Weekly Algorithm Review: 12/24/2022 to 12/30/2022
Algorithm Performance This Week
Overall Market: -0.13%
Sector Neutral: -0.55%
Market Neutral: -0.84%
Base Algorithm: -0.88%
Long Term Portfolio: -0.97%
Variable Market Neutral: -1.03%
Variable Sector Neutral: -1.23%
This week’s performance is interesting. It’s not good for us - but it’s interesting. First of all: we underperformed this week. The market beat us. The main for this is the long term portfolio underperforming. In terms of technicals, we did well enough. The base algorithm outperformed the long term portfolio - meaning that our technical analysis was a net positive for us this week.
In order to explain the lagging of the long term portfolio, let’s examine it’s makeup versus that of the S&P 500.
Our portfolio is defensive. As such, it over-allocates into Consumer Defensive, Healthcare, and slightly on Utilities. Consequently, it under-allocates into all other sectors. Looking at market performance by sector this week, we can see why we underperformed.
This graph is a bit crowded, but if we look at the sectors we over-allocated: Consumer Defensive, Healthcare, and Utilities, we can see all 3 of them near the bottom by the end of the week. This is also the reason our sector neutral algorithm outperformed the market neutral, when the opposite has recently been the case.
In summary: we underperformed this week because the long term portfolio underperformed. The long term portfolio underperformed because we chose to allocate into more defensive sectors, which performed poorly this week. We don’t see this as a red flag for our strategy - we’re still bearish for the near future. Further, the market has a tendency to perform abnormally around the holidays and end of the year.
I’d also like to take a moment to address our variable systems, as they also fell outside their normal range. Specifically, I want to address why this can happen.
The variable algorithms have a slight modeling difference compared to our non-variable algorithms.
Our base algorithm is a machine-learning system. It seeks to buy stocks that will perform well. Our market neutral algorithm seeks to buy stocks that will perform well - when hedged to achieve 0 market beta. And the sector neutral algorithm does the same, but hedged to achieve 0 beta to selected sector ETF’s.
The variable algorithms seek stocks that they expect to perform well, after some percentage of their market/sector betas have been hedged away. These percentages are determined by the responses we receive from our users. Since the training data they use each day has slightly different rewards for each day in each stock’s history, this can result in them deciding to allocate more or less into stocks they’ve chosen. This, for example, is why our sector neutral system made a profit trading ABT this week, but our variable sector neutral never allocated into it.
This is something I would consider changing. I can’t say for with my usual certainty how well it will achieve its goals, because we can’t backtest this training system. Well, not until we have a longer history of responses from our users.
All in all, this week was sub-par for us, but not a red flag. We remain optimistic about our algorithms’ future.
That’s all I have for you tonight. Thank you for reading, now let’s start 2023 off right.
