Daily Summary for 2026-10-27

daily

Transcript

Introduction to the show: ident: Quantitative Finance. The day in quantitative finance: models of markets, risk and prices.

Ruth: It's the twenty-seventh of October, twenty twenty-six, and this is the day's research.

Warren: 6 new papers came out today.

Ruth: I'm Ruth.

Warren: And I'm Warren, here with Iris.

Ruth: We'll take the day in one pass, then pull out the papers we're staying with.

Warren: Start us off with Simple Portfolios that Deliver Consistent Decent Returns.

The day's research: Ruth: Simple portfolios deliver consistent returns because investors lack short selling and leverage.

Warren: This paper addresses private investors facing those realistic constraints differently than most academic work.

Iris: The main finding is that simple portfolios from as few as three ETFs can achieve decent returns without constant tuning.

Ruth: This idea builds on general portfolio construction studies but focuses on liquid US ETFs and monthly trading costs.

Warren: When no return forecast is made, the portfolios use risk metrics like risk parity or minimum volatility derived from a statistical factor model.

Ruth: When a return forecast is included, researchers combine the Markowitz formulation with momentum signals and ridge regression.

Warren: How does that combination help identify added value?

Iris: This framework lets us backtest hundreds of these simple portfolios over twenty years or more on standard hardware.

Ruth: The empirical testing shows those proposed portfolios actually outperform the traditional sixty-twenty allocation.

Warren: That suggests the inclusion of those signals has a measurable benefit.

Iris: Another piece examines market structure by looking at supplier income response to liquidity in pooled lending markets.

Ruth: How does adding liquidity affect total interest income for a supplier?

Warren: It can sometimes reduce that income even as their balance grows because constraints shape outcomes.

Iris: This relationship mirrors how the choice of which ETFs to include shapes portfolio performance.

Ruth: We see work on pricing complex derivatives in incomplete markets using deep learning operators.

Warren: That suggests a different modeling complexity than simple asset allocation focus.

Iris: How does that deep learning scheme approximate those high-dimensional variational inequalities?

Ruth: It uses physics-informed neural networks to approximate them effectively.

Warren: Does this method account for transaction costs in the option pricing problems?

Iris: Yes, the operator splitting scheme incorporates transaction costs explicitly.

Ruth: That links it back to valuing American options in those incomplete markets.

Warren: And what about the research on catastrophe bonds using geometric Riemannian pricing algorithms?

Iris: That approach uses financial entropy to resolve ambiguity between frequency and severity risk premia.

Ruth: So, even with complex markets, entropy helps select a more meaningful risk measure?

Warren: It suggests a physically meaningful way to assess risk when dealing with complexity.

Iris: That contrasts sharply with the simple asset allocation focus of the first paper.

Ruth: Indeed, that distinction highlights different modeling focuses in this research day.

Warren: We also have work on optimal order placement under a state-dependent Hawkes process model.

Iris: What specific findings relate to terminal wealth compared to fixed-spread benchmarks?

Ruth: The study finds that optimal strategies can increase terminal wealth under certain inventory penalties.

Warren: That outcome depends heavily on the specific inventory penalty structure used in the model.

Iris: It shows how order placement decisions impact performance under those conditions.

Ruth: We also have research on contemporary transmission networks and systemic risk in European banking.

Warren: How does that network model identify institutions acting as transmitters or receivers of risk?

Iris: It maps how equity returns spread through banks, showing their roles in systemic risk propagation.

Ruth: That gives us a structural view of interconnectedness within the banking system.

Warren: So, the research covers derivatives, bonds, order placement, and systemic risk modeling.

Iris: Indeed, it shows a broad spectrum of financial complexity addressed today.

Ruth: These four papers are Simple Portfolios that Deliver Consistent Decent Returns for simple portfolio construction; Depositing More Should Not Earn Less for managing supplier income elasticity; A Deep Learning-Driven Operator Splitting Scheme for Free Boundary Problems to solve complex American option pricing problems with transaction costs and volatility; An Application of Geometric Riemannian Pricing Algorithm for Valuing Catastrophe Bonds to price catastrophe bonds by resolving ambiguity in frequency and severity risk premia; and Optimal Order Placement under a State-Dependent Hawkes Process Limit Order Book model to find optimal order placement strategies that increase terminal wealth compared to fixed-spread benchmarks.

Ruth: Simple Portfolios that Deliver Consistent Decent Returns is worth hearing for simple portfolio construction.

Warren: Depositing More Should Not Earn Less is worth hearing because it shows how supplier income can be managed.

Iris: A Deep Learning-Driven Operator Splitting Scheme for Free Boundary Problems is worth hearing to see how deep learning solves complex option pricing.

Ruth: An Application of Geometric Riemannian Pricing Algorithm for Valuing Catastrophe Bonds is worth hearing because it uses entropy to price catastrophe bonds.

Warren: Optimal Order Placement under a State-Dependent Hawkes Process Limit Order Book model is worth hearing to learn strategies that increase terminal wealth.

Iris: That concludes the review of today's research papers.

Ruth: These are the lucky papers for today.

Lucky paper: 2610.36491: Ruth: Alright, let's get into the details of Simple Portfolios that Deliver Consistent Decent Returns.

Warren: So, the core idea here is that private investors face constraints like no short selling or leverage, which is different from what much academic work looks at.

Iris: The paper suggests that a simple portfolio built from as few as three ETFs can achieve consistent and decent returns without needing to tune it to past performance.

Ruth: That’s the big claim—three ETFs instead of the usual sixty-twenty split, and no need for constant tuning.

Warren: But how do they build those portfolios when there's no return forecast involved?

Iris: When there is no forecast, the portfolios are constructed purely from risk alone, using methods like risk parity or minimum volatility derived from a statistical factor model fitted via expectation maximization.

Ruth: Risk parity sounds like it’s trying to balance the portfolio based on different levels of volatility.

Warren: Exactly. They use that statistical factor model to figure out those inputs for the optimization problem posed as convex optimization.

Iris: When they do include a return forecast, they combine the Markowitz formulation with momentum signals and ridge regression to see what adds value.

Ruth: So, it’s a hybrid approach: risk-based construction without forecasts, and then adding signals when you have some guidance.

Warren: And that framework is what allows them to backtest hundreds of these simple portfolios on a standard laptop over twenty years or more.

Iris: That empirical testing shows that these proposed portfolios outperform the standard sixty/forty allocation in their tests.

Ruth: Outperforming sixty-twenty over twenty years is a pretty strong result, especially given the constraints they are operating under.

Warren: That suggests that simply sticking to those three ETFs might be more robust than trying to perfectly time markets with complex models.

Iris: It's interesting how they weigh the gain against the complexity and effort demanded of the investor when combining those components.

Ruth: And what about those monthly trading costs they mentioned in their setup?

Warren: Those monthly trading costs are explicitly factored into their construction, so we can’t just ignore them.

Iris: They show what works and what each component adds, weighing its gain against its complexity and the effort it demands of the investor.

Ruth: So you're saying they aren't just suggesting a static list of assets, but a dynamic way to build that list based on risk metrics.

Warren: That’s right. It’s about building the portfolio structure itself using those specific optimization inputs derived from public data.

Lucky paper: 2610.36899: Ruth: We're moving on to the paper, Depositing More Should Not Earn Less: Supplier Income Monotonicity and Rate Curve Design in Pooled Lending Markets.

Warren: This one deals with utilization dependent interest rate models, specifically how adding liquidity affects a supplier's total interest income even as the supplier's balance increases.

Ruth: So if you add more liquidity, you might actually reduce the total interest income for the supplier?

Iris: Exactly, because when a supplier holds a fraction alpha of final pool supply, their income stays nondecreasing with a marginal deposit exactly when alpha times one plus eta is at most one where eta is the elasticity of the borrow rate with respect to utilization.

Warren: That condition sounds like it sets a threshold for how much liquidity matters for the supplier's income stability.

Ruth: And if that condition fails, what happens next?

Iris: When that condition fails, the supplier can instead raise their income by withdrawing, which pushes utilization further up.

Warren: That means the market structure forces a choice on the lender about what to protect at each utilization level.

Ruth: The paper says on two public drift markets, a lender held more of the pool than the curve protected in at least one hour in nine percent of SOL and thirteen percent of USDC.

Iris: That statistical finding shows that this structural choice happens frequently across different protocols.

Warren: So this choice turns into a design constraint for the protocol by choosing the largest supplier share it intends to protect at each utilization level.

Ruth: That choice effectively caps the elasticity, and integrating that cap describes every admissible curve for interest rates.

Iris: If you protect the same share at every utilization, that results in a power law boundary, but protecting every fresh depositor results in a hyperbolic one.

Warren: So the shape of that boundary depends entirely on whether you are protecting the existing pool or individual new deposits.

Ruth: This is interesting because it shows how income response is governed by the constraint governing how income responds to a deposit, not just the price level.

Iris: That distinction is key when thinking about compliant repairs in a multiple kink USDT market, where these constraints can either lower or raise supplier income.

Warren: So this paper highlights that the mechanism controlling interest response is tied to utilization limits rather than just market prices.

Lucky paper: 2610.37171: Ruth: Alright, we're moving on to paper number three of four today. We’re looking at An Application of Geometric Riemannian Pricing Algorithm for Valuing Catastrophe Bonds in Incomplete Markets: Market-Consistent Financial Entropy and the Split of Frequency and Severity Risk Premia.

Warren: That title sounds dense. It's about pricing catastrophe bonds in incomplete markets using geometric optimization and financial entropy.

Iris: The paper uses a maximum financial entropy problem on a finite-dimensional manifold of equivalent measures, parameterized by the non-hedgeable market prices of catastrophe frequency and severity risk.

Ruth: So the authors are treating the pricing measure selection as this optimization problem involving those two risk premia.

Warren: Since unconstrained entropy maximization recovers the physical measure, they incorporate observed secondary-market spreads, both pooled and date by date.

Iris: They use a compound Poisson-generalized Pareto loss model with Wang distortion to derive closed-form expressions for financial entropy and the Riemannian metric.

Ruth: The key result seems to be that CAT bond prices depend on frequency and severity premia primarily through the risk-neutral trigger intensity, which makes those two premia unidentifiable from market data alone.

Warren: How does that unidentifiability happen in practice?

Iris: The entropy criterion resolves this ambiguity by using a minimal-entropy measure, and when they calibrate it to a hypothetical earthquake CAT bond calibrated to quarterly spreads of Golden State Re II Ltd., they obtain risk-neutral trigger intensities of six point five eight to ten point four five times the physical intensity.

Ruth: That range seems like a concrete output from their calibration, but what about the comparison with market data?

Warren: The model-implied risk multiple aligns with the market spread-to-expected-loss ratio of eight point eight seven versus eight point eight one at issuance, which is pretty close.

Iris: But the paper also points out that the minimal-entropy measure attributes fifty-three point five percent to sixty-three point seven percent of the log risk multiple to frequency risk, reflecting selection principle rather than just market identification.

Ruth: So the entropy criterion is picking a measure based on selection principles, not just matching market data perfectly for both risks.

Warren: And what about the final comparison they made regarding coupons and spreads?

Iris: Equating the coupon with the quoted spread understates the risk multiple by approximately half, which is another important detail from that paper.

Lucky paper: 2610.37324: Ruth: Alright, let's take a look at this paper, Optimal Order Placement under a State-Dependent Hawkes Process Limit Order Book model.

Warren: This study looks at optimal quote placement for a market maker facing order flow modeled by a state-dependent Hawkes process.

Iris: A block-based linear demand specification links market-order counts to execution volumes, providing the demand moments required by a continuous time discrete state dynamic programming formulation.

Ruth: So they are using that to build this framework for quote placement. How does the inventory penalty fit into that dynamic programming setup?

Iris: They use a quadratic terminal inventory penalty which lets them characterize those inventory-dependent quotes and derive backward recursions for the coefficients of a quadratic value function.

Warren: A quadratic terminal penalty means the cost of holding more or less stock grows quadratically with the inventory level.

Ruth: That sounds like they are trying to find an optimal balance between quoting aggressively and controlling that inventory risk.

Iris: Exactly, and they ran numerical experiments using LOBSTER data to examine quote distributions, inventory exposure, and terminal wealth.

Warren: What did those numerical experiments actually show regarding the strategy?

Iris: The implemented strategy yields higher aggregate terminal wealth than fixed-spread benchmarks at one and three ticks.

Ruth: Higher than fixed spread at one and three ticks, but lower than the two-tick benchmark. That’s a specific comparison there.

Warren: So it performs better in those intermediate tick scenarios compared to the standard two-tick benchmark strategy.

Iris: That difference illustrates how the framework's numerical implementation works and distinguishes terminal wealth comparisons from evaluation under the inventory-penalized objective.

Ruth: It means that when you factor in that penalty, the optimal placement isn't just a simple spread adjustment.

Warren: If we look at it from an execution perspective, what does that mean for real-time deployment?

Iris: It means the model suggests adjusting quotes based on the immediate order flow pattern described by that Hawkes process.

Ruth: So, if you're running this live, you need to model that state-dependent process in real time.

Warren: Right, and what about the LOBSTER data they used? Was that representative of current market conditions?

Iris: The paper uses LOBSTER data for their numerical experiments, but they are focused on demonstrating the framework's capabilities.

Ruth: So, it’s a demonstration of the mathematical approach rather than necessarily a prediction of future market behavior.

Warren: That makes sense. We also have to consider if this framework is robust when you move away from LOBSTER data to live traffic.

Iris: The authors are showing that the framework can be applied, and they derived backward recursions for the coefficients of that quadratic value function which is a key part of the mechanism.

Ruth: So, the core contribution here seems to be linking order flow dynamics directly into a dynamic programming solution for inventory management.

Warren: That shifts the focus from static pricing models to dynamic, inventory-aware quoting strategies.

Iris: It does that by characterizing those quotes and deriving those recursions based on the demand moments from that process.

Ruth: For someone just listening, it means instead of setting a fixed price spread, you are trying to place orders strategically depending on what the market is doing right now.

Warren: That’s a big shift in thinking for market making operations.

Iris: The results show higher terminal wealth at one and three ticks compared to the fixed-spread benchmarks they tested against.

Ruth: So, while it underperformed the two-tick benchmark in that specific test, it showed superiority in other realistic tick scenarios.

Warren: It really highlights that optimization isn't just about maximizing profit on a single trade; it’s about managing inventory exposure over time.

Iris: The framework itself is designed to handle those inventory exposures by using the quadratic penalty term within the value function derivation.

Ruth: That penalty term is what forces the model to be sensitive to how much stock you hold, which is a key constraint for private investors too.

Warren: It’s interesting seeing how concepts from high-frequency trading modeling are applied here under those specific constraints.

Iris: The entire structure of the paper, Optimal Order Placement under a State-Dependent Hawkes Process Limit Order Book model, shows how dynamic programming handles these state transitions effectively.

More episodes

← Home