Independent quantitative research
A quantitative research lab for volatility. We study how it is measured, how it is modelled, and how far ahead it can honestly be forecast — across equity, rates, FX and commodity markets. Everything we learn gets published, in writing and on video.
01 — What we work on
Volatility is the most forecastable thing in finance and still routinely forecast badly. Most of the damage happens before any model is fitted: in how the target is measured and how the result is evaluated.
Realized variance, realized kernels, microstructure noise, jump separation. You cannot forecast a quantity you have not defined properly.
Smile, skew and term structure; arbitrage-free parameterisation; what the surface implies about the risk-neutral distribution.
Horizon by horizon, with honest out-of-sample protocol, sensible loss functions and significance testing rather than a single flattering backtest.
Clustering, long memory, roughness and structural breaks — why persistence estimated in calm markets fails exactly when it matters.
02 — Model zoo
Four families, each answering a different question. Nothing here is a silver bullet; most of the skill is in knowing which family the problem belongs to.
Discrete time, daily returns
Intraday data, observable target
Pricing, surfaces, roughness
Where it helps, and where it does not
03 — Notes
Working notes, written to be read by someone who wants the reasoning and not just the conclusion.
ARCH begat GARCH, and GARCH begat a zoo. What each extension is actually fixing — asymmetry, persistence, long memory — and how to tell which one your data is asking for.
Corsi's three-lag regression is embarrassingly hard to beat. We set up an honest horse race against it and look at where the added complexity of rough volatility and deep learning earns its keep.
Trees cannot extrapolate, and volatility spikes are extrapolation. A practical account of feature design, target transforms and the leakage traps that make backtests look far better than they are.
04 — Video
We publish explainers on @volatility_insights — a ladder that starts from nothing and ends at the current research frontier. Every episode is built so the previous rung is the only prerequisite.
05 — Contact
Collaboration, data, a question about a note, or a topic you want covered on the channel — all welcome.
Volatility