With no fear of the future and full trust in my model, Geometric Balancing can theoretically achieve eyepopping returns. A backtest with “extreme leverage” produces 56% annual arithmetic average return. To do so, you would have to fully push the strategy with no safety factors, averaging 5x leverage, and driving up to 12x at times. While this is a spectacular result, you would be bat shit crazy to push the strategy that far.
Predicting Volatility
I started studying the predictability of volatility using over 100 years of weekly Dow returns. After developing an early model for predicting the future volatility, I started evaluating its accuracy. An extreme outlier from May 10th, 1940 jumped off the screen. The prediction wasn’t remotely accurate for the following week, missing by over 8 standard deviations.1 What happened here?
Calculations are my own. Results are theoretical, do not represent returns that any investor actually attained, and are not an indicator of future results.
I’m pretty knowledgeable with history, yet I did not know what occurred on this date. No one ever talks about the financial crash of May 1940. Wikipedia doesn’t even recognize a financial crash then. As far as I knew, nothing meaningful happened that day, and I suspect most Americans wouldn’t recognize this date either.
Germany Invades France
Turns out this was the start of Germany’s invasion of France. The Dow Jones fell as fast at the Maginot line, collapsing nearly 16% over the next week.
I don’t think it’s a coincidence that the largest drawdown in the unlevered strategy ends in late September 2001. While I believe the strategy has a fairly good handle on predicting “market” based volatility, no one can predict non-market volatility. The strategy wasn’t prepared for a terrorist attack on New York City, and it would have been totally unprepared for the German invasion as well.
The un-levered portfolio likely would have fallen by well over 10% that week. Bonds didn’t really respond to the invasion, and gold was pegged at $35/oz. The portfolio would have felt the brunt of this fall, nearly in its entirety.
A levered portfolio would have been flattened.
Financially, volatility looked contained leading into that week. The markets were calm in early May 1940. I suspect the extreme leverage strategy would have called for around 6 times leverage.2 A 16% fall in the Dow literally would have decimated the portfolio’s value.3
Financial Data Does Not Predict Non-Financial Events.
This is why I wouldn’t dream of using extreme leverage with Geometric Balancing, even though on paper the returns are through the roof. All models have limits that must be respected.
I can’t predict terrorist attacks, I can’t predict acts of war, and I can’t predict natural disasters. These risks lie totally outside my formulas (and probably anyone else’s). And as such, using leverage leaves the portfolio open to disastrous consequences due to negative, fast moving events outside the financial markets.
Leverage exaggerates prediction errors. You have to be careful with leverage in any strategy because estimates of financial conditions are inherently going to have some error. But there are certain “errors” in the world you’re never going to predict.
Keep leverage limited because over your investing life you’re likely to experience a Blitzkrieg or two.
Geometric Balancing Leverage, Live
I added leverage to the live banner at the top for your viewing pleasure. The basic portfolio components are the same, with a leverage value to show where the strategy would scale to if it could. As an example, x 1.5 means multiply each portfolio component shown by 50%.
Calculations are my own. Results are theoretical, do not represent returns that any investor actually attained, and are not an indicator of future results. Returns do not reflect trading fees. Dividends are re-invested.
This isn’t the 3x leverage strategy discussed in the earlier post, but a tamer version that allows leverage up to 2x and aims to limit drawdowns to 15% (2.85).
The live strategy is tuned for weekly rebalancing. The difference between rebalancing weekly and daily is not overly large when leverage is not involved. It matters much more when leverage is being used. Therefore I post this for entertainment purposes only.4 It is not a recommendation. As this post explains, you can get yourself in real trouble with leverage if you’re not paying attention.
1-To clarify, this is not a typical analysis. My predictions for future volatility are very dynamic. There were worse weeks than the weeks following May 10th, 1940. But my volatility prediction was much more prepared for that volatility. In a way it expected those moves. As an example the week following October 16th, 1987 hit at half the standardized error as May 1940. Black Monday wasn’t exactly “expected”, but the prediction did expect high volatility for the week.
2-The worst week from this perspective within the timeframe of my strategy was the week of October 6th, 2008. The “extreme leverage” strategy held around 1.5x leverage and lost 8% that week. Bonds and gold performed well, helping to offset the steep fall in stocks. The 3x strategy discussed in the last post was holding 30% cash on this date due to drawdown limits and fell half as much.
3-Decimated technically means to break into 10 pieces.
4-I used to trade this strategy. I don’t any longer because my job sometimes doesn’t allow me to make trades for a few days during the week. The strategy with leverage changes rapidly in a few days. I didn’t like the idea of having too much leverage when events changed without the ability to lighten up right away. So I stopped until I am able to rebalance more frequently, or I build a program for the computer to trade for me.
Hello, I think I have reviewed most of the relevant posts on your site through today. This topic seems to be a key weakness in the strategy. The strategy appears to be a volatility timing strategy based on the past month’s volatility. As you outline here, outliers will occur at greater frequencies than you can predict, potentially setting you up for a huge loss depending on your allocation. Have you considered maintaining a minimum allocation to a “riskless” asset in case your volatility timing model breaks?
Sorry this got flagged as spam for some reason. I have not considered maintaining a minimum allocation to the riskless asset, but would understand if someone wanted to. I’m not sure outliers will occur at a “greater” frequency than I expect, but it is certainly possible they will occur which could cause large losses. This is investing, so there’s always a chance of a large loss.
How do you determine your scale/leverage?
Essentially, this:
https://breakingthemarket.com/how-to-balance-a-portfolio/