Worst Drawdown is Always Ahead

Today I want to explain and provide a caveat to the claim “backtesting fails because my worst drawdown is always ahead.” *

Regardless of backtesitng interval, future market conditions can always be worse than historical. The only case where this would not be true is one where the underlying goes to zero. Were that to happen, of course, the underlying would not be tradeable today.

The worst drawdown being ahead is a consideration to be made in trading system development rather than a failure of said development. Backtesting is a phase of trading system development. Monte Carlo simulation might be another phase… position sizing yet another. The latter two can potentially offer better perspective on likely worst-case future drawdowns based on historical backtrades.

Any systematic approach to trading is based on backtesting and live-trading experience. Nevertheless, drawdown is always a nervous time. The deeper the drawdown, the fewer similar occurrences to anything seen in the past. This gives rise to uncertainty at the right edge of the chart. I can hold through a -10%, -25%, etc. correction and say “well this has only happened X, Y, etc. times in the last 15 years,” but to then believe this can in any way forecast a market turnaround is flawed. The chance of Tails on the next coin flip is 50% regardless of how many consecutive Heads have just turned up.

I like the idea of position sizing based on historical drawdowns plus additional margin of error in case of a worse-case future drawdown. Ultimately, though, I am always just crossing my fingers for a reversal in fortune. A drawdown 50% worse than 2008 is enormous and can certainly come to pass. The deeper a correction becomes, the more I will be hoping the Big One is not taking place now.

Two things can benefit me in the unlikely case the worst drawdown in history is in progress. First is the alluring case for insurance. Second is trading with constant position size because the profits already generated can then serve to offset drawdowns proportionally. Other position sizing algorithms will be position sized largest at portfolio highs, which can lead to the largest gross losses.

* — This post was written in May 2016 but never completed. I find this interesting
       to revisit in lieu of my recent algorithmic trading experience.