Algorithmic Week In Review: 11/06/2022
Recent Performance - Friday
Long Term Portfolio: -0.02%
The Market: -0.17%
Market Neutral: -0.32%
Variable Market Neutral: -0.35%
Base Algorithm: -0.38%
Variable Sector Neutral: -0.47%
Sector Neutral: -0.55%
Friday was, again, a bad day for technicals. This continued a pattern we saw at the end of last week. Following a major shakeup day, technical signals in general have been shocked and will need some time to readjust. We’ve been allocating into the LTP on our pilot account, and will likely continue to do so. This isn’t out of the ordinary, and is something we’ve seen many times before. There’s not much we can do to force technicals to come back into style, we just have to wait them out.
Current Projects - Live Trading Bot
You might have noticed the HK trading bot was significantly more active last week. That’s because we’ve expanded it. Instead of only including tickers in our long term portfolio, we now include the entire S&P 500 in its signals.
This has introduced some new challenges. If we keep every ticker in the bot, it will simply send too many messages, to the point of being unusable. So we have to filter out less desirable trades - but how do we categorize a trade as less desirable?
The first thing we can examine is the risk/reward ratio, that being the ratio of maximum loss when stopped out to the target return. Allen tells me that you all typically ignore trade alerts with a r/r greater than 1 (where the max loss is greater than the target return), so we’ve filtered those out.
The next thing is restricting messages during market open. Right after open, it’s easy to get tied up find yourself busy. At that point, you aren’t looking for new trades to make. This works out, because the algo finds many more buy and sell signals during the first ~15 minutes of market open. By suppressing these, the we can further reduce the number of undesirable trade alerts.
Our next step is to optimize both of these. <1 r/r and only after 15 minutes open are fine enough rules, but can we optimize these parameters to increase profitability? These parameters were selected to suppress trade alerts that are inconvenient to trade, but I want to suppress trade alerts that are less likely to be profitable.
I have a comprehensive backtest of our bot running now. In the interest of full disclosure, we are double dipping into our training data here. This test is looking at all trading data since March 1 of this year, which the bot has already been trained on. I’m looking to see if time of day and r/r significantly affect expected profit. Maybe a higher percentage of trade alerts in the afternoon are profitable? Maybe high r/r are a better bet? These results will take some time, but when we get them, we might adjust the bot. If so, this will yield higher expected returns from trading with it.
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:
That’s all for tonight. Thank you for reading, and best of luck this week.
-Asher
