Algorithm And Portfolio Stats: 01/27/2024 - 02/02/2024
Our final numbers are in for the month, and they’re slightly negative, but leave us room for optimism. If you’d allocated 1/72nd of your capital to every trade - the most conservative strategy, which guarantees never needing margin - you’d be down about 6.6 bps. Conversely, if you’d allocated 100% of your capital into every trade, using whatever margin was needed to achieve this, you’d be down about 5.26% this month. Depending on how you allocate your capital, you’d be somewhere between those 2.
For this system, that’s fairly close to the black. We’re obviously no happy with this result - every loss is bad - we’re optimistic about its ability to perform in the future. Let me elaborate.
First off, we pushed a few changes to this algorithm fairly late in the month. Of these, we can simulate 1 of them - ie, see how we would have performed this month if we’d made that change from the beginning. And that’s our new rule against taking positions near open. The current version of the algorithm won’t take new positions during the first 30 minutes of each trading day, since these contain a high proportion of fake-outs.
If this had been in place at the start of the month, instead of being down 5.26%, we’d be up 9.85%. That’s a major improvement - all from avoiding more fake outs for more of the month - but there’s more.
We’re also looking into avoiding holding positions overnight. Our system looks at 5 minute candles, and from the closing candle one day to the open of the next is the most common time we’ll see the biggest gaps. These can throw our technical analysis out of whack, inhibiting our ability to profit consistently. If we’d held no positions overnight, treating everything as a day trade, we’d be up 9.83%. If we both took no positions near open, and held nothing overnight, we’d be up 24.11%.
Obviously, this is all easy to see in hindsight. We’re making decisions about how to best trade when we already have all of the data. Furthermore, these all assume massive amounts of margin. A real trader would likely only make a small percentage of that, but in terms of consistency of profits, we would have seen a large return.
We’re going to take a few weeks to examine the data and run some more backtests. In the meantime, we’re going to keep running the system as is, but with the change preventing overnight holdings. We’ll also be doing our scheduled monthly reset. To those unaware, most of the model is dedicated to determining which tickers it wants to watch, and when one is eligible to be bought/shorted - execution is a relatively small part of this. Each month, we close out all open positions, and have the model choose a new set of tickers for itself to watch.
For the time being, the algorithm will continue running with only these changes in place. If any major updates are coming, we don’t expect to see them until mid-late February.
Now then, let’s examine the best and worst we’ve seen this month.
The best trade we made this month was this shot on APTV. We got into it shortly after activating the bot for the first time on January 3rd, at $85.15, and overnight the stock gapped down, right past our target exit of $80.77. Assuming we’d gotten out right at our exit, we would’ve seen a return of 5.14% here.
Our best long was similar, with the stock gapping up. Our model assumes that, in the event of the stock passing our stop loss price or target exit, we get out at exactly our price. In this case, we’re pricing in an exit of $39.06, for a return of 4.17%. In reality, we would have gotten out at more than $42, for a return of more than 12%!
Our worst trade of the entire month is this long on WBA. We took a loss of 2.15% here. That’s pretty tame relative to our big winners, but with the algorithm designed to be lower-risk, that’s to be expected.
In total, we took on 1044 positions this month: 505 long and 539 short, for an average of about 45 positions per day.
In other news, our portfolio had a great week! We beat SPY again, primarily off the strength of our Technology and Communication holdings. Well-done to Vlad, our head analyst! Our portfolio returned roughly 2.5% this week, with upside/downside capture ratios of 106.7% and 97.1% respectively, based on the overall market.
Lastly, for a quick macro update, the yield curve has gotten slightly flatter since the start of the year. It’s still fairly inverted, but this a good sign, albeit a small one.
That’s all I have for you tonight. Thanks for reading, and happy trading!
