Algorithmic Week In Review: 09/11/2022
Recent Performance - Friday
The Market: +0.96%
Long Term Portfolio: +0.57%
Base Algorithm: +0.31%
Experimental Sector Neutral: -0.02%
Sector Neutral: -0.13%
Market Neutral: -0.24%
Experimental Market Neutral: -0.28%
Friday was more in-line with what we usually expect to see when trading with our conservative portfolio. All in all, last week was a tough one for us.
Tuesday’s results were particularly erratic, but since that was just the result of a change we pushed (and have since reverted), I don’t think it’s worth further examination.
Wednesday was much more on-track for us. Only 1 of our 5 systems underperformed its baseline - and only by 0.03%.
Thursday, similar to Friday, we saw our base algorithm underperform the long term portfolio by a decent stretch.
I’m not sure what to think about this. With Tuesday a wash, and markets closed last Monday, we only have 3 days of effective data from last week. Of those 3, we have 2 solid underperformances from the base, and 1 approximate match from it. It’s possible we just got unlucky twice, and equally possible that our base algorithm is making the choice to play more conservatively. With the markets deep red the week prior, I wouldn’t blame it for this.
Regardless, a loss is a loss. If this pattern continues into this week, I’ll be looking to retraining it.
New Developments
We’re primarily looking into how we process technical signals. This is similar to a previous project - in which we used different formulas to calculate the “strength” of each technical signal - and it is. This time, we’re looking at more advanced ways to determine this.
We also have one new hedging system we’d like to test, but we’re mostly happy with the ones in-place, given that its an overall market crash we’re worried about.
A step further from sector neutrality. Can we choose individual stocks that best represent the risk in a single stock? This will vary from ticker to ticker, and determining it mathematically will be difficult. For example, hedging AAPL with MSFT seems like a safe bet, but how would we hedge companies with fewer comparables?
Amalgamated technical signals. What would happen if we gave each signal “bonus points” for the performance of similar signals? My concern here is about putting all of our eggs into one basket. I suspect this could result in us leaning into the same few signals too hard, and getting a more volatile portfolio as a result.
Evolving our portfolios over time. As is, each day’s recommendations are independent of all previous days. What if instead, our algorithms recommended and adjusted allocation of each ticker based on the previous day’s allocation? If nothing else, our holdings would be more consistent. Would this allow us to benefit from repeated pro-signals, or hinder our ability to adapt to changing conditions?
Tomorrow’s Outlook
The full algorithm reports will be published tomorrow morning, once Allen has had a chance to vet its recommendations. In the meantime, here are our tentative exposures for the trading day tomorrow:
I want to take a moment and wish everyone good luck this week. Let’s all make some money and be ready to celebrate.
-Asher
