In the arsenal of sales tools the finance industry uses, the backtest is surely one of the most paradoxical.
Investopedia defines backtesting as “the process of testing a trading strategy on relevant historical data to ensure its viability before the trader risks any actual capital.” Essentially, it’s a simulation of the past with the added benefit of hindsight bias. Jason Zweig in The Devil’s Financial Dictionary, is more acerbic, if not accurate, with his definition:
Backtest, v. and n. To comb through financial databases to determine which investing strategy would have worked the best if anyone had known about them at the time. Many asset managers then use the backtest as a way to extract money from clients in the present—and disappoint them in the future.
Backtesting is placing the bet after the race. This is, perhaps, a little unfair as backtesting has the potential to yield interesting, if not always useful, insights. Still, much depends on the quality of the backtest and, particularly, its timeframe. The self-selective nature of backtests means that one tends to hear a lot more about investing strategies that worked under certain parameters and specific timeframes than those that didn’t. In this way, the backtest offers little in the way of learning for the new investor.1
Might then I propose the concept of forward-testing, wherein you take a past prediction and see how it actually performed.2
This allows you to put yourself in the head of an actual investor. You can contextual their mindset and sometime gauge their reasoning because, at the time they were as limited in their ability to predict the future as you are in the present.
With a backtest, the model or strategy is formulated in the present and you test backward. With a forward-test, the investment prediction is one that was made in the past and you measure how it’s performed since then. The difference is subtle, but I suspect meaningful because it inherently forces its users to recognize the future’s opaqueness.
This idea was born when I read a piece on NPR’s website, titled “3 Investment Gurus Share Their Portfolios.” The article, which compared model portfolios designed by David Swenson, Gretchen Tai and Jack Bogle, offered a clear view inside the minds of investing legends. What it didn’t do was offer a chance to look at how these model portfolios would perform in the real world. Given that I “discovered” this article more than two years after it was published, it afforded me the perfect opportunity to forward-test.
I opened up an excel spreadsheet and started inputting each of the three models with an imaginary investment of $100,000. I selected the equities the article mentioned and figured out how they performed with the help of Morningstar.
Here’s what I learned:
- Jack Bogle’s simple two-fund strategy—which, for my age bracket meant a 70/30 stock-bond split—performed best. It grew the most over the two and a half year-long period; a $100,000 investment made on October 17, 2015 (the date the article was published) would have yielded a gross return of $128,405.3 It also had the lowest expense ratio of the three model portfolios when expense ratios were weighted based on fund allocations. As per Ben Carlson, simpler is often better.
- All of three portfolios performed well. The bull market essentially rewarded Bogle’s the most since it was more heavily weighted in stocks. It’s likely that in other market conditions Swenson’s and Tai’s portfolios would have outperformed Bogle’s. The biggest failure would have been not to invest at all.
- Forward-testing is time-consuming. It is no wonder most leave this type of thing to professionals. However, with the help of Excel, a calculator, and the Morningstar website, it is amazing how much one can DIY it.
- Satisficing gets the job done.4 My spreadsheet wasn’t perfect. While I noted expense ratios, I didn’t remove those costs. Also, I don’t think I took into account dividends; I’m still learning.5 That being said, I’m confident that I still learned which portfolio performed best and why.
- Two and half years is not long a long time. Sure, Bogle’s portfolio performed admirably well, but given a longer time horizon (with its attendant ups and downs) any of the other two portfolios might of come out on top.
- These portfolios are investor-agnostic. They fail to take into account any of the investors essential attributes among which I count age, risk-tolerance, time horizon, etc. With the exception of Jack Bogle’s portfolio, which considers age, these portfolios are blind to any individuals needs. In ignoring this reality, these portfolios set their investors up for failure.6
I found conceptualizing forward-testing invigorating and using it stimulating. It’s certainly a concept I’ll revisit. If you’d like me to forward-test a past prediction to see how it actually performed, please write: firstname.lastname@example.org. Or better yet, let me know if you’ve encountered or created any examples yourself.
1And since past performances is not evidence of future potential returns, the backtest doesn’t purport to offer much more for the experienced investor either.
2I recognize that there are other, more generally-accepted definitions of the term forward-testing, like these ones, but felt my take was of a similar enough spirit to co-op the term.
3An annualized return of roughly 10.52%.
4More on the magic of satisficing in a future post.
5If you’d like to teach me, feel free to reach out. Or, at the very least, email me and point me in the direction of some resources so I can teach myself.
6Admittedly, it was a not an easy or enviable task to design a portfolio for everyone.