Objective research to aid investing decisions

Value Investing Strategy (Strategy Overview)

Allocations for May 2020 (Preliminary)

Momentum Investing Strategy (Strategy Overview)

Allocations for May 2020 (Final)
1st ETF 2nd ETF 3rd ETF

Add Position Stop-loss to SACEMS?

| | Posted in: Momentum Investing, Strategic Allocation, Technical Trading

Does adding a position stop-loss rule improve the performance of the “Simple Asset Class ETF Momentum Strategy” (SACEMS) by avoiding some downside volatility? SACEMS each months picks winners from among the a set of eight asset class exchange-traded fund (ETF) proxies plus cash based on past returns over a specified interval. To investigate the value of stop-losses, we augment SACEMS with a simple rule that: (1) exits to Cash from any current winner ETF when its intra-month return falls below a specified threshold; and, (2) re-sets positions per winners at the end of the month. We focus on gross compound annual growth rate (CAGR) and gross maximum drawdown (MaxDD) as key performance statistics for the Top 1, equally weighted (EW) Top 2 and EW Top 3 portfolios of monthly winners. Using monthly total (dividend-adjusted) returns and intra-month drawdowns for the specified assets during February 2006 through March 2020, we find that:

Specific trading assumptions are:

  • At the end of each month, reform SACEMS portfolios using ETFs with the highest total returns over the past four months.
  • If intra-month drawdown for any current position falls below the stop-loss threshold, sell that position at the threshold and put proceeds in Cash. Otherwise, hold all positions to the end of the month.
  • Ignore trading frictions for monthly portfolio reformations and stop-outs.
  • Ignore return on Cash while stopped out (close to accurate during the available sample period).

The following chart summarizes numbers of stop-outs for first, second and third ranks of ETFs for intra-month stop-loss thresholds ranging from -1% to -23% over the available sample period. Numbers of stop-outs decrease rapidly as stop-loss threshold deepens. For example, the total number of stop-outs for threshold -5% (-10%) is 116 (28) over the sample period.

Results suggest that top-ranked ETFs are more likely to stop out than those ranked second or third, perhaps because winners tend to be more volatile.

How do these stop-outs translate to SACEMS portfolio performance?

The next chart summarizes CAGRs for SACEMS Top 1, EW Top 2 and EW Top 3 portfolios for intra-month stop-loss thresholds ranging from -1% to -23% over the available sample period. The Baseline portfolio CAGRs are for no stop-outs. CAGRs mostly increase with magnitude of the stop-loss threshold, but only one of 69 thresholds outperforms Baseline portfolios (and then only slightly).

How do stop-losses affect MaxDDs?

The final chart summarizes MaxDDs for SACEMS Top 1, EW Top 2 and EW Top 3 portfolios for intra-month stop-loss thresholds ranging from -1% to -23% over the available sample period. Baseline portfolio MaxDDs are for no stop-outs. Except for threshold -1%, MaxDDs with stop-losses are at least as deep as those for Baseline portfolios. 

The following two tables provide data used to construct the preceding two charts.

In summary, evidence from tests on available data does not support belief that a simple position stop-loss rule improves SACEMS performance.

It may be that stop-losses do not work for ETFs the same way they work for individual stocks. Many ETFs are diversified, such that their risks (volatilities) are inherently more random than those of individual stocks, so intra-month stop-losses tend to freeze bad luck rather than avoid bad trends. In other words, stop-losses tend to miss upside reversions more than miss further downsides.

Cautions regarding findings include:

  • For thresholds of large magnitude, numbers of stop-losses are small, undermining confidence in results. Said differently, the available sample is very short for testing deep thresholds.
  • As noted, analyses ignore trading frictions, but these frictions are generally higher with than without stop-losses.
  • Different ETFs have different volatilities, so different stop-loss thresholds may be optimal across ETFs.
  • Brute force testing of different stop-loss thresholds introduces data snooping bias, such that the best-performing thresholds overstate expectations. A small number of stop-outs amplifies this bias.
  • Some other stop-loss rule (such as different thresholds across ETFs, thresholds adaptive to recent volatility, a portfolio-level threshold or intra-month re-entry) may work better, but brute force experimentation would increase snooping bias, again amplified by small sample size.
Daily Email Updates
Filter Research
  • Research Categories (select one or more)