Hi YXI friends,

This email is to share some of my thoughts on the direction of the YX Insights service.

We have recently upgraded our machine learning models to focus on the market regime changes rather than very short-term forecasting. This aims to reduce position turnover (and costs), decrease the “noise” from frequent signal changes, and capture the “meat” of a broader move. This was done at no expense to the theoretical Sharpe ratios, and should, in fact, improve them over time.

The second upgrade we introduced was a strategy overlay on top of our models to enable optional leverage and short strategies. You can read the full details here.

I will review our upgraded model results and strategies near the end of Q2, allowing at least a full quarter to play out before making any tweaks. In reality, a quarter is fewer than 63 trading days, which are not statistically significant either. But at least we can catch some glaring mistakes, without overfitting to the recent data (the most common pitfall of algorithmic strategies).

I do have plans to expand our ticker list for this service, which will further enrich the value of our service offering.

Specifically, I plan to split our service into two offerings: Macro and Equities.

Equities as a standalone service

At the moment, I am slightly split-minded on what the equities offering should include. There are two approaches:

We currently cover more than 30 charts a day with commentary. It is a long and exhausting list. I understand this newsletter format clearly works and is very well received (every time I suggest changes, there is a lot of resistance against it).

However, there are several difficulties I face with the current format:

I hold my work to a very high standard, and while the existing format may be satisfactory to everyone already (allowing me to “coast”), I believe there is much more potential.

Moreover, I do want to spend more time conducting research to deepen my understanding of the names we cover and expand our reach. That, in turn, improves the quality of our service in the long run.

Of course, not all is bad with our approach. By looking at everything every day and writing about it, I am developing a pretty strong intuitive sense of what the market is doing. I have felt calmer and much more confident in my analysis.

In the background, I am working on a new set of models that project further into the future than our existing ML Signals. I will share them in due course - they likely further uplift the quality of our analysis.

At the same time, technical analysis will take a back seat. We currently rely on them everyday in our analysis, but I want to treat them simply as another toolkit in our box. Our edge comes from the combination of analytical approaches we readily deploy, which allows us to adapt to the market quickly.

Moreover, modelling and quantitative methods are far more objective, measurable, and trackable, which allows us to improve them tangibly over time.

We still run our quarterly earnings updates. They are still useful in helping understand a stock’s longer-term trajectory.

I would think about the time horizon vs techniques as per below

As we split our Daily Signals into two camps, our daily briefing will do the same. As a result, my current plan is to rotate between them each day:

The intention is to free up half the space each day to focus on additional observations and more diversified methods.

If you have any ideas or feedback, please click on the poll below. It will take you to a comment page to share your thoughts with me privately.

It could be about which equity ticker list to use, the macro vs equities split, or the Daily Briefing content.

I may say no to many of the requests when designing the updated service. Please do not take it personally. I will also do a lot of trial and error during the discovery process. Many things will no doubt fail, but that is life.