Weekly Algorithm Review: 10/07/2023 to 10/13/2023
Performance Rankings
Variable Market Neutral: +1.03%
Market Neutral: +0.98%
Base Algorithm: +0.41%
Sector Neutral: +0.11%
Long Term Portfolio: +0.08%
Overall Market: +0.01%
Variable Sector Neutral: -0.04%
Our first week with the new algorithm is good, but not conclusively impressive. It initially looks great, with an out-performance of 33 bps, but that was mostly carried by Monday, where it won by 27 bps. And on Monday, we were still using the previous algorithm. That leaves us with a win of 6 bps from the following 4 days. That’s certainly good - a win is a win - but you see why I’m not impressed at this time. Our first 4 days with the new algorithm were, all things considered, rather slow. Despite the market (and consequently, the portfolio) being somewhat erratic this week, the algorithm’s returns in excess of the portfolio were extremely tame. The difference consistently hovered around +/- 10 bps per day.
The portfolio also did deceptively well this week. It out-performed the overall market (market-cap weighted) by 7 bps. A small win, though a win. But that’s the number I’m focusing on.
Looking at only the new portfolio (and only Tuesday-Friday; we were still using the old portfolio on Monday) we see an out-performance of roughly 15 bps. But it gets better. If we compare our portfolio to an equal-weighted S&P 500 ETF, we get an out-performance of roughly 36 bps.
As mentioned previously, we’re using the equal-weighted S&P 500 as the benchmark for the long term portfolio’s performance. The algorithm doesn’t have a good way to account for recommended allocations in the tickers we feed it, so we construct our portfolio under the assumption of equal weighting to account for this. This is contrary to the S&P 500, which is traditionally weighted by market cap.
We will continue publishing both numbers, because we understand that many traders prefer the market-weighted S&P as a benchmark, but internally, we are gauging the portfolio’s success by its performance relative to the equal-weighted market etf.
I’ve generated these graphs based on Tuesday - Friday, the dates where we were using the new algorithm. It surprises me how close the algorithm and portfolio were during this time, since these graphs do show significant differences. The algorithm’s distribution, as we often see, is much flatter, with a few key winners and losers doing the bulk of the work. We successfully dodged much of the loss from NFLX, as well as some of the portfolio’s other big losers. On the other hand, most of the average winners paled in comparison to NEE and LLY, which really carried the algorithm this week.
This is largely speculative on my part, but given the significance of allocation differences: I don’t expect the algorithm’s performance to stay this close to the portfolio’s indefinitely. I suspect that we’ll look back on this 4 day period as an outlier, as more volatility for it is coming. At the same time, if it keeps bringing positive returns with this level of consistency, I’ll be fairly happy with it.
All in all, this is a fine start to the beta test, but not a particularly impressive one. I’m going to want to see stronger performance, or at least equally consistent performance in the next ~2 weeks before deciding if it’s time to move to a full release. If the algorithm maintains this level of consistency, earning 7 bps for every 4 trading days we run it, we’re looking at roughly 4.5% returns in excess of the portfolio per year. Not a major profit, but if it’s consistent enough, I can be more than happy with that.
To summarize: it’s been a fine week to start the beta, but I wouldn’t get excited just yet. Every win is great to see, but I’m not confident in it yet.
What’s In The Pipeline
Research into the main algorithm is ongoing, but I expect to pivot some attention towards the new intraday bot during this week. In the long run, the consistency of returns we can get trading 5-minute or even 1-minute candles will dwarf what we can get trading daily candles. The shorter term candles will of course add their own obstacles, but I’m confident in saying that there’s a lot of potential here.
We’ll send out updates on intraday development as they come. At this time, we aren’t able to give an estimated release date for any new intraday algorithms or bots.
Misc. Data For The Week
