Stochastic Risk Modeling for CDP Defaults

Collateralized Debt Positions (CDPs) are the backbone of many DeFi lending and stablecoin systems, such as Sky’s USDS. They allow users to borrow against crypto collateral while keeping their exposure to the underlying asset. To protect solvency, protocols impose overcollateralization requirements: borrowers must lock collateral worth significantly more than the debt they issue.

The central question is: how likely is a CDP to default under realistic collateral dynamics? Instead of treating defaults as anecdotes, we can use stochastic processes to quantify liquidation probabilities, simulate tail events, and design safer parameters.

---

Collateral Dynamics

We model collateral price $S_t$ as a stochastic process. The simplest benchmark is geometric Brownian motion (GBM):

$$dS_t = \mu S_t , dt + \sigma S_t , dW_t,$$

where $\mu$ is expected drift, $\sigma$ volatility, and $W_t$ a standard Brownian motion.

If a user locks $Q$ units of collateral and borrows debt $D$, the collateralization ratio is

$\text{CR}_t = \frac{Q S_t}{D}.$

The loan is liquidated if $\text{CR}_t \leq c^*$, where $c^*$ is the liquidation threshold.

Thus, liquidation corresponds to the price hitting a barrier

$S_t \leq H = \frac{c^* D}{Q}.$

The default time is the first passage time:

$\tau = \inf \{ t > 0 \mid S_t \leq H \}.$

Our main object of interest is

$\mathbb{P}(\tau \leq T),$

the probability that liquidation occurs within horizon $T$.

---

Barrier Option Analogy

This is mathematically equivalent to a down-and-out barrier option problem. The probability that a log-normal asset crosses a barrier within $T$ has closed-form approximations.

For $S_0 > H$, the probability that the collateral hits the barrier before $T$ is

$\mathbb{P}(\min_{0 \leq t \leq T} S_t \leq H) =\Phi\left( \frac{-\ln(S_0/H) + (\mu - 0.5\sigma^2)T}{\sigma \sqrt{T}} \right)+ \left(\frac{H}{S_0}\right)^{2\lambda} \Phi\left( \frac{-\ln(S_0/H) - (\mu - 0.5\sigma^2)T}{\sigma \sqrt{T}} \right)$

where $\lambda = \frac{\mu}{\sigma^2} + 0.5$, and $\Phi$ is the standard normal CDF.

This lets us estimate the default probability directly from observed volatility $\sigma$, debt size $D$, and collateral ratio.

---

Monte Carlo Simulation

Closed-form results are elegant but fragile. Crypto markets exhibit fat tails, jumps, and volatility clustering that GBM cannot capture. Monte Carlo simulation provides flexibility:

1. Generate $N$ paths of $S_t$ under stochastic dynamics.

2. Record the first time each path hits the barrier $H$.

3. Estimate default probability as

$\hat{p} = \frac{1}{N} \sum_{i=1}^N 1_{\{\tau_i \leq T\}}.$

Simulations can incorporate:

• Jump-diffusion models (Merton or Kou) to capture sudden price crashes.

• Stochastic volatility models (Heston) to match clustered volatility.

• Correlated assets if multiple collaterals are accepted (ETH, WBTC, stETH).

This approach yields not only expected default probability but also the distribution of losses, enabling stress testing.

---

Risk Management Insights

From a protocol perspective, these models guide parameter optimization:

• Choosing liquidation thresholds $c^*$: too low, and insolvency risk is high; too high, and capital efficiency suffers.

• Dynamic parameter adjustment: thresholds or stability fees could adapt to realized volatility $\sigma$.

• Collateral onboarding: new assets can be stress-tested under stochastic models before being approved.

From an investor’s perspective, understanding the tail distribution of defaults informs risk premia. Quantifying default probabilities helps price that risk.

Conclusion

Liquidation risk in DeFi lending is not anecdotal—it can be quantified rigorously. By treating CDPs as barrier options on stochastic collateral processes, we can compute default probabilities, simulate extreme scenarios, and design safer parameters.

Stochastic modeling shifts the conversation from “will the protocol survive the next crash?” to “what is the probability of ruin under given parameters?”, a framing familiar from actuarial science and risk management.

As lending protocols mature, governance can increasingly rely on these quantitative tools to balance capital efficiency and systemic safety.