formula section below). This compensates for the gap between public liquidity (sell an ETF in seconds) and private fund redemption timelines (weeks to months).
Over five years (March 2021 to February 2026), the dampened proxy basket returned 7.53% annualized with 5.22% annualized volatility and a Sharpe ratio (return per unit of risk) of 0.48, using a 5% risk free rate.
Proxy basket cumulative return vs mF-ONE NAV (%, overlap period)
4. How Well Does the Proxy Track?
A proxy basket is only useful if it actually correlates with the real asset. So we tested it two ways: daily return regression and cumulative trend comparison.
On daily returns, the regression gives R² = 1.5%, a measure of how much of the fund's movement the proxy explains (4.6% weekly). Nearly zero.
The biggest NAV drops (December 4: -2.07%, January 19: -0.90%) happened when public credit markets were calm. These were fund specific events: redemption processing, portfolio marks, operational rebalancing. The proxy did not see them coming, and it was not supposed to.
But look at the cumulative chart above. Strip out the two drops, and both lines climb at nearly the same pace: the proxy returned ~4.7% over the overlap period, mF-ONE returned ~5.0%. Both are earning credit income from the same asset class (private lending, floating rate, middle market). The trend tracks. The shocks do not.
This is the right behavior for a proxy. It captures the structural credit return that both portfolios share. What it does not capture is fund specific operational risk: NAV reductions from loan losses, redemption processing hits, and shifts in the fund's cash buffer. Those show up in the onchain signals, not the proxy. The two layers are complementary, not redundant.
The data is also limited. mF-ONE has only been live since May 2025, giving us roughly nine months of oracle price history and 152 overlapping trading days. That is not enough to draw strong statistical conclusions. With more history, the daily R² could shift.
And there is a more important reason the proxy matters: we have not yet seen a market wide credit event during mF-ONE's lifetime. The proxy's five year history includes the 2022 rate shock and the April 2025 tariff selloff, where ARCC and MAIN each dropped over 7% in a single day. Those scenarios have not hit mF-ONE yet. When one does, the proxy is what catches it and widens the discount before the fund's smoothed NAV reflects the damage.
Daily returns: proxy basket vs mF-ONE (R² = 1.5%)
Red dots: mF-ONE loss days > 0.5%. These cluster near zero on the proxy axis, confirming the drops were fund specific, not market wide.
5. Value at Risk
Given a confidence level and a time horizon, Value at Risk (VaR) is the loss threshold you should not expect to exceed. At 95% confidence, VaR says: "on 95% of days, your loss will be smaller than this number."
Three computation methods:
Historical. Sort actual returns, pick the percentile. No distribution assumptions. Just the data.
VaRα = −R⌊(1−α) × N⌋
Parametric. Assume returns are normally distributed. Fast, but breaks down with fat tails.
VaRα = −(μ + zα × σ)
Monte Carlo. Simulate thousands of scenarios from fitted distributions. Most flexible, most computationally expensive.
We use Historical because private credit returns are decidedly not normal. They exhibit negative skew (losses are larger than gains) and fat tails (extreme days happen more often than a bell curve predicts): small gains most days, occasional sharp drops.
The worst day in our proxy dataset was -1.99% (April 4, 2025, when the VIX volatility index spiked above 40). The 20 worst days cluster around periods of acute market stress.
| Confidence | Daily VaR | Monthly VaR | Tail Obs |
|---|
| 90% | 0.328% | 1.50% | 125 / 1,255 |
| 95% | 0.506% | 2.32% | 62 / 1,255 |
| 99% | 1.054% | 4.83% | 12 / 1,255 |
| 99.7% | 1.742% | 7.98% | 3 / 1,255 |
Daily return distribution (synthetic proxy, 1,255 trading days)
Red bins: losses beyond VaR 95 (0.506% daily). Amber bins: between VaR 90 (0.328%) and VaR 95. The left tail is where the discount gets priced.
6. Expected Shortfall (CVaR)
VaR answers: "what is the worst I would expect at this confidence?" But it says nothing about what happens when things ARE worse. You know the boundary. You do not know the depth.
Expected Shortfall (also called CVaR or Conditional VaR) answers: "when the loss exceeds VaR, how bad is it on average?"
ESα = (1/k) × Σ(losses beyond VaR) where k = number of observations in the tail
ES is wider than VaR by construction. It captures tail severity, not just frequency. At VaR 95, the boundary is a 0.506% daily loss. But within that worst 5% of days, the average loss is 0.807%, the median is 0.671%, and the worst single day was 1.989%. The tail is 1.6x the boundary, and the worst day is 3.9x.
This is why RAVA calibrates discounts to ES rather than VaR alone. If a vault is buying distressed assets, it needs to price for the average bad outcome, not just the boundary.
| Confidence | VaR (monthly) | ES (monthly) | ES/VaR |
|---|
| 90% | 1.50% | 2.79% | 1.85x |
| 95% | 2.32% | 3.70% | 1.59x |
| 99% | 4.83% | 6.24% | 1.29x |
| 99.7% | 7.98% | 8.58% | 1.07x |
VaR vs Expected Shortfall by confidence level (monthly horizon)
These four CVaR tiers drive vault pricing, how yield products split returns between protected and higher risk depositors, how much collateral a borrower must post, and vault entry pricing. For the full framework, see One Dial, Three Products: How CVaR Drives Yield, Lending, and Vaults.
7. Onchain Signals: What the Fund Data Shows
Since the proxy only captures market risk, we need to measure fund specific risk directly. mF-ONE publishes daily transparency reports: total assets, token supply, settlement reserve, portfolio allocation, and (since January 2026) pending burns and redemptions. We analyzed all 165 available reports. Here is what the data shows.
The supply is declining rapidly
Peak: 160.7M tokens ($168.5M) on October 30, 2025. As of February 28, 2026, gross supply is 98.27M tokens, but 34.92M sit in the redemption vault, leaving 63.35M in net circulation ($67.2M). That is a 61% decline in net circulating supply over four months.
The price per token went up 1.2% during this period. Nearly all of the market cap decline came from redemptions, not losses. Token holders are redeeming, not taking losses on price.
Token supply (M tokens) and price per token over time
The Market Cap Decline Is From Redemptions
Peak market cap: $168.5M (October 30, 2025). Current net circulating market cap: $67.2M (63.35M tokens at $1.0608). That is a $101.3M decline.
But the token price went from $1.0486 to $1.0608. It went up 1.2%. Substantially all of the market cap decline came from supply reduction. Holders redeemed their tokens. No holder lost money on price over this observation period.
The fund is not declining in value. It is shrinking in size. That distinction matters because a shrinking fund creates its own liquidity dynamics independent of credit quality.
Market cap decomposition: supply effect vs price effect ($M)
The green dashed line (price effect) stays flat or slightly rises. The blue dashed line (supply effect) tracks the actual market cap almost exactly. Supply drove 101% of the decline. Price contributed a small positive offset.
The liquidity buffer composition shifted
Midas maintains a liquidity buffer for instant redemptions. We traced the onchain history of the liquidity provider wallet (Fordefi 2) at historical block heights. The buffer is held as mTBILL (Midas US Treasury Bill Token), not USDC. During the growth phase (June through October), it sat consistently at $7M to $16.3M, peaking at $16.3M on October 30.
By November 14, the LP wallet held zero mTBILL. At the same time, 1.65M mF-ONE tokens ($1.74M) appeared in the redemption vault as the first pending redemption requests. By November 21, the queue had grown to 4.01M tokens ($4.24M) while the buffer remained at zero. With no buffer and a growing queue, instant redemptions were not available during this window.
The buffer was partially restored by December 4 ($2.4M), returned to $30K by December 18, then rebuilt to $8.4M by December 31. It has held between $5.5M and $9.1M since January. The composition is worth noting: the buffer is held in mTBILL, not stablecoins. The instant redemption fee is 1.00%, and redeemInstant() (the contract function for immediate redemptions) pulls from the LP wallet via transferFrom (a standard ERC-20 transfer call), meaning the mTBILL is converted before a redeemer receives stablecoins.
The underlying F ONE fund allows quarterly redemptions at 5% of NAV. As of February 28, 35.5% of gross supply sits in the redemption vault, which exceeds what a single quarterly cycle can process.
As of February 28, the redemption vault holds 34.92M mF-ONE tokens ($37M at oracle price) across 8 pending redemption requests, representing 35.5% of gross supply. Two requesters account for 21.4M of that total. Every token in the vault matches a pending request down to the smallest decimal.
Where the money actually sits
The Midas dashboard displays $6.81M in "instant redemption capacity." We traced every onchain address referenced by the four published contracts. Here is where the money actually sits:
| Wallet | Role | Holdings | Value |
|---|
| Fordefi 2 (LP) | liquidityProvider | 6,882,669 mTBILL | $7.25M |
| Fordefi 1 | tokensReceiver | 420,463 USDC | $420K |
| Fordefi 3 | requestRedeemer | ~0 | -$40K |
| All contracts | vaults, token, oracle | ~$15 USDC | ~$0 |
As of February 28, the liquidity provider wallet holds 6,882,669 mTBILL ($7.25M). The redeemInstant() function pulls funds from this address via transferFrom, giving holders an instant exit path at a 1.00% fee. The $37M in the redemption queue represents holders who chose redeemRequest() (the queued redemption function) instead, which has no fee but requires waiting for the underlying fund to process redemptions.
The Midas transparency page (as of February 26) broke down $103.78M in total TVL: $84.41M in Fasanara Capital (the fund), $11.87M in "Fasanara Capital redemption process" (money being pulled from the fund), $7.25M in mTBILL as the liquidity buffer, and $254K in USDC. The instant liquidity buffer remains available for holders willing to use the 1.00% instant redemption path. The queue reflects holders who opted for the fee free path instead.
LP liquidity buffer ($M, mTBILL in LP wallet) vs redemption queue ($M, mF-ONE in redemption vault). All values verified from historical onchain block queries.
The liquidity chain: what backs the buffer
The LP wallet holds $7.25M in mTBILL. But mTBILL is not a stablecoin. It is a tokenized claim on Superstate's USTB, which is itself a tokenized share of a US Treasury fund. When a holder calls redeemInstant(), the transaction passes through three contracts in a single atomic execution:
mF-ONE Redemption Vault
Burns mF-ONE from the holder, pulls mTBILL from the LP wallet via transferFrom
0x44b0...9C1.00% fee
mTBILL Redemption Vault
Receives mTBILL, redeems against its USTB reserves (4.12M USTB, ~$45M backing)
0x569D...Ec0.07% fee
Superstate RedemptionIdle
Burns USTB, sends USDC from its settlement pool to the redeemer
0x4c21...CfShared across ALL USTB holders
Total combined fee: ~1.07%Execution: 1 atomic transaction
The USTB settlement pool: the binding constraint
The effective instant redemption capacity is the minimum across all three layers. The LP wallet may hold $7.25M in mTBILL, but if the Superstate settlement pool has $980 in USDC (as it did on February 22), the effective capacity is $980.
mTBILL is a tokenized Treasury bill holding another tokenized Treasury bill. The mTBILL redemption vault holds 4.12M USTB tokens (~$45.36M at $11.00/USTB), representing approximately 97.5% of mTBILL's $46.5M market cap.
USTB is Superstate's Short Duration US Government Securities Fund, holding short maturity US Treasury Bills. That layer has ample reserves. The binding constraint is the final hop: Superstate's onchain settlement pool, which holds USDC for instant USTB redemptions.
This settlement pool is shared across all USTB holders, not just Midas. When other protocols or holders redeem USTB, they draw from the same pool. Superstate periodically replenishes it ($5M to $7M at a time), but between refills, the balance can swing from $9.9M to under $100 in a single day depending on redemption activity from other participants.
USTB redemption paths: onchain and offchain
USTB has two redemption paths. The onchain path (via the RedemptionIdle contract) is atomic: burn USTB, receive USDC, single transaction, available 24/7. There is currently no fee (the contract has a configurable fee parameter set to zero).
However, the onchain path has no queuing mechanism. If the pool does not hold enough USDC to fill the redemption, the transaction reverts. There are no partial fills and no IOUs.
The contract exposes a maxUstbRedemptionAmount() view function that returns the maximum redeemable amount at any given moment, currently ~618,969 USTB (~$6.81M).
The alternative is offchain redemption: transfer USTB to a designated address or call offchainRedeem(). This path settles same day if submitted before 1:00 PM ET, or next business day after that, with a two business day buffer documented in the fund terms. It only processes on days when both the NYSE and the Federal Reserve Bank of Philadelphia are open.
The offchain path has no cap (it liquidates underlying Treasury bills through Federated Hermes, with proceeds flowing through UMB Bank to Circle for USDC minting), but it is not atomic (instant and all or nothing) and cannot be used within automated DeFi transactions.
Replenishment of the onchain pool is manual. Superstate describes it as "regular" but discloses no schedule. The pool launched with $10M.
When it drains, the only recourse is to wait for a refill or use the offchain path. For a protocol like mF-ONE that chains through this pool atomically, a dry pool means the instant redemption function reverts entirely, regardless of how much mTBILL the LP wallet holds.
Effective capacity over 30 days
Over a 30 day observation window (January 31 to February 28), the USTB settlement pool was the binding bottleneck on 23 out of 30 days. Average effective daily capacity ranged from under $100K to $8.7M depending on the day. On two days the pool held under $100:
| Date | USTB Settlement | LP mTBILL | Effective Max | Bottleneck |
|---|
| Jan 31 | $55 | $8.70M | $55 | USTB Settlement |
| Feb 1 | $25 | $8.70M | $25 | USTB Settlement |
| Feb 18 | $9.92M | $7.49M | $7.49M | LP mTBILL |
| Feb 22 | $980 | $7.53M | $980 | USTB Settlement |
| Feb 25 | $1.67M | $6.68M | $1.67M | USTB Settlement |
| Feb 28 | $6.81M | $7.25M | $6.81M | USTB Settlement |
30 day instant redemption capacity: USTB settlement vs LP mTBILL vs effective max ($M)
The effective max (amber) tracks whichever layer is lowest. On 23 of 30 days, USTB settlement (red dashed) was the binding constraint, not the LP wallet (purple dashed).
Implications for RAVA
This multilayer chain means monitoring the LP wallet alone is not sufficient. The Liquidity Sleeve Health signal in RAVA's discount formula needs to account for the effective capacity across the entire redemption chain, not just the top layer. A buffer that shows $7.25M in mTBILL but can only deliver $25 in USDC is not a $7.25M buffer.
This also suggests a second application. A vault using this framework could provide liquidity not just at the mF-ONE level (buying mF-ONE at a discount), but within the sleeve itself: providing USDC liquidity at the USTB settlement layer, or holding mTBILL directly to backstop the LP wallet when it runs thin. Any tokenized fund asset that relies on a multihop redemption chain has this same structural pattern, and each hop is a potential point where a liquidity provider could deploy capital and earn a spread.
The contracts involved in this chain are all verified onchain:
| Layer | Contract | Balance |
|---|
| LP Wallet (mTBILL) | 0x4dc2...3071 | 6.88M mTBILL ($7.25M) |
| mTBILL Vault (USTB) | 0x569D...f0Ec | 4.12M USTB ($45.36M) |
| USTB Settlement (USDC) | 0x4c21...54Cf | $6.81M USDC |
8. The Dynamic Discount Formula
This is where RAVA's illiquidity focus becomes concrete. The discount is not a credit spread or a default probability. It is the price of liquidity: what it costs to convert this position to cash, right now, given current fund conditions.
The discount has two components: static and dynamic.
Static component (2.8%)
Structural illiquidity that changes rarely:
- 2.0% Base illiquidity premium: private credit, monthly NAV, limited portfolio visibility
- 0.5% Legal/structural complexity: SPV (special purpose vehicle) wrapper, offshore domicile, ERC-20 layer
- 0.3% Execution spread: the discount buyers demand when they know the seller has to sell
Dynamic component (0% to 8%)
Four weighted signals, updated daily:
20% weightProxy CVaR (Market Risk)
Input: CVaR 95 from the dampened proxy basket (monthly horizon)
Source: 5 year historical returns from BKLN/HYG/ARCC/MAIN
The market risk floor. Even if the fund has perfect operational health (full buffer, no queue, stable supply), there is still a cost to holding illiquid credit during a market wide selloff. The proxy CVaR captures this: at 95% confidence, the average loss in the worst 5% of months is 3.70%. This sets the minimum discount regardless of fund specific conditions.
CVaR 95 ≤ 2% = score 0 | 3% = 25 | 4% = 50 | 6%+ = 100
20% weightBleed Rate
Input: 30 day rolling supply change
Source: onchain totalSupply() snapshots
The velocity of outflows matters more than the absolute size of the fund. A fund with stable, balanced flows (investors entering and leaving at similar rates) can manage redemptions orderly. A fund with 10%+ supply decline per month may need to liquidate portfolio positions on shorter timelines. The rate of change is what this signal captures.
0% change = score 0 | -10% = 50 | -30% = 100
30% weightLiquidity Sleeve Health
Input: Effective instant redemption capacity / Total Assets
Source: onchain balance queries across the full redemption chain
When the sleeve is full, holders can exit at 1.00% instant. When empty, the effective exit cost jumps to 4% to 8%. This gap is what RAVA's vault fills. Critically, the "sleeve" is not a single balance. For mF-ONE, the instant redemption path passes through three contracts (LP wallet, mTBILL vault, USTB settlement), and the effective capacity is the minimum across all three. The USTB settlement pool is the binding constraint on most days.
10%+ = score 0 | 5% = 25 | 2% = 50 | <0.1% = 100
30% weightRedemption Queue Pressure
Input: Queue dollar value / Quarterly redemption capacity
Source: transparency reports, onchain pending burns
What matters is the rate at which the queue is growing relative to the fund's capacity to process it. A stable queue that clears each quarter is healthy. A queue that doubles in two weeks signals a run. When the queue exceeds one quarter's redemption capacity, it starts backing up. At 3x+ capacity, the fund faces a structural liquidity constraint.
0x = score 0 | 1x = 50 | 2x = 75 | 4x+ = 100
Dynamic score = 0.20 × cvar + 0.20 × bleed + 0.30 × liquidity + 0.30 × queue
Dynamic discount = score × 8% (maximum additional spread)
CVaR scaling applied to the total discount:
| CVaR Level | Multiplier |
|---|
| CVaR 90 | total × 0.72 |
| CVaR 95 | total × 1.00 (base case) |
| CVaR 99 | total × 1.18 |
| CVaR 99.7 | total × 1.42 |
Dynamic discount (CVaR 99.7) vs static Morpho oracle over time
9. The Timeline
Key dates with the computed RAVA discount alongside the static Morpho oracle:
Jul to Oct
RAVA 4.0%Morpho 7.7%
Growth phase. Sleeve at 10%. Static only.
Nov 3
RAVA 4.3%Morpho 7.7%
Supply starts declining. Buffer still at $16.9M.
Nov 14
RAVA 10.1%Morpho 7.7%
Buffer at zero. Queue appears: 1.65M tokens.
Nov 21
RAVA 10.2%Morpho 7.7%
Buffer still zero. Queue grows to 4.01M tokens.
Dec 4
RAVA 8.7%Morpho 7.7%
NAV drops -2.07%. Buffer partially back ($2.4M). Queue: 4.22M.
Dec 18
RAVA 8.2%Morpho 7.7%
Buffer at $26K. Queue mostly processed (0.4M).
Dec 31
RAVA 7.4%Morpho 7.7%
Buffer rebuilt to $8.4M. Queue grows to 10.75M tokens.
Jan 13
RAVA 8.0%Morpho 7.7%
Buffer stable ($8.5M mTBILL). Queue partially cleared: 6.38M.
Jan 30
RAVA 9.7%Morpho 7.7%
Queue grows to 18.6M tokens. Buffer $9.1M.
Jan 31
RAVA 9.7%Morpho 7.7%
USTB settlement pool drops to $55. Effective instant capacity: $55.
Feb 1
RAVA 9.7%Morpho 7.7%
USTB settlement at $25. LP $8.7M mTBILL but zero effective capacity.
Feb 22
RAVA 10.2%Morpho 7.7%
USTB settlement at $980. LP $7.53M mTBILL. Effective: $980.
Feb 26
RAVA 9.2%Morpho 7.7%
Queue 11.25M. Buffer $7.25M. USTB settlement $9.75M.
Feb 28
RAVA 11.1%Morpho 7.7%
Queue 34.9M ($37M). Buffer $7.25M. USTB $6.81M.
| Date | Event | RAVA | Morpho |
|---|
| Jul to Oct | Growth phase. Sleeve at 10%. Static only. | 4.0% | 7.7% |
| Nov 3 | Supply starts declining. Buffer still at $16.9M. | 4.3% | 7.7% |
| Nov 14 | Buffer at zero. Queue appears: 1.65M tokens. | 10.1% | 7.7% |
| Nov 21 | Buffer still zero. Queue grows to 4.01M tokens. | 10.2% | 7.7% |
| Dec 4 | NAV drops -2.07%. Buffer partially back ($2.4M). Queue: 4.22M. | 8.7% | 7.7% |
| Dec 18 | Buffer at $26K. Queue mostly processed (0.4M). | 8.2% | 7.7% |
| Dec 31 | Buffer rebuilt to $8.4M. Queue grows to 10.75M tokens. | 7.4% | 7.7% |
| Jan 13 | Buffer stable ($8.5M mTBILL). Queue partially cleared: 6.38M. | 8.0% | 7.7% |
| Jan 30 | Queue grows to 18.6M tokens. Buffer $9.1M. | 9.7% | 7.7% |
| Jan 31 | USTB settlement pool drops to $55. Effective instant capacity: $55. | 9.7% | 7.7% |
| Feb 1 | USTB settlement at $25. LP $8.7M mTBILL but zero effective capacity. | 9.7% | 7.7% |
| Feb 22 | USTB settlement at $980. LP $7.53M mTBILL. Effective: $980. | 10.2% | 7.7% |
| Feb 26 | Queue 11.25M. Buffer $7.25M. USTB settlement $9.75M. | 9.2% | 7.7% |
| Feb 28 | Queue 34.9M ($37M). Buffer $7.25M. USTB $6.81M. | 11.1% | 7.7% |
The LP buffer went to zero on November 14 and stayed there through November 21, while the redemption queue grew from 1.65M to 4.01M tokens. This preceded the December 4 NAV adjustment (−2.07%) by 20 days. A dynamic model would have widened the discount to 10.1% on November 14. A static oracle does not adjust to these changes.
The USTB settlement events on January 31 ($55), February 1 ($25), and February 22 ($980) illustrate a second dimension: even when the LP wallet holds $7M+ in mTBILL, the effective instant capacity can be near zero because of downstream constraints that have nothing to do with mF-ONE. A dynamic model that monitors the full redemption chain would widen the discount on those days. A model that only watches the LP wallet would not.
10. Who Is Borrowing Against mF ONE
On chain data from the Morpho mF ONE market shows 9 active borrowers with $13.6M in total borrows against $25M in collateral. Tracing each wallet's funding source, transaction history, and connected addresses reveals the financial profile behind each position.
Borrower profiles (on chain data, March 2026)
Type: Safe multisig
LTV: 25%
Collateral: $10.3M
Net worth: $64.9M
Funded by: Safe deployer
Other assets: $14.9M USDC, $5.1M iUSD, $1.7M WBTC, $41.8M on Avalanche
Type: EOA
LTV: 70%
Collateral: $6.6M
Net worth: $6.7M
Funded by: Binance
Other assets: $6.7M in aWBTC on Aave, $31 liquid
Type: EOA
LTV: 81%
Collateral: $5.9M
Net worth: $7.4M
Funded by: Bybit
Other assets: $1.5M in mHYPER, $651 liquid
Type: EOA
LTV: 67%
Collateral: $824K
Net worth: $827K
Funded by: DeFi wallet ($3.1K)
Other assets: Purpose built wallet. Parent holds $3.1K across 81 tokens.
Type: EOA
LTV: 71%
Collateral: $446K
Net worth: $456K
Funded by: DeFi wallet ($9.3K)
Other assets: Purpose built wallet. Parent holds $9.3K (57% USDC).
Borrower #6
Uncertain riskType: EOA
LTV: 73%
Collateral: $349K
Net worth: Unknown
Funded by: Kraken
Other assets: $198 liquid. CEX assets unverifiable.
Type: Safe multisig
LTV: 75%
Collateral: $249K
Net worth: $262K
Funded by: hubirb.eth + ferrari-strategist.eth
Other assets: Signers hold $10K combined. Curve/Stake DAO ecosystem.
Borrower #8
Uncertain riskType: EOA
LTV: 69%
Collateral: $220K
Net worth: Unknown
Funded by: Coinbase
Other assets: $4K in PENDLE, MORPHO, EUL. Active DeFi user (822 txns).
Borrower #9
Uncertain riskType: EOA
LTV: 80%
Collateral: $184K
Net worth: Unknown
Funded by: Cross chain bridge (Bungee)
Other assets: $16.3K liquid. $12.1K staked AVAX on Avalanche. True source wallet on another chain.
| ID | Type | Collateral | LTV | Total net worth | Funded by | Key assets beyond mF ONE |
|---|
| #1 | Safe | $10.3M | 25% | $64.9M | Safe deployer | $14.9M USDC, $5.1M iUSD, $1.7M WBTC, $41.8M on Avax |
| #2 | EOA | $6.6M | 70% | $6.7M | Binance | $6.7M aWBTC on Aave. $31 liquid. |
| #3 | EOA | $5.9M | 81% | $7.4M | Bybit | $1.5M mHYPER. $651 liquid. |
| #4 | EOA | $824K | 67% | $827K | DeFi wallet ($3.1K) | Purpose built. Parent: $3.1K across 81 tokens. |
| #5 | EOA | $446K | 71% | $456K | DeFi wallet ($9.3K) | Purpose built. Parent: $9.3K. |
| #6 | EOA | $349K | 73% | Unknown | Kraken | $198 liquid. CEX assets unverifiable. |
| #7 | Safe | $249K | 75% | $262K | hubirb.eth | Signers hold $10K combined. |
| #8 | EOA | $220K | 69% | Unknown | Coinbase | $4K in DeFi governance tokens. 822 txns. |
| #9 | EOA | $184K | 80% | Unknown | Bridge (Bungee) | $16.3K liquid. $12.1K staked AVAX. |
Only borrower #1 has clearly sufficient liquid assets ($14.9M in USDC alone) to cover their position in a margin call. Borrowers #2 and #3 hold large positions but are concentrated in single assets (aWBTC on Aave, mHYPER) with virtually no liquid reserves.
Borrowers #4 and #5 were funded by small DeFi wallets holding $3K and $9K respectively. These are purpose built borrowing wallets with no visible backup. Borrowers #6, #8, and #9 were funded from centralized exchanges or cross chain bridges, meaning their real net worth exists off chain or on other chains and cannot be verified.
Borrower #7 is a Safe multisig controlled by two ENS named users active in the Curve ecosystem. Technically sophisticated but the combined visible assets of both signers total roughly $10K against a $249K position.
A traditional NAV lender would assess the borrower's ability to service the loan and respond to margin calls as part of the underwriting process. On chain, this information is partially available through wallet analysis but requires active monitoring that no curator currently performs. The gap between the visible net worth of most borrowers and the size of their positions raises questions about who would defend these positions in a markdown scenario.
11. Takeaway
Other curators evaluate the intrinsic quality of the asset: is the strategy sound, are the borrowers creditworthy, is the portfolio diversified? That work matters. RAVA does not replace it.
What RAVA adds is a liquidity cost layer: how much does it cost to exit this position today? How does that cost change when redemption queues build, when the liquidity sleeve balance shifts, when supply is declining?
Static discounts assume liquidity cost is constant. It is not. The same asset can be cheap to exit in October and prohibitively expensive to exit in January, with zero change in credit quality.
The four dynamic signals (proxy CVaR, bleed rate, sleeve health, queue depth) are all observable. The proxy basket uses public market data. Two fund signals are onchain (supply snapshots, pending burns). One comes from transparency reports that the fund is already publishing.
The sleeve health signal extends beyond the immediate buffer: for assets with multihop redemption chains, it traces the effective capacity across every layer down to the final USDC settlement.
This analysis also surfaces a second application beyond pricing discounts: providing liquidity within the redemption chain itself. When the binding constraint is a downstream settlement pool (as with USTB), a vault that holds USDC at that layer could backstop the instant redemption path and earn a spread on the liquidity it provides. Any tokenized fund with a multihop exit path has these same structural bottlenecks, and each one is a potential point of intervention.
The inputs are computable today for any tokenized fund that publishes basic operational data. The core question for a DeFi lending vault remains: not "is this asset good?" but "if I need to sell this, what will it cost me?"