The last few months has been crazy. Driven by the need to get an A for both of my courses in order to graduate, my life assumed a certain relentless discipline. Work, home, school, studying in the evenings, studying over the weekends, hanging out with Shaune, exercising once in a while. It was just so long. A full 13 weeks of this schedule. Towards the end, I did feel extremely tired. I miss hanging out with my friends. I miss the feeling of just switching off your mind, turning off the need to constantly ask, “what’s next”. I miss going on purposeless meanders, to dwell, and to take luxurious indulgent breaks. To breathe slowly, and to soak up each moment of the day.
Beneath the grind, I have to admit that it was extremely fulfilling. I have covered the breadth of deep learning, a truly transformative technology of our times. Not only have I dipped my toes into understanding more advanced topics such as normalizing flows or graph neural networks or VAEs, I’ve also managed to - through gritting my teeth and slowing my breath - worked on the mathematical foundations essential for the understanding of machine learning. I’ve covered topics such as probability, expectations, calculus, and linear algebra. I’ve gained an insight into how to gain competence within those subjects. Crucially, I’ve also developed a system of taking mathematical notes, something that I’ve never done in my life (spoiler alert, it involves Obsidian). I’ve also created a site where I share the models that I’ve trained, host the models so that I can actually play around with them and see the results of their use, to witness the power of the models in the flesh. The course has sparked a further desire to really master those mathematical fundamentals, play with more models, and fully exploring what this technology is capable of. And also to share my journey with the world.
And on the other hand, I also had that algorithmic trading course, which, frankly speaking, despite being somewhat eye opening, was somewhat more of a let down. Let’s put it this way. After studying it for 13 weeks, I am still not convinced of this approach. To be convinced, you’d need a robust system of backtesting, and way more experience playing around with different strategies, documenting those strategies, and seeing what works and what doesn’t, and where. Just like the machine learning models, it requires way more time and energy spent experimenting and documenting your findings. Nothing quite work out of the box.
However, I do appreciate the instructor’s general advice on this topic. Some of the stuff he said was golden - for instance, whenever you run a strategy, the true gems of each run lies in how you conduct your analysis after that. I think what’s more important is how this paradigm of trading - of thinking about trades in terms of expected values and standard deviations - challenges me to think about trading/investing in a different light. Especially in light of what happened (something I’ll perhaps get into in a later post), I truly wonder whether my entire approach, or mental model, was wrong, or requires updating.
Briefly, where I erred since around 2019 was this: I was trying to wait for a recession to enter the markets. This was a 7 year wait. In 2021, I thought it was going to be the next year - it never came, and so in 2022, I thought it was going to be the next, and it never came, etc. There were a few flash corrections in between, during which I was waiting for the price to drop further. That never happened. The corrections practically rubber banded upwards.
The problem with waiting for so long is that you miss out on the opportunities for riding the wave when it’s already halfway up, which could be significant gains. The other more acute conundrum is that the more you wait, the more it makes sense for you to keep waiting, because the probabilities of a correction increases with time.
From the perspective of probability, if I am awaiting a 40% market drawdown before entering, then the probability of it being a random quarter is probably 0.1% or something (I’ll need to go do the math for this). Which means most of the time, I am not doing anything. Which means much of my currently predicament is expected behavior.
Expensive and cheap, I realized, is relative. Things might look expensive now, but if it’s only going to be more expensive in the future, then currently, it’s cheap. So you see - it’s some twisted logic, but there’s logic to that.
In any case, I’ve digressed. My era of school-work busyness is officially over. I can take my saturday mornings off typing this post, recollecting my thoughts, whilst being under no pressure to do this thing or that. I operate best and enjoy what I do best when I operate in this way. These modes of being can also be surprisingly productive; but I recognize that it can also lack consistency and focus. For instance, I doubt I would have been able to get to this level of competence with deep learning if not for the structure that the course provides. My body and soul reacts against the structure, but the structure keeps me accountable to my own goals. So bless be to structure. Bless be to the school.