Skip to main content
Track Sharpe ratio
3 / 12
Search the archiveLibrary
Library

1995issue C021-10

Evaluating mechanical switch rules with a stop-loss order and Sharpe ratio

A historical Mechanical trading system treated entry, cash abstention, and a Stop-loss order as one procedure, then compared that full path with buy-and-hold using a Sharpe ratio instead of isolated indicator stories.

  • A ranked 26-week yield-change map was treated as roughly linear, and the extreme tails were used as the informative region for later timing rules.
  • A one-rule yield switch, and later a dual-indicator switch, specified buying, moving to cash, and staying out as a single Mechanical trading system.
  • A Stop-loss order on the dual-indicator path bounded the largest single loss and the worst peak-to-valley equity drop relative to buy-and-hold.
  • A monthly nonannualized Sharpe ratio compared the whole invested-and-cash path with a dividend-preserving buy-and-hold baseline.
Entries in this reading3 entries
Current trackSharpe ratio14 readingsStop-loss order144 readingsMechanical trading system158 readings

Read the switch as one procedure

Editorial interpretation: judge a complete switch as one testable object. Entry, exit, time in cash, and a pre-set loss bound belong to the same Mechanical trading system. The useful comparison is the whole path against buy-and-hold with a Sharpe ratio, not a story about any single indicator.

Yield change ranked against later stock moves

A weekly 26-week percent change in long-term government bond yields was paired with the next 26-week equity-index move. Those pairs were ranked and split into ten equal groups. The pattern was roughly linear: faster yield declines lined up with stronger subsequent stock gains, and faster yield rises lined up with weaker or negative subsequent stock results.

The extreme tails of that yield-change distribution were treated as the most informative region for later timing rules, because turning points in stocks often lined up with periods of rapidly changing rates.

A one-rule yield switch

A one-rule switch bought stocks after a 4% 26-week yield decline and moved to cash after a 14% 26-week yield rise. It was tested on a no-load index fund with dividends reinvested in the market and cash parked in commercial paper, so the buy-and-hold comparison would not drop the dividend stream.

That one-rule procedure stayed in the market 72% of the time from 1980 and 74% of the time from 1957, and every recorded buy signal in those windows was profitable in the backtest.

An adjusted short-sale ratio

A public-to-specialist short-sale ratio was recentered as a 10-week average divided by a 261-week average so that later derivative-era readings could be compared with their own five-year norm rather than with an unadjusted historical scale.

The same ten-group ranking applied to the adjusted short-sale ratio showed subsequent 26-week index gains rising toward the high end of the ratio and falling toward the low end, with the top group near a 24% average annualized gain and the bottom group below 1%.

Two non-price rules, a Stop-loss order, and a Sharpe ratio

A dual-indicator switch combined the two independent, non-price inputs and added a Stop-loss order. That Stop-loss order fired seven times, kept the largest single loss under 4%, and limited the worst peak-to-valley equity drop to 16% in late 1974 versus 45% for buy-and-hold.

Because the dual-indicator procedure was invested only 51% of the time and otherwise earned cash interest, its monthly nonannualized Sharpe ratio was 0.24 against 0.11 for buy-and-hold, or more than twice the tested return per unit of risk.

S&P 500 later return by 26-week bond-yield change

When the 26-week change in long Treasury yields is ranked into ten equal groups, later S&P 500 results fade as yields rise. The source states the tails exactly: yields down 14.3% over 26 weeks go with an 18% average annualized stock gain; yields up 17.3% go with a −4.2% average annualized stock change. Remaining decile bars are read from Figure 1. Traders should treat the extreme tails as the usable signal, not the middle of the distribution.
When the 26-week change in long Treasury yields is ranked into ten equal groups, later S&P 500 results fade as yields rise. The source states the tails exactly: yields down 14.3% over 26 weeks go with an 18% average annualized stock gain; yields up 17.3% go with a −4.2% average annualized stock change. Remaining decile bars are read from Figure 1. Traders should treat the extreme tails as the usable signal, not the middle of the distribution.S&P 500 vs long-term U.S. Treasury bond yield · 26-week · 1957-01-01T00:00:00.000Z to 1994-12-31T00:00:00.000Z

About 1,950 weekly observations from 1957 to mid-1994, ranked by 26-week percent change in the long-term U.S. government bond yield and grouped into ten equal clusters. Each observation is paired with the subsequent 26-week S&P 500 change, shown here as a mean annualized percent change.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
3 of 12 in the Sharpe ratio track
19951-5 pp.Next on Sharpe ratioIntermediate-term allocation with drawdown filtersA working allocation design may be fully invested, fully defensive, or anywhere between, and is meant to catch major uptrends and avoid major downtrends on an intermediate-term horizon.
All readings on this track · 12 readings
  1. 1986Auditing stochastic crossovers with moving-average baselines
  2. 1994Evaluating system changes with chi-square, Sharpe, and leverage
  3. 1995Evaluating mechanical switch rules with a stop-loss order and Sharpe ratio
  4. 1995Intermediate-term allocation with drawdown filters
  5. 1996Evaluating a multi-market book without picking winners
  6. 1996Regime-aware allocation beyond a single equity trend
  7. 1997Evaluating managed futures as portfolio diversifiers
  8. 2008Audit an out-of-the-money covered-call overlay against a Sharpe control
  9. 2013Constructing the Sharpe ratio as return over variability
  10. 2014Expected value and bet size are separate controls
  11. 2015Constructing a Sharpe-style score from profit and loss variability
  12. 2019Continuous futures series and long-horizon allocation evaluation
By fieldEntry and exit rules206 readingsBy year199575 readings
All 14 readings tagged Sharpe ratio
Also on Sharpe ratio5 readings
1984Evaluating managed account portfolios on the risk-return frontier1986Auditing stochastic crossovers with moving-average baselines1994Evaluating system changes with chi-square, Sharpe, and leverage1995Intermediate-term allocation with drawdown filters1996Evaluating a multi-market book without picking winners