Static Pool Backtest: Simulated Advance Rate Performance Through 2008 and 2020

1. Scope and Purpose

1.This backtest answers one question: if Ravariant had originated a pool of NAV loans at our proposed advance rates before 2008 or 2020, what would the warehouse bank have experienced?

2.The specific outputs are pool weighted average LTV trajectory, number and timing of margin calls, cash sweep activity, collateral release events, realized losses at the senior (warehouse bank) level, and first loss tranche absorption.

3.The backtest complements the methodology document. The methodology describes what we do. The backtest shows what the methodology would have produced under historical stress.

2. Data Sources

4.Every input is publicly sourced and verifiable. No proprietary data is used.

3. Simulation Methodology

5.We construct a hypothetical pool, apply the advance rate formula from the methodology at origination, then walk the pool quarter by quarter through the stress period using actual realized NAV moves from the benchmark indices.

4. Hypothetical Pool Construction

6.The pool is sized at $100M notional with 10 funds. Composition reflects the strategy mix our methodology targets and the eligibility gates in Section 2.

7.Pool weighted average advance rate: 52.6%. Starting loan balance: $52.6M against $100M NAV. HHI: 0.12 (within the 0.25 pool limit). All eligibility gates pass by construction.

5. Scenario 1: 2008 to 2010 Global Financial Crisis

8.Origination: Q2 2006. Stress period: Q3 2008 to Q4 2009. The pool is walked quarter by quarter using CDLI and Cambridge Associates realized returns blended by strategy weight.

GFC Scenario Results

9.The dynamic margin threshold tightened before NAV declined materially. This produced a preemptive margin call in Q4 2008 when live LTV (loan + interest / NAV) was 59% and the tightened margin threshold had moved to 45%. The 75% liquidation cap was never approached.

10.The cash sweep captured distribution flow (CDLI distribution yield was approximately 9 to 10% annualized through the crisis). The sweep reduced loan balance by $12.1M over four quarters, well in excess of the NAV decline.

11.No realized loss was required to emerge from the GFC under this methodology. The cushion, the dynamic margin threshold, and the distribution sweep together produced full recovery without any collateral release.

6. Scenario 2: 2020 COVID Shock

12.Origination: Q2 2018. Stress period: Q1 2020 to Q3 2020. The shock was faster and shallower than 2008 (CDLI peak drawdown of 5.8% vs 9.4% in 2008) with a more rapid rebound.

13.COVID produced a faster dynamic threshold tightening because HY OAS, VIX, and NFCI moved further and faster than in 2008. The system tightened the margin threshold to 40% in Q1 2020. Live LTV was 54%, so the margin call fired immediately. The 75% liquidation cap remained 21pp away.

14.Recovery was also faster. Indicators normalized by Q2 2020 as the Federal Reserve deployed emergency facilities. The loosening branch of the dynamic threshold released the margin call within one quarter.

7. Scenario 3: 2015 to 2016 Energy and HY Credit Crisis

15.Origination: Q1 2013. Stress period: Q4 2015 to Q2 2016. The dominant factor in this crisis was credit spread, not equity or rates. Energy sector default wave drove HY OAS from 350bps to 870bps. CDLI drew down 4.8% over 3 quarters, a slower bleed than 2008 but extended.

16.Peak live LTV 51%. Distance to 75% cap: 24pp. 1 margin call. 0 realized loss. Credit spread factor dominated; the indicator framework caught it through HY OAS and credit spread indices. Equity factor was muted (S&P 500 was -0.7% in 2015, +9.5% in 2016).

8. Scenario 4: 2022 Rate Shock

17.Origination: Q1 2020 (weathered COVID per Scenario 2, refinanced). Stress period: Q1 to Q4 2022. The dominant factor was rates: 10yr UST moved from 1.5% to 4.3% in 12 months, the largest rate shock in a generation.

18.Direct lending CDLI held up remarkably well because loans are floating-rate (SOFR based); rising rates increased interest income. Equity beta contribution was sharp (S&P 500 down 19%) but fund NAV was largely insulated. CDLI finished 2022 at +8.5%, better than any comparable asset class.

19.No margin call. The rate factor decomposition (Section 5 Factor Decomposition) correctly predicted direct lending's low rate sensitivity: beta_rates = 0.18 is small, and floating-rate loans actually benefit from rising SOFR. A PE buyout pool would have shown different behavior in this period because PE has higher equity beta.

9. Scenario 5: 1998 LTCM and Russia Default

20.Origination: Q1 1996. Stress period: Q3 to Q4 1998. Dominant factor: global liquidity shock with idiosyncratic spike (LTCM collapse, Russia default). This is a historical benchmark period for testing liquidity-driven tail events.

21.CDLI did not exist in 1998 (started 2004). We substitute Cliffwater US Private Credit Analog (pro-forma extension of CDLI using contemporaneous leveraged loan and HY returns). Analog peak drawdown: 3.2% over 2 quarters.

22.Peak live LTV 50%. 1 margin call. 0 realized loss. The LTCM episode shows the indicator framework catches liquidity-driven shocks even where the NAV impact is small, because factor convergence (HY OAS, swap spreads, Treasury-agency basis) exceeded the stress threshold.

Cross-scenario summary

23.The methodology produces consistent results across five regimes with five different dominant factors. No single factor decomposition drives the outcome. The proxy indicator system fires margin calls on the factor that is actually stressed, not on a preset trigger. This is the value of factor decomposition: the system adapts to the regime.

10. Tail Scenario: 50% NAV Decline

24.This scenario has no historical analog for private credit. It is modeled to test the first loss tranche and warehouse bank protection.

25.Assumption: pool NAV declines 50% simultaneously across all funds. No distribution recovery. Gate events at all funds. Secondary sale required.

26.The first loss tranche is sized to absorb the tail scenario. Warehouse bank protection in the 50% simultaneous decline scenario is preserved, with $4.9M of first loss remaining as residual buffer.

11. Reverse Stress Test

27.The reverse stress test asks: what combination of events produces a warehouse bank loss?

28.No historical period has produced all six conditions simultaneously. The 2008 crisis met two (NAV decline, recovery delay) but not the others. The breaking point requires a scenario materially worse than any observed in the modern private credit era.

12. Sensitivity Analysis

29.Each parameter is shifted and the GFC scenario is re-run to isolate its effect on the outcome.

30.The dynamic margin threshold and distribution sweep each contribute roughly 3 to 6 percentage points of peak live LTV relief. The advance rate set at origination is the dominant parameter: +10pp on the advance rate brings peak LTV within 4pp of the 75% liquidation cap, indicating the methodology is well calibrated and additional advance rate adds material risk.

13. Reconciliation to Public Benchmarks

31.To check that the simulation is plausible, the pool NAV trajectory is compared to the underlying benchmark indices for the same periods.

32.Our simulated pool drawdown of 8.4% sits within the observed range for private credit strategies (8.8 to 9.4%). The strategy mix (90% private credit, 10% PE) explains the slight difference from a pure CDLI trajectory.

14. Limitations

15. Next Steps

33.Once the facility is operational, the simulated backtest is replaced by actual pool performance. The quarterly reporting cycle described in the methodology (Section 19) includes borrowing base certificates, margin call activity, and realized loss data.

34.Year 1 to 2: the pool accumulates real loss and cure history. The backtest remains the reference document but is supplemented with live performance.

35.Year 3+: the live performance dataset becomes the primary validation artifact. The simulation is retained as an archived reference to confirm methodology consistency.

Related Publications