Types of Modelling - Statistical vs Discretionary

So this recently occurred to me while trying to apply statistical modelling to a problem (in my opinion) that was clearly not suited to it. Rather, my prior is that the problem could be solved via discretionary modelling, aka having a hypothesis or thesis about how a series of events in the future would play out. Read more...

Intuiting the Frisch-Waugh-Lovell (FWL) Theorem

In this article, I take a look at the Frisch-Waugh-Lovell (FWL) theorem, which lets us understand the interpretation of individual betas in a multiple regression after partialling out correlation effects, and lets us view the intercept as orthogonalizing the features with regard to a constant vector. Featuring stuff from Gregory Gundersen, RyxCommar. Read more...

🍁 Analyzing the Impact of the Trans Mountain Pipeline Expansion (TMX) on Crude Markets

In this article, I analyze an oil analyst (Big Orrin's) discretionary take on 'the most overhyped (upcoming) flop in the oil market': the Trans Mountain Pipeline, using his Twitter discourse and a freight analyst (Ed Finley Richardson's) articles/tweets. The aim is for TMX to be a gateway for WCS to Asian markets, improving egress & bringing in more dollars for Canada's economy. But some argue it may have the opposite effect due to the difficulty of the crude, the lack of demand and exorbitant freight costs to Asia, cheapening the marginal barrel. Read more...

🛢️Understanding Crude & Product Trade Flows: Visualising the BP Review of World Energy

Every year since 1952, BP publishes a report called the BP Statistical Review of World Energy. In it contains historic data on world energy markets: facts & figures on global energy markets from 1960 to 2022. I visualise and analyze the data via charts. My aim is to gain an intuitive geographic feel of global crude and product flows. Read more...

📊 Crypto Stat Arb I: Quantifying & Combining Alphas

I look at some alphas for a crypto stat arb strategy. We use momentum, carry and breakout - simple price action features. I evaluate how 'good' they are: looking their mean returns across deciles, the info coefficients and how they decay over time. Then I combine these with 'hardcoded' weights to see the hypothetical returns of trading these signals frictionlessly. Read more...

Paper Replication: High Frequency Trading in a Limit Order Book

Based off Avellaneda and Stoikov's 2006 paper, we simulate a market-making strategy that sets quotes off an indifference price: an offset to the midprice based on inventory. I calibrate the model in Python to sample L2 order book data. I learn basic limit order book mechanics and the problems of simulating a HFT strat. 📚 Read more...

Paper Replication: Principles of Principal Components

Based off Salomon Brother’s 2000 paper which showed how PCA on the yield curve can be used to weigh curve-neutral butterfly trades (theoretically) free of level and slope directional bias. I replicate the (no longer stationary, of course) PCA-weighted butterfly spreads on treasury yields from 2012-2022 in Python. ⚖️ Read more...

Thinking About Thinking (Metacognition)

In this post, I explore the notion that we can observe experts in different fields and distill their underlying patterns of thought via the right syntax in order to adapt the “optimal style of thought” for that field (e.g natural sciences, computation, humanities). Read more...