Max Wait

There's a phrase or attitude in and around Java called pasrah.

Maybe it's dying out only surviving in some out of the way places which haven't been totally submerged in bit and bytes.

Pasrah means 'resignation'.

Train doesn't come on time? Pasrah.

Stuck in a humungous queue? Pasrah.

Stranded in your car in the middle of Jakarta due to flooding? Pasrah.

I.e. random acts of gratuitous incompetence are met by pasrah. Perhaps that's changing. Hopefully.

In any case it hasn't changed in finance.

Go buy some fancy financial product and prepare to be underwhelmed!

In the spirit of pasrah, I included a 'Max Wait' analytic in my Lazy Backtesting IDE. Rather than report the maximum peak to trough loss over a time period in the context of much larger gains; better to report how long it may take for the strategy under investigation to end up in the black.

In order to calculate Max Wait, you break your total back test history into contiguous chunks, and find the minimum chunk size where all chunks have positive returns.

If you do this for the S&P 500 over the last 65 years, the Max Wait is 8 years.

I.e. sometimes you would have had to wait longer than 7 years to just turn a profit!

This should give everyone pause for thought when investing in even the humble S&P 500, let alone anything more exotic.

Let's have a look at the Max Wait of a few strategies and explore it further.

Realised Steady Vol

Logic: inversely weight current month's holding vis a vis last month's realised vol

Max Wait: 8 years

Backtesting Period: 55 years

Implied Steady Vol

Logic: inversely weight this month's holding vis a vis the Vix

Max Wait: 6 years

Backtesting Period: 25 years

Skewerage

Logic: weight the implied skew index by month on month changes, short when the skew increases and vice versa

Max Wait: 5 years

Backtesting Period: 25 years

Momentum + Mean Reversion

Logic: go long when previous day's return was positive and momentum detected or when previous return was negative and mean reversion detected. Otherwise short

Max Wait: 7 years

Back test period: 55 years

Steady Vol + Skewerage

Logic: combination of steady vol and skew strategies

Max Wait: 4 years

Backtesting Period: 25 years

Dual Momentum

Logic: each month invest in best performing asset over previous year.

(For this run, I tested with only the S&P 500)

Max Wait: 11 years

Backtesting Period: 55 years

War of Ideas Strategy

Logic: maximise portfolio entropy

Max Wait: 3 years

Backtesting Period: 10 years

(Bubbles moved around for legibility)

What's interesting when you put it all together is how little the Max Wait varies with time.

Sure longer backtest periods lead to longer waits, but not by that much.

For example the S&P 500's Max Wait for 55 years is 8, and for 25 years is 7.

Comparing the same strategies across 10,25 and 55 years reinforces this lack of change in Max Waits. Only Dual Momentum benefits a lot from a reduction in backtest period.

Comparability across different back test periods (within reason) is a property I really like.

Taking the S&P as a benchmark, we'd expect to add a year Max Wait for every doubling of back test period.

Timeliness or lack thereof is super important to avoiding those pasrah moments (if only all investors were as patient as the Javanese) and at the very minimum set expectations properly.