#221: Muscle Memory

micksonmatch
January 29, 2025

S2N Spotlight

I have run 5 marathons, including an ultra. People speak of muscle memory. If you haven’t heard of the term, it refers to the body’s ability to quickly relearn or regain specific motor skills and strength after a period of inactivity.

Yesterday I decided to get my fat ass into gear and go for a run. I have done no exercise in a month and almost none in a year. Shame on me. I ran 2.5 km at a brisk 6 minutes a km, not bad for 108 kg of quivering passion. I can officially say that my muscles have no memory, as today I am as stiff as I was on the dance floor at my bar mitzvah. Double figures before the end of the year, here we come.

Speaking of memory, I can remember the moment when I was trading full-time for an investment bank in 2001 and I came up with the theory that financial time series have memory. I thought I had made a huge discovery. You see, if you went through a classic finance education, you would have been grounded in the belief that the markets work according to the Efficient Market Hypothesis, which would suggest that the market’s next price move is random, as all available information is in the current price, and therefore the future is unknowable. Trends be damned.

It seemed clear to me that the price action of yesterday had an effect on today. It turns out price memory is a well-known effect with a number of different statistical ways to measure it. For today we will focus on the most popular, the autocorrelation.

Autocorrelation measures the correlation of the same time series with lags. So if today is Wednesday, then what is the correlation of the S&P 500 today with the S&P 500 on Tuesday and Monday?

Tip: Don’t use price; use the return. The larger the price, the higher the autocorrelation. It is kinda of obvious. If Wednesday is 6020 and Tuesday is 6000, then naturally those numbers are highly correlated, but it is a distortion. With return, you start at 0% each day and then produce a return, so it is a cleaner and more correct way to measure the effect of memory. I hope I am making sense.

As I know I am speaking to a sophisticated audience, I thought I should apply more precision than I usually do, so I am using 99% confidence bands. What you should be looking out for is whether any lag points peep over their confidence bands. I am looking at a very long time series here of 100 years, and sadly the market seems to have a case of Alzheimer’s. Very little, if any, memory.

You should not give up so quickly. If you are a true observer of the market, you will see memory; you just need to know how and when to look. When looking at the last 5 years, there is clear memory in the market.

The one takeaway I would like to leave you with is that the secret to understanding the market is figuring out what market regime we are in. Do that, and you have discovered the keys to unlock the door to a very rich kingdom.

S2N Observations

I am running around the city in and out of meetings today; I have lots to comment on but sadly no time to comment.

I will share this one obsession of mine. A few years ago I became obsessed with writing some code that would run like this in a terminal. I had no need for it, just some weird obsession that I would dream about night after night. I scratched the itch in a number of ways over the years but never quite curing the itch. More recently I have been scratching again, which has sucked a number of hours out of my day. Either I need to see a psychiatrist or a dermatologist.