How Much is “Too Much Information?”
QiT has data going back to 2007 for all the QiT portfolios.
You did know that we have data going back that far, didn’t you? No? Well heck, check it out. On each performance page you’ll see a link that says, “Click here for the Monthly Profit, Trade Count and Trade List tables” where you will find all your monthly profit percentages back to 2007, monthly trade counts back to 2010 and a list of the last month’s trades so you can check your trades to the strategies trades.
Then you have “Click here for Year to Date and Since launch charts” where you will find how the strategy has done since the beginning of time – just kidding, since the beginning of the year and an equity curve that starts at the portfolio’s launch date.
Now to get to the topic of this week’s newsletter, just how far back should a developer backtest and why does QiT not go back to 2000? It’s like the question, which type of strategy is best, mean reversion or momentum? The answer is, it depends on the developer, the strategy and what you are trying to backtest.
My answer, however, is two-fold. First, data all the way back to 2000 is just TOO MUCH INFORMATION (TMI) and would be overwhelming for most. It is called data overload and you would get lost in it. Of this, I guarantee because I’ve done it – many times.
The second reason, the most important, is how relevant the years from 2000 to 2006 really are to today’s trading environment. I have a constant dialogue with my algo contractor on how far back you should go to backtest and we both agree you need to have a time frame that covers bull markets, bear markets, high volatility and low volatility. Once you have those environments covered anything else is just noise.
You need data for all environments
The ten years from 2007 to 2017 encompasses all four environments. Certainly a bear market from 2007 to 2009. Then an awesome bull market since. We had very high volatility in 2008 – DUH! And very low volatility in 2014 – long time QiT members will remember that year with anything but fondness.
Another point is those two years skew the outcome upwards and I like to have the metrics on the website reflect reality as close as possible.
This is why we use only data back to 2007 but with the caveat, be careful if you take into account the years 2007 – 2009 because the data has been skewed by 2008. Personally, I only use data back to 2010 for my own trading.
There is another argument you can make for using only 2010 and that has to do with frequency of trading. There is no day trading algo developer who would use minute charts beyond 2010 or probably even 2014 because of the frequency of trading. But that is a discussion for another day.
Michael Lewis, who authored the iconic books, Moneyball, and the Big Short released a new book in December, The Undoing Project. Here is a synopsis of the book,
“Forty years ago, Israeli psychologists Daniel Kahneman and Amos Tversky wrote a series of breathtakingly original studies undoing our assumptions about the decision-making process. Their papers showed the ways in which the human mind erred, systematically, when forced to make judgments about uncertain situations. Their work created the field of behavioral economics, revolutionized Big Data studies, advanced evidence-based medicine, led to a new approach to government regulation and made much of Michael Lewis’s own work possible. Kahneman and Tversky are more responsible than anybody for the powerful trend to mistrust human intuition and defer to algorithms.”
I have it read it and it’s excellent
We’re the Plan in “Plan your Trade and Trade your Plan”
Check our out the Performance Matrix for our algorithms and sign up today.