The day in conversation for 2026-11-18

discussion

Transcript

The discussion: Ruth: The idea that we can recover a risky position by evaluating increments using a dynamic risk measure rho is interesting, but I keep coming back to how we formally define that recovery process when the underlying market dynamics are non-Markovian, which is where I see my blind spot. Warren: When you mention non-Markovian dynamics, are you referring to the jump processes we modeled in the earlier simulations or the continuous diffusion models? Because if rho needs to account for those jumps explicitly rather than just relying on local volatility assumptions, that changes how we calculate the effective drift term entirely. Iris: I read the section on BSDEs in that paper; it frames rho as a worst-case conditional expectation, which ties back neatly to our earlier work on path-dependent hedging strategies. But what specifically about the convexity requirement for rho is providing a tangible advantage over, say, using standard Conditional Value at Risk measures applied incrementally? Ruth: The convexity ensures that the resulting drift function remains tractable within the framework of stochastic control theory; without that, solving for the worst-case expectation becomes analytically intractable because the set of possible future outcomes doesn't maintain a simple lower bound property. Warren: That tractability is key. But if we are applying this to portfolio selection, say in that second paper regarding disappointment aversion,

More episodes

← Home