Roman Paolucci

People that don’t know what inflation is vote

Nothing like a 4am deployment!

I just realized telling an LLM “it’s not working” is roughly equivalent to “turning it off and turning it back on again” - maybe it’ll solve the problem, fingers crossed 😭

Oftentimes we forget the sheer number of folks that engage in a retail trading capacity. Regardless of the average outcome, probabilistically, some must accumulate extraordinary wealth - they may even think they have a well defined edge, but it’s all just a lucky random walk…

Much appreciated @montykarlo42 !

This week hinges on Jobs and Powell. The market is already expecting a cut, risk coming off the books early in the week, some profit taking, bad news is good news (maybe?) inflation concerns always looming, rates gotta come down eventually (right?) - what a week!

I’m looking forward to beginning Live Classes next week with the fall cohort, there are some open seats remaining if anyone has interest in joining! Working hard on a lot of exciting things for the channel and Quant Guild toward the winter - stay tuned!

Everyone has something to say about inflation with the PPI numbers, some use it to point fingers and play the blame game for media attention or politics: empty noise. Avoid this nonsense altogether, it is what it is - look at the entire chess board and start trading that edge

I have to acknowledge how big of a fan I am of Carr-Madan and their identifying of the relationship between the characteristic function of log prices and corresponding option prices via the Fourier transform. Such elegance and efficiency - expect a long video on this topic soon!

Everyone’s posting GPT-5 criticisms and gotchas… “8.11 - 8.33 > 0…” “This is not Ph.D. Level knowledge…” “…” Scientific criticisms are useful and encouraged, empty noise and commentary from folks who have no intention of moving the needle? Not too sure about that…

There is a place for ML/AI in finance, but rather than throwing the kitchen sink at your feature vector I’ve always preferred a more conservative approach where economic interpretation takes priority. For example, I’d like to include a vol feature, let me engineer one of those.

Calibration to a volatility surface may be trivial from the standpoint of problem formulation “just minimize the distance between model vol and market vol” but wow if you don’t have (quasi)closed form solutions is this impractical for efficient and consistent pricing

Asking someone to “teach you to trade” is like asking someone to “teach you to bat” they can show you what it means to swing at high probability pitches, or take some bigger risks to knock it out of the park - but you still gotta swing the bat

Check out my latest video on Itô integration clearly and VISUALLY explained!

Devoid of inside information, Wallstreet quants with Ph.Ds don’t know anything you can’t!

Never underestimate a quant’s ability to find edge in a qualitative capacity! Decomposing the structure in a quantitative way is useful, but the concept itself is so abstract - intuition from models, experience trading, opinions, all contribute to wealth generating abilities!

Stochastic integrals are random variables, sometimes the integrand is also a random variable - lots of moving parts (literally)! It’s impossible to discuss the “solution” (numerical or analytical) without Itô’s Lemma or mentioning SDEs. All of which are necessary for pricing!

When EV is a function of action, that’s where there is structure for a person, model, or agent to learn. That’s where we find edge. Think blackjack (card counting), poker, and of course, trading - unlike games of chance where EV is fixed, we can influence EV based on our actions!

The only thing a person or even a reinforcement learning agent can learn (or approximate) using data from a game of chance is the probability mass or density function. There is no “optimal” action in the sense of maximizing EV - action does not influence EV in that setting!

Is there a better way to represent that correlated high-dimensional data in a lower-dimensional subspace? Is a model that captures richer characteristics an extension of others that live in lower dimensional manifolds? This is how all of the pieces of the puzzle fit together!