Hello, so it is now December 1st. I'd like to recap November 2020 and share how my quant trading journey is going.
In regards to live trading, the system has been running since the end of September. In November it made $401.97 USD. I haven't made any changes to the strategy so far. It's still just been running since deployment.
In regards to strategy research, I have published a new strategy to the QuantConnect Strategy Library. This strategy uses Gaussian Naive Bayes (GNB) classifiers to forecast the upcoming price movements of the technology sector. You can read all about it on the Gaussian Naive Bayes model tutorial page. The tutorial reviews the theory behind the strategy, walks through how to implement the strategy with the QuantConnect backtesting engine, and discusses the results.
Here is the equity curve of backtesting the strategy over the last 5 years:
This equity curve out-performs the S&P 500 index ETF, SPY. It has a greater Sharpe ratio over the entire backtest, during the 2020 stock market crash, but not during the 2020 stock market recovery. All of the performance statistics are available on the tutorial page.
I also delivered a presentation for the co-op department at my university during November. For this, I reviewed the internship I have at QuantConnect. It was a lot of fun to review my progress over the last 7 months and see all of the different things I've been working on during that time. Doing this presentation improved my presentation skills and gave me an opportunity to network with other students.
Of the people shown above, the three on the right are continuing with the team this year to compete again in February. During November, I did a little preparations for the upcoming competition. I discussed with my team, reviewed some of the case files from last year, and tested my practice server.
This year, the competition is going to be remote. It'll be interesting to see how Rotman deals with the latency of the trading data. If all of the traders are scattered throughout the world, some may get information sooner than others.
This is the equity curve of the strategy that has been backtested over 13 years. It achieved a 1.859 Sharpe ratio.
Menno Dreischor provided some very insightful comments, explaining how the strategy is likely overfit because of the large number of optimized parameters. Menno explained how applying the same strategy to a cross-validated training set leads to much poorer performance, suggesting the original strategy is over-fit.
I recommmend reviewing this forum thread. It highlights the importance of avoiding over-fitting by utilizing cross-validation.
Kevin pointed out that you should focus on the inputs to your trading business, not the outputs. He defined the outputs as the return or drawdown a trading system is experiencing. To a certain degree, you can't really control those outcomes. What you can control are the inputs that you put into your trading business, meaning the number of hours you spend researching strategies, reading books, or listening to podcasts.
This got me thinking about my schedule and how I'm going to allocate my time during December. Here are some of the tasks I'll be spending my time on:
To approach this, I'm going to use Google Calendars. I went through the month of December ahead of time and allocated my time to these activities.
To assist with this boost of productivity, I created a Strategy Pipeline spreadsheet (blurred out).
This is a list of strategies that I intend to test out in the future. I've been collecting all of these strategies for a couple years now. It's convenient having them all organized in one place. I organized it so the ones with the highest priority are at the top of the list. This is something I recommend everyone does. It's just a simple file containing the name of the strategy, links to some sources, and a short description.
Anyways, that's it for this month. I'll provide another update after December.