Weekly Algorithm Review: 09/16/2023 to 09/22/2023
Performance Rankings
Sector Neutral: +0.72%
Variable Sector Neutral: -0.38%
Market Neutral: -1.18%
Variable Market Neutral: -1.45%
Long Term Portfolio: -1.63%
Base Algorithm: -1.8%
Overall Market: -2.51%
This was another strong week for the portfolio. We go from a win of 127 bps to a win of 88 bps. Not as much, but still a major success for the portfolio.
On the other hand, the algorithm did poorly, but much better than last week. It went from an under-performance of 90 bps, to just 17. That’s obviously still not a good performance, but it’s not the major failure we saw previously.
As expected with performance that close, there aren’t a lot of standout tickers that explain it here. We managed to avoid some big losses on EOG, FCX, and TGT for instance, but at the same time we missed out on some fairly large potential wins. We only allocated into UNH once this week, and missed out on most of its gains as a consequence. Interestingly, the algorithm did fairly well with trading MPC, one of the common high allocations on our reports. This one was profitable for us off the strength of our technical analysis, rather than the success of the MPC stock itself.
The biggest change the algorithm made in terms of its sector distribution was an under-allocation into Utilities and Consumer Defensive, most of which went into Financial Services. With Utilities and Consumer Staples doing well this week, and Financial Services under-performing, this helps to explain the slight under-performance of the algorithm overall this week.
You may have guessed, but we will not be releasing the hold at this time. As of late, the algorithm has performed poorly, outside of our expected bounds for it. Research is ongoing on improvements to the system, but I wouldn’t expect anything soon. Even if we find a version we like today, we’ll want to paper trade it for several weeks before we release it to users.
If I wanted to trade our systems right now, I would not hold anything the algorithm outputs. Instead, I would simply hold the long term portfolio. This portfolio represents our fundamental analysis of each company, as well as our overall market outlook. Even if the markets don’t currently favor our technical analysis, we believe in these aspects of our strategy.
What’s In The Pipeline
At this time, work is ongoing with testing updated versions of the algorithm. We have an internal version now that reduces the amount of trend-chasing it does, to positive results within our development data. Future tests will likely concern reworking how trading signals are calculated, and how we determine which ones to use. I determined the original calculations when I was still new to most of them, in addition to when I was less experienced with trading algorithms as a whole. I suspect that, with a stronger understanding of how traders use some of the more advanced signals at our disposal, I can give the algorithm more useful data to work with. In general, I would expect development to follow this pattern: updating older parts of the algorithm first. These will be larger tests, taking longer than the relatively minor changes to trend chasing, but they have the potential to yield much greater improvements.
Additionally, we are not releasing an updated intraday bot at this time. Earlier this week I was doing a few final deep dives to find any problems and, unfortunately, I found one that we couldn’t ignore. The bot had, to that point, been significantly less profitable than we believed, and we aren’t comfortable releasing it in its current state. I know this comes as unpleasant news - I don’t enjoy delivering it - but integrity is our #1 priority at HaiKhuu. We will always delay the release of a new system, rather than roll something out we don’t believe in.
With focus primarily on the daily algorithm at this time, I don’t have the extra man hours needed to create a new model to live test for the intraday bot. But there’s something I can do to progress this. There are a few things that I theorize might be the problem with the intraday bot we’re testing - since it backtests fairly well. In the near future, I’ll be making these tweaks to the intraday system, and testing it with these changes.
Given more time, I would prefer to try a new model at this point. However, this will allow us to make progress on intraday development while keeping our focus on the main algorithm. I apologize for the delay in the release of the updated intraday bot, and I hope that, when the time comes, we’ll release something truly special.
Misc. Data For The Week
