Memorizing a few linear stochastic matrix expressions is pretty useful.

But what about higher order formulas?
Things quickly get weird...
🧵1/4

a cool article about backtest:

have you ever wondered why practitioners run regression as a pseudo-backtest? (one) reason is backtest ≈ predictive regression. this is only 1 page, the full book is a must read IMO.

Inefficiently Efficient - Lasse Heje Pedersen

Solving linear systems comes in different flavors. en.wikipedia.org/wiki/Least_squ…

You've trained a machine learning model.

Now you take one of the features and increase its value beyond its maximum range.

What will the model predict? Let's find ou

How do I test a strategy without any portfolio optimization or even forecasting?

Let's talk about a quick and dirty method for testing alphas.

A thread 🧵

[1/N]

A question to my #CardioTwitter friends.
Now that we now that a significant share of ASCVD is attributable to Lp(a), which is (nearly) environment independent, don't we need to revise the mean and, especially, variance, of past computed risk factors?

On March 12, the Biden Admin finally announced a new Drawdown package for 🇺🇦, ending an 11 week moratorium.

For 11 weeks they maintained they would not provide any new aid. They watched as ammo ran out. They watched as Avdiivka fell. Then they reversed course.

Why?

1/109

🧵

If you've not got many features, it is rare that you'll use anything more than a regression to model it.

Why?

You can express all non-linearity in the feature if you want. Just add a mask to it.

If I think there is a log relationship between my feature and it's alpha, then…

What this means is something you know INTIMATELY:

As time goes by, the probability of been far from your starting point ( wealth, position, etc), increases..

The is a section in his thesis: PROBABILITY IS A FUNCTION OF TIME.

Something I am finally grasping from Bachelier’s work:

-The distribution can remain the same.
-The first moment can remain the same.

While the probabilities evolve with time.

The square root of time (connected to the scale of the distribution) also changes the probabilities.

This is just science - go for a walk.

This paper found walking (whether outdoors or a treadmill) increased key types of creative thinking for over 80% of undergraduates. The reasons are not fully clear, but there seem to be direct effects on the brain apa.org/pubs/journals/…

McKinsey's bizarre regressions of executive diversity against firm-level financial performance are not only potentially reverse-causal, they don't replicate in newer data.

🧵Effect of commission on the minimum win rate required for a trading method or strategy to break even.

In my book Profitability and Systematic Trading (Wiley, 2008), which is free to download (link at the end), I derived a formula for the win rate, which I called profitability…

Notice how no one ever claimed that they were an “R programmer”, just an “R user”.

Luke Burgis Gurwinder wrote a brilliant piece about this phenomena.

Rather than drawing from an unknown source, they are all responding to the desires of the audience which seeks suck drivel. The 'influencer' is influenced - and deeply so.
gurwinder.blog/p/the-perils-o…

Not getting personally attached to your mental models is a superpower. Don’t be too trigger happy to change them, but don’t hesitate to do so when the overwhelming new information warrants.

Stanley's probability textbook (probability4datascience.com) is fantastic, so I imagine this tutorial on diffusion is good as well. However, I am no longer of a fan of presenting the discrete time version - the flow matching interpretation seems more elegant.

I often teach about Decorators in Python.

Many know how to use them, but few can write them.

These are tricky because nested functions make our brains hurt.

Here are some hints for grokking them.

1/

One way that an inflated ego often manifests itself: a priest, a founder or a CEO, a teacher, truly believes that they can't step away from their company or their parish or their school for a week-long retreat without everything falling apart in their absence.

Secretly, they…

I often get asked which LLM library I use for my experiments, side projects, and to help build solutions for customers. I actually use the one I help develop, LitGPT: github.com/Lightning-AI/l…
It's focused on performance and ease of use but is also designed to be easily extendable…