Q4 Update & Automated Portfolio Managers
We're Gearing Up For Live Test #2
Important: There is no guarantee that ATS strategies will have the same performance in the future. I use backtests and forward tests to compare historical strategy performance. Backtests are based on historical data, not real-time data so the results shared are hypothetical, not real. Forward tests are based on live data, however, they use a simulated account. Any success I have with live trading is untypical. Trading futures is extremely risky. You should only use risk capital to fund live futures accounts and if you do trade live, be prepared to lose your entire account. There are no guarantees that any performance you see here will continue in the future. I recommend using ATS strategies in simulated trading until you/we find the holy grail of trading strategy. This is strictly for learning purposes.
I’m on a quest for the holy grail of automated trading. Questions? Check the FAQs or feel free to reach out directly:
AutomatedTradingStrategies@protonmail.com.
A lot has happened over the last three weeks so I thought I’d give a quick update. Not only am I changing VPS providers (you can read more about that here), but I’m trying to apply some of the portfolio management techniques we’ve discovered over the last six months to the Q4 Forward Test. I am most excited about a group of portfolio management methods that achieved 1,785% improvement over baseline.
I’m still working on adding strategies to the Incubator. The hope was to finalize the process in Q4, but now the priority has shifted to building the best Automated Portfolio Managers.
Why the shift? HMM.
HMM showed that there was a way to make informed decisions about which strategies to trade in a portfolio that yielded 1,785% improvement. And the results are so impressive that I decided to push the original time line out a bit so I can get these Portfolio Managers up and running. A walk-forward analysis on historical data is one thing, but I want to see if this really works on live data.
These are truly exciting times!
Automated Portfolio Managers
I’ve been hyped on the idea of creating Agentic and/or Automated Portfolio Managers (APMs) for a few months now, but couldn’t figure out the best approach. The first challenge was creating the “brain”, the second was designing the most efficient process or architecture —the “body”.
After many iterations of different set-ups and countless hours spent building, testing and rebuilding set-ups, I finally found a system that works. It’s easy to scale and less complicated than any other architecture. What’s more, it allows the APM to update in real-time which is important.
Why is this important?
The walk forward analysis that I conducted—the one that created the 1,785% improvement over baseline—was based on an update frequency of 50 trades. That means the model updated thresholds every 50 trades. It’s important to note that you can go a month without updates and still achieve a measurable improvement over baseline, but if you can figure out a way to pull it off in real-time without the use of an API or some other universal configuration file, it will not only improve performance and accuracy, but save time.
What else improves performance and efficiency?
I’ve been looking at various HMM feature sets, some are novel (% pullback), others not so much (Sharpe ratio). I’ve also been exploring the use of more or less states. So far seven is the sweet spot for the forward test (using 1 year of historical trade data). There are also additional risk management applications that can increase the confidence level of trades. I am currently testing the use of expected value in this way.
Here are some other design decisions to be mindful of: