Sizing the Cushion
Off chain assets cannot be independently valued quickly or cheaply. We do not price the asset. We price the leverage. The question is: how much LTV buffer does a lender need given the uncertainty in the reported NAV?
1. Why We Don't Price the Asset
Most tokenized real world assets are Level 2 or Level 3 under ASC 820. Level 1 means exchange traded with observable prices. Level 2 means the asset can be valued using observable inputs from similar instruments. Level 3 means the valuation depends on unobservable inputs: internal models, GP assumptions, appraisals.
For private credit funds, the reported NAV comes from the administrator. The administrator relies on GP marks. The GP marks the book using internal models, recovery assumptions, and borrower financials that are months stale. There is no ticker. There is no closing price. The number is an estimate published on a schedule, typically monthly or quarterly.
Independent valuation exists but it is slow and expensive. Firms like Kroll, Houlihan Lokey, and Lincoln International perform annual or semi annual fair value assessments. A full portfolio valuation for a mid size private credit fund takes months and costs six figures. You cannot call Kroll at 2 PM and ask for a mark by 3 PM.
A lending protocol that waits for an independent valuation will never liquidate on time. A lending protocol that trusts the GP mark without adjustment will eventually lend against a stale number and lose capital. The solution is to stop trying to determine what the asset is worth and instead measure how wrong the reported number could be.
2. What We Price Instead
We build a proxy basket of liquid, publicly traded instruments whose return characteristics approximate the underlying fund. For a direct lending fund, that basket might include leveraged loan ETFs (BKLN), high yield bond ETFs (HYG), and listed BDCs (ARCC, MAIN). The proxy basket trades every day. It has observable prices, observable volatility, and observable drawdowns.
We run CVaR (Conditional Value at Risk) on the proxy basket to answer a specific question: in the worst X% of scenarios over a given time horizon, how much does this basket lose? That loss number becomes the leverage cushion. If the CVaR 95 over 21 trading days is 3.8%, then a lender operating at CVaR 95 needs at least 3.8% of LTV buffer between the loan amount and the reported NAV.
The proxy basket does not tell you what the fund is worth. It tells you how much the fund's value could move before the next reliable mark arrives. The cushion absorbs that uncertainty. If the cushion is larger than the actual move, the lender is whole. If it is smaller, the lender takes a loss.
A wider cushion protects the lender but reduces the borrower's capital efficiency. A thinner cushion improves capital efficiency but exposes the lender to mark to market loss. CVaR confidence levels let each participant choose where they sit on that curve.
3. Case Study: mF ONE
Maple's mF ONE is a tokenized private credit fund investing primarily in senior secured direct lending. The proxy basket we constructed for mF ONE uses four instruments: BKLN (senior leveraged loans), HYG (high yield corporates), ARCC (Ares Capital, listed BDC), and MAIN (Main Street Capital, listed BDC). The basket weights reflect the fund's exposure profile across floating rate senior secured, high yield, and broadly syndicated credit.
The proxy basket tracked mF ONE's NAV reasonably well during normal conditions through the second half of 2025. Both moved in the same direction with the proxy showing slightly higher volatility, which is expected for liquid instruments versus smoothed GP marks.
The breakdown came in early December 2025. mF ONE's NAV dropped approximately 2% in a single administrator update. The cause was a write down on First Brands Group, a portfolio company that had been experiencing operational deterioration. The proxy basket barely moved that day. BKLN, HYG, ARCC, and MAIN are diversified across hundreds of issuers. A single name event in a concentrated portfolio does not register in a broad proxy.
This is the known limitation. Proxy baskets capture systematic risk well. They miss idiosyncratic loss events entirely. A lender relying only on proxy CVaR would have been caught by the First Brands write down.
| Metric | Value |
|---|---|
| Redemption queue | $37M |
| Instant redemption capacity | $3M to $8M typical, below $1M on 3 of 18 sampled days |
| NAV drop (single update) | ~2% |
| Proxy basket move (same day) | <0.1% |
The redemption queue growing to $37M while instant capacity approached zero illustrates why the cushion must account for more than market risk. When a lender cannot exit the position on demand, the relevant time horizon for the cushion is not the market's settlement cycle. It is the fund's redemption cycle. For mF ONE at that moment, the effective exit horizon was measured in weeks.
4. The Discount Tiers
CVaR at different confidence levels produces different cushion sizes. The choice of confidence level is a risk preference, not a calculation. Two lenders looking at the same asset can rationally choose different tiers.
| Tier | Cushion | Max LTV | Lender profile |
|---|---|---|---|
| CVaR 90 | 2.7% | 97.3% | Aggressive. Thin buffer, higher yield, accepts tail exposure. |
| CVaR 95 | 3.8% | 96.2% | Base case. Standard protection for institutional lending. |
| CVaR 99 | 4.5% | 95.5% | Conservative. Covers nearly all systematic scenarios. |
| CVaR 99.7 | 5.4% | 94.6% | Maximum protection. Wide cushion absorbs deep tail events. |
The numbers above are illustrative for the mF ONE proxy basket. Each asset on the platform has its own proxy basket and its own CVaR surface. A fund holding exclusively senior secured first lien loans will produce a tighter cushion than a fund holding mezzanine or equity tranches. The tiers scale proportionally: a wider base discount at CVaR 95 produces proportionally wider cushions at 90, 99, and 99.7.
Idiosyncratic events like the First Brands write down fall outside the proxy CVaR distribution. The protocol handles this with an additional illiquidity premium layered on top of the proxy CVaR. This premium incorporates fund specific signals: redemption queue depth, gate provisions, lock up periods, and the age of the most recent administrator mark. When a fund's redemption queue grows or its mark goes stale, the illiquidity premium widens, increasing the total cushion automatically. Because this methodology is standardized and reusable, the cost of sizing each new loan drops. That reduction flows through to borrowers as tighter spreads and faster origination.
Sources
- ASC 820 Fair Value Measurement, FASB Accounting Standards Codification
- Kroll (fka Duff & Phelps), Portfolio Valuation Best Practices, 2024
- BKLN, HYG daily returns via Yahoo Finance, Jan 2020 to Mar 2026
- ARCC, MAIN daily returns via Yahoo Finance, Jan 2020 to Mar 2026
- mF ONE NAV feed history, 158 data points, Midas
- First Brands Group credit event coverage, LCD News, December 2025
- Maple Finance mF ONE redemption disclosures, December 2025