Motivation & Scope
Curve recently approved the expansion of the YieldBasis (YB) credit line to 1 bn crvUSD. YieldBasis scaling its credit line materially increases the size and frequency of flows it can route through PegKeeper pools, potentially amplifying stress during adverse market conditions.
This analysis aims to determine whether the system can safely accommodate such scaling. We examine how crvUSD supply expands and contracts conditional on peg deviations, attributing observed behaviour directly to PegKeepers, mint-market actions, liquidations, and YieldBasis trades routed through PegKeeper pools.
We focus on four questions relevant for a 1 bn rollout:
- Which mechanisms actually defend the peg when it is stressed?
- Is YieldBasis structurally peg-supporting or peg-destabilising, conditional on the peg?
- When YieldBasis flow occurs, is it the sole stressor on the PegKeeper, or are there additional flows?
- Is the monitoring framework for safe credit line expansion proposed by YieldBasis sufficient?
The results are used to (i) validate the proposed liquidity-anchored monitoring framework for YieldBasis supply, and (ii) derive practical guidance on PegKeeper liquidity requirements needed to support a safe, staged expansion toward higher credit limits.
Acknowledgments: We would like to thank Pangea for providing the raw data used in this analysis. We also appreciate their support in facilitating access to the underlying datasets.
Analysis
Supply Contracting and Expanding Actions Conditional on Peg
We analyze crvUSD peg dynamics between 2025-07-24 and 2025-11-30. For each block timestamp т, we compute the instantaneous peg deviation \text{pegdev}_t = P_t - 1, expressed in basis points (bps). We then record all observable on-chain actions that directly affect the circulating supply of crvUSD at that timestamp.
The actions considered are:
- PegKeeper liquidity provision and withdrawal,
- mint-market borrowing and repayment,
- liquidations and
- YieldBasis trades restricted to PegKeeper pools (buying or selling crvUSD).
No temporal aggregation or smoothing is applied; each observation reflects contemporaneous system behavior.
Each action is mapped to a signed supply effect based solely on its mechanical impact on crvUSD circulation: actions that add supply (PegKeeper provision, minting, YieldBasis selling) are assigned positive values, while actions that remove supply (PegKeeper withdrawal, repayment, liquidation, YieldBasis buying) are assigned negative values.
We plot peg deviation against signed action magnitude in a single four-quadrant scatter plot, with points colored by action type. This provides a non-parametric view of which actions occur, and in which direction, conditional on the peg being above or below parity. All figures use winsorized axis limits based on empirical quantiles to reduce the influence of extreme outliers.
For convenience and easier interpretation, we have added the following table below and will review each action separately.
| Quadrant | Peg deviation | Flow | Meaning |
|---|---|---|---|
| Top-left | Below peg | Positive | Undesirable response - system expanding supply |
| Bottom-left | Below peg | Negative | Desirable response - system shrinking supply |
| Top-right | Above peg | Positive | Desirable response - system expanding supply |
| Bottom-right | Above peg | Negative | Undesirable response - system shrinking supply |
Review of actions:
- PK Provide: PegKeeper operates desirably by expanding the system supply when above peg.
- PK Withdraw: PegKeeper operates desirably by shrinking the system supply when below peg.
- Mint: Mint actions are not directly peg supporting, with many mints occurring when the peg is less than 1 dollar. However, deviations are minor when minted in size. Given that mint, repay, and liquidation actions are behavioural, they may occur with a lag. In addition, 5-10bps is very small considering the overall average cost of leverage on volatile collateral types.
- Repay: Repay actions largely occur when the peg is below 1 dollar, removing crvUSD supply from circulation. This is desirable.
- Liquidation: Liquidation actions largely occur when the peg is below 1 dollar and are negative in nature. Hence, they exhibit desirable behaviour.
- YB Buys: Buys of crvUSD associated with YB in PegKeeper pools largely occur below peg, putting upward pressure on the peg. This is a desirable property.
- YB Sells: Sells of crvUSD associated with YB in PegKeeper pools also largely occur below peg, putting additional downward pressure on the peg. This is undesirable for peg stability.
The culprit in the crvUSD system design with YB seems to originate from YB sells during the observation period when crvUSD is below peg, which are not particularly large but happen frequently.
Delineation into Pre- and Post-YB
We delineate the analysis into pre- and post-YB. We mark the introduction of YB at the timestamp 2025-09-24 16:12:59+00:00.
The pre-YB period experienced less market sources of stress, which must be kept in mind when interpreting this chart or making any inference that the peg has been more volatile due to YB. Yet overall, we can see that the system reacts directionally in a desirable way.
Post-YB, all PegKeeper and mint actions are more dispersed, reflecting greater crvUSD system stressors. Mint actions seem at times directionally undesirable. However, they are low in size as the peg deviation turns more negative. Only YB sell actions are additive, showing undesirable behaviour.
Overall, we observe that when the peg was stressed, YB actions showed partially peg-supporting actions with YB buys and partially destabilised actions with YB sells.
Limitation: The analysis is pointwise and does not assume instantaneous reactions; observed actions may reflect lagged responses to prior peg deviations due to execution delays, smoothed inputs, or internal state constraints. Consequently, the figures are interpreted as conditional associations rather than causal or synchronous responses.
Flow of PegKeeper Stable Reserves
PegKeeper stable reserves are defined as the sum of non-crvUSD stablecoin balances held in PegKeeper pools (USDC, USDT, PYUSD, frxUSD). These reserves represent the inventory capacity available to PegKeepers for peg defence via Withdraw and Provide actions against crvUSD.
For this, we enhanced our dataset by merging reserve data of the respective PegKeeper pools, restricting the analysis to the common coverage window from 2025-07-24 01:28:23 UTC to 2025-11-18 00:00:47 UTC.
We then construct a bivariate state space with peg deviation (bps) on the x-axis and PegKeeper stable reserves on the y-axis. Observations are split into pre- and post-YieldBasis regimes. The space is discretized using hexagonal bins; each bin is coloured by an aggregation metric count, which indicates the frequency spent in this given state.
Before YieldBasis, the system remained tightly clustered around the peg with high and stable PegKeeper reserves. After YieldBasis, the state space broadened substantially: larger peg deviations coincide with a much wider range of PegKeeper reserve levels, including repeated visits to low-reserve states. The interesting part is that below the peg, PegKeeper reserves deplete non-linearly, indicating that downward pegs increasingly deplete stable reserve consumption.
Note: The mechanical intuition here is that when crvUSD trades below peg, getting the PegKeeper pool to a balanced state with respect to its paired stablecoins requires either removing crvUSD from the pool or supplying more of the paired stablecoins.
This likely explains the previously introduced figure, suggesting that YB sell actions are damaging to peg stability. In particular, the sale of crvUSD consumes PegKeeper stable reserves. Given that we observe a similar pattern in the left side of the chart (pre-YB), the fundamental dynamics are likely not introduced by YieldBasis.
Digging Deeper: Stable Reserve Changes to YB
We construct changes in total Pegkeeper stable reserves. YieldBasis activity is measured as net crvUSD flows associated with Pegkeeper Pools over the same intervals.
To focus on economically meaningful interactions, we restrict the analysis to an economically meaningful regime, defined by events where both PK reserve changes and YB net flows exceed nominal thresholds (|ΔPK| ≥ 10,000; |YB| ≥ 1,000). This conditioning isolates periods where both mechanisms are actively adjusting.
Within this active regime, we estimate a robust linear model (Huber RLM) of YB net flows on PK stable reserve changes. Robust estimation mitigates the influence of large outliers while preserving slope identification from economically relevant variation.
=== RLM summary (active regime) ===
Robust linear Model Regression Results
==============================================================================
Dep. Variable: yb_m No. Observations: 7745
Model: RLM Df Residuals: 7743
Method: IRLS Df Model: 1
Norm: HuberT
Scale Est.: mad
Cov Type: H1
Date: Wed, 17 Dec 2025
Time: 11:54:13
No. Iterations: 50
==============================================================================
coef std err z P>|z| [0.025 0.975]
------------------------------------------------------------------------------
const 0.0001 8.67e-06 12.036 0.000 8.74e-05 0.000
pk_m 0.9981 4.61e-05 2.16e+04 0.000 0.998 0.998
==============================================================================
If the model instance has been used for another fit with different fit parameters,
then the fit options might not be the correct ones anymore.
In economically active periods, this relationship is visible directly in the left panel, where each point represents a time interval with both YieldBasis (YB) flows and PegKeeper (PK) reserve changes. The tight alignment of points along the 45-degree diagonal shows that increases or decreases in PK stable reserves are almost exactly matched by YB net flows in the same direction and of similar magnitude. Visually, this indicates that large PK reserve movements rarely occur without a corresponding YB flow, implying that YB activity explains the dominant share of PK reserve changes when both mechanisms are active.
The middle panel examines the difference between YB flows and PK reserve changes across these intervals. The sharp concentration of mass around zero indicates that mismatches between the two are generally small and infrequent relative to the size of active events. This visual evidence suggests that while minor discrepancies exist - due to timing, aggregation, or isolated non-YB actions - they do not materially affect the overall one-for-one relationship seen in the left panel.
Finally, the right panel plots the remaining PK reserve changes after accounting for YB flows against peg deviation. The absence of any visible slope or structure, with points clustered around zero across the full range of peg deviations, shows that once YB activity is removed, there is no systematic peg-dependent pattern left.
Taken together, the three panels indicate that during economically active periods (i.e. Pegkeeper flows and YB flows are present), YieldBasis routing is the primary mechanism driving large PegKeeper stable reserve adjustments, consistent with the mechanical effect of routing stablecoin flows into PegKeeper pools, up to timing noise and occasional non-YB interventions.
Strain of YieldBasis on crvUSD
To assess how YieldBasis (YB) activity strains the crvUSD peg, we focus on misaligned YB flows - defined as buying crvUSD when the peg is above 1 and selling crvUSD when the peg is below 1. These actions mechanically move supply in the wrong direction relative to peg restoration and therefore represent direct sources of peg stress.
We construct histograms summarizing the distribution of misaligned YieldBasis flow sizes under different peg conditions. The horizontal axis shows flow magnitude, measured as log1p(strain in crvUSD) to place small and large flows on a common scale, while the vertical axis reports their frequency. The red dashed line denotes the median flow size for each distribution.
We first examine the distribution of misaligned YB trade sizes across successive credit-line expansion regimes. The figure below shows that both the central tendency and the tails of wrong-way YB buys and sells increase as available credit expands. This indicates that larger credit lines do not merely increase activity frequency, but also allow materially larger adverse flows to occur.
The accompanying table summarizes this behaviour numerically. For each credit-line regime, it reports (i) the frequency of misaligned YB events and (ii) the expected wrong-way strain per bar. Together, these metrics show that higher credit limits are associated with both more frequent and more severe misaligned flows, increasing the realized burden placed on peg defence mechanisms.
We next condition on peg deviation severity. Expressing wrong-way flows in basis points of peg deviation, we find that extreme (p95) misaligned YB activity rises as the peg moves further from parity and volatility increases.
We relate these observations to broader market conditions. The strong co-movement between BTC volatility and the p95 magnitude of misaligned YB flows suggests a shared external driver: periods of elevated market volatility simultaneously worsen peg instability and increase the likelihood and size of adverse YB trades.
Conclusion
The first three questions we set out to answer in the Motivation & Scope are answered directly through empirical and mechanistic analysis. The fourth, which evaluates the cap framework, depends on these results and is therefore addressed as a synthesis of the preceding findings rather than as a standalone analysis:
1. Who defends the peg under stress?
Across the full sample, PegKeepers remain the primary and most consistently peg-aligned mechanism. Withdrawals below peg and provisions above peg dominate corrective supply adjustments. Mint-market repays, and liquidations also contribute in the correct direction but appear behaviourally lagged and smaller in magnitude. Peg defence is therefore structurally anchored in PegKeeper inventory capacity.
2. Is YieldBasis peg-supporting or destabilising?
YieldBasis is directionally mixed. YB buys below the peg are clearly peg-supporting and remove circulating crvUSD. However, YB sells below peg are frequent and this mechanically expands the circulating supply at precisely the wrong time. YieldBasis seems, therefore, to operate on factors independent from, and often in opposition to, crvUSD peg stability.
3. When PegKeeper reserves change materially, is YieldBasis the dominant driver?
In economically active periods - where both PegKeeper reserve changes and YB flows are non-trivial - YieldBasis explains nearly all observed PegKeeper stable reserve movements. Reserve changes and YB net flows match almost one-for-one, with little residual dependence on peg deviation. This indicates that YB is the dominant marginal stressor on PegKeeper inventories when both are active, rather than one of many comparable sources.
YB Scaling Recommendation
We now translate the empirical findings into operational guidance for safely scaling the YieldBasis credit line, and to answer the fourth question we set out to answer in the in Motivation & Scope.
Evaluation of Proposed Monitoring Framework
Michael Egorov put forward a proposal and methodology for scaling the YB credit line safely. A key element of this framework is the interpretation of pool imbalance as a stock constraint, rather than a marginal absorption margin. Empirical observations suggest that crvUSD/stable pools remain functional even at imbalances of approximately 70/30, implying that up to 70-75% of pool TVL can be safely held as crvUSD in total. This quantity represents a tolerable crvUSD inventory level, not incremental headroom from a balanced 50/50 state. Separately, historical YieldBasis pool data indicate that extreme YB imbalances (≈70/30) translate into net crvUSD sell pressure of roughly 40% of YB TVL. By combining these two empirical bounds, he produces a supply cap proportional to 2.5×0.75 x PegKeeper TVL.
The methodology has several assumptions. We highlight the main assumption we deem relevant below:
A1. YieldBasis is the dominant (or sole) source of crvUSD pressure
Assessment: In our analysis, we review reserve changes greater than 10_000 and YB_flows greater than 1000. The analysis suggests that when both coincide, YB is the primary driver for the reserve change. This supports the assumption.
Suggestion: Regardless, we suggest reviewing if this continues to be true. If YB, as well as additional flow, hit the PegKeeper pools simultaneously, cap measures may be underestimating risk.
A2. The historical worst imbalance for YB pools is representative of future imbalance and hence peg pressure
Assessment: Given the short history, yet unusually volatile market conditions, we see this as a sufficiently conservative assumption.
Suggestion: We recommend dynamically monitoring this imbalance as well as the overall cap framework and updating the YB cap recommendation accordingly.
A3. PegKeeper pools TVL is balanced and therefore a valid proxy for usable liquidity
Assessment: We see the need to account for the relative balance within PegKeeper pools when stress from YB occurs. Our analysis highlights that YB sell actions are directionally misaligned: YB sells remove the paired stablecoin and add crvUSD to the PegKeeper pool.
The chart below aggregates PegKeeper pool balances and YieldBasis trades into 15-minute intervals. PegKeeper imbalance is defined as ((\text{crvUSD} - \text{stable}) / (\text{crvUSD} + \text{stable})). YieldBasis buy and sell flows into PegKeeper pools are summed per interval, filtered to exclude trades below 1,000 crvUSD, and plotted to assess directional pressure relative to pool balance.
We observe that a ratio > 0 (indicating more crvUSD in the pool than the paired stablecoin) repeatedly reaches levels of ~0.6 over the observed timeframe, with the aggregated imbalance reaching as high as 30/70.
To explore a longer timeframe, we study the USDC PegKeeper pool, which has reached an imbalance of 70/30 several times since May 2023.
We therefore recommend, instead of using the more relaxed assumption of 0.75, to stick with the empirically more conservative “safe” imbalance level of 70/30, translating to a factor of 0.7. In addition, we implicitly assume PegKeeper pools are evaluated from a balanced pool state of 50/50 and measure additional absorbable capacity from that starting point. Empirically - as shown in the previous charts - we observe that pools frequently operate away from a balanced state, highlighting that this assumption may not hold. A more consistent adjustment is therefore to make the YB cap dependent on PegKeeper imbalance, reflecting remaining headroom rather than total ex-ante capacity.
Suggestion: To make this concrete, we suggest extending the current formula:
S_{max} = m_{yb} \times f_{safe\_crvusd\_{fraction}} \times tvl_{pk}
where
- m_{yb} = YB imbalance multiplier (e.g. 2.5)
- f_{safe\_crvusd\_{fraction}} = fraction crvUSD the aggregated PegKeeper can safely hold during extreme imbalance (e.g. fixed factor of 0.75)
- tvl_{pk} = PegKeeper pool total value
with (1) a more conservative “safe” imbalance scenario based on the empirically observed aggregated imbalance of 70/30, translating into a factor of 0.7; and (2) an adjustment of the cap by the absorbable capacity based on the current pool state. This can be expressed as:
S_{\max}(t) = m_{yb}\Bigl[f_{safe\_crvusd\_{fraction}}-\max\bigl(0, f_{current\_crvusd\_fraction}(t)-0.5\bigr)\Bigr]tvl_{pk}(t)
where
- \max\bigl(0, f_{current\_crvusd\_fraction}(t)-0.5\bigr) represents the deviation of the (EMA-smoothed) crvUSD fraction from a balanced pool, bounded at zero to penalize only crvUSD-heavy pool imbalance.
Backtesting this revised framework using an EMA with a half-life of 12 observations (≈3 days, given the sampling frequency) results in a reduction in the recommended YB cap.
Further Recommended Interventions
Beyond the monitoring framework, we see three elements crucial for overall health for the co-existence of YB and crvUSD.
Liquidity of PegKeeper Pools
We saw from our analysis that PegKeeper YB sell flows are especially misaligned with the peg. YB imbalances are, by design, short-term. Mechanically, a YB sell transaction adds crvUSD to the pools and removes the respective paired stablecoin. During a stress state, when PegKeeper reserves are depleted, no immediate defence is available. Such a state will result in a borrow rate spike, incentivizing mint market repays. Hence, deeper liquidity in PegKeeper pools is essential to manage short-term instability and preserve the crvUSD peg.
We see two logical interventions:
- Further incentivise liquidity in PegKeeper pools: This will provide more capacity to deal with short-term YB flows. It is important to maintain liquidity levels. If liquidity depth degrades, action must be taken to reduce YB caps.
- Additional PegKeeper pools: In addition, we suggest adding additional strategic PegKeeper pools that may bring unincentivised or protocol-owned liquidity (i.e., durable liquidity) into the crvUSD system.
Deleveraging Mechanism for YB
We want to see an active mechanism where the YB credit line can be contracted if liquidity or pool balance conditions change. As of now, no active mechanism is available apart from deposit caps. Yet, currently, there is no obvious means to encourage deleveraging of YB users and contraction of YB credit lines to safe exposure levels.
Growth of crvUSD
Mint markets act as a Peg defence backstop. First steps have been taken to make crvUSD more attractive to borrowers by EMA smoothing borrow rates according to PegKeeper debt levels. Robust mint markets remain essential to the crvUSD system’s health, and we suggest adding additional collateral types. We notice that mint market repays often act in a peg-aligned manner. Hence, a larger crvUSD market will introduce more balancing pressures. In addition, a large mint market will reduce the existential threat to crvUSD. This is because mint markets remain the most direct revenue driver for crvUSD (creating demand for the stablecoin via scrvUSD), and can directly defend the peg when PegKeeper debt is depleted.
Again, we see two primary interventions:
- Competitive rates on existing markets: We see this as an ongoing effort with our existing proposal to introduce smoothing adjustment of the Monetary Policy lever to provide a more competitive rate.
- Additional mint markets: In addition, we believe there is a need for additional mint markets. This enables use cases where leverage is, ideally, independent of market cycles to avoid contraction in bearish conditions and attract size.
Proposed Next Steps
We would like that YieldBasis:
- Define and document a deleveraging mechanism that can be activated if liquidity deteriorates. This mechanism does not need to be fully automatic, but should provide a clear, protocol- or governance-level pathway to contract outstanding exposure when conditions warrant a reduction.
- Incorporate PegKeeper pool state into credit-line monitoring by conditioning the cap on pool imbalance in PegKeeper pools. Alternatively, LlamaRisk can independently operate and publish this monitoring framework in the near term.
LlamaRisk will:
- Identify and evaluate additional mint markets that can act as structural backstops to peg stability. We will publish suitable assets on the Governance Forum with the respective analysis.
- Assess opportunities for additional PegKeeper pools that introduce more stable or protocol-aligned liquidity into the crvUSD system. We will put forward suitable pools on the Governance Forum with the respective analysis.
We are supportive of YieldBasis and view its success as an important component of the crvUSD ecosystem. The steps above are intended to strengthen system resilience as YieldBasis scales, ensuring that growth remains aligned with PegKeeper capacity and peg stability under stress.