Thread by @HangukQuant

Sub-Thread 3

The Research Process

The Alpha Leak

“Vertical Alpha Research”, A Thought Process

📌 The Research Process

1)
Type 1: To keep my knowledge of finance, I both subscribe to financial literature, academic or otherwise. That means reading books/papers on finance, trading, economics, podcasts for general knowledge, and a working knowledge “Mathematical Finance”.

2)
Type 2: My knowledge on Computer Science and Statistical methods comes from my background in academia. I also keep up to date on new state-of-the-art research by reading academic literature.

3) However, I would say that this learning process is heavily skewed in favor of post-grad science degrees. Despite it, one can still make remarkable advances in this area with consistent effort.

4) One of my favorite follows is @QuantSymplectic , who regularly posts non-academic reading/literature that can help widely boost appreciation for modern computational methods.

📌 The Alpha Leak

5) Let’s look at a sample alpha using a Hybrid (Type 3) approach. In my alpha research process, analyzing price information led me to believe that daily return windows of varying historical length and its relationships contain information about market pricing.

6) It seemed to suggest that co-movement of the windows was negatively priced.

📌 “Vertical Alpha Research”, A Thought Process

7) A thought process from market pricing factors (Type 1):
One of the market’s pricing mechanisms is a measure of risk, or uncertainty.

8) The fundamental idea behind risk premia is that traders are provided excess returns for taking on risk, as in the fire insurance example. Another stylized fact of returns is that stock returns have low autocorrelative properties.

9) If you fit a simple linear model of today’s returns on yesterday’s returns, r(t) = b + m * r(t-1) + e, we find that cor(r(t), r(t-1)) is low (in a SLM, the fit = cor^2), and therefore today’s returns provide little predictive power about tomorrow’s profitability.

10) This is FIN/STAT101. EMH says that “returns today are a summary of all available information”, and r(t) is a measure of all such information. Now, it is true that r(t) is a weak predictor for tomorrow, but what if some of the assets are more “predictable” than others?

11) The autocorrelations are sure to be cross-sectionally unstable, indicating that for some assets, information today “has better predictive power”.

12) If today’s information has better predictive power, then those assets with less useful information today should be more uncertain; the market should price this in and award excess returns to those who assume the risk of holding such assets.

13) MAIN THREAD

https://twitter.com/hangukquant/status/1445464869587746818