The full title is “How Markets Really Work: A Quantitative Guide to Stock Market Behavior” by Laurence A. Connors with Cesar Alvarez and Connors Research. This is the 2nd edition of the book.
I was attracted by the catchy title of the book, and picked it up to find out what it is all about. Basically the book back-tested various strategies over a 22-year period and showed that the U.S. market (SPX and NDX) tends to reverse when there is a string of good days or a string of bad days. It is something that I think all traders intuitively know and play in their daily trades, and it is good to finally get some data to back it up.
Each chapter specifies what was tested, plots graphs showing the equity curves following the strategy, and gives long numerical tables summarizing the various statistics such as gain/loss, # winners, #days, % profitable, etc. Due to the many pages of graphs and tables, the actual text does not take up many pages, and is a quick read for someone who just wants to know the high-level results.
One note of caution is that the tests were done on the U.S. markets using SPX and NDX, and the results may not apply for other markets (by instrument or geography). For example, trending strategies work better for the DAX while countertrend strategies work better for the S&P.
Overriding Themes: (1) U.S. Markets Tend to Reverse in the Short-Term, (2) Trade with the Trend
- On a short-term basis, oversold markets tend to move higher over the next few days and overbought markets tend to move lower on a short-term basis over the next few days. Markets do not move in only one direction neither short-term, nor long-term. They move from overbought to oversold and vice versa.
- It is better to be a buyer on short-term weakness versus being a buyer after market strength has occurred.
- The results are more pronounced when the trend is taken into consideration (i.e. position of the index with respect to its 200-day moving average).
- The 2-period RSI (the authors sometimes use ‘period’ rather than ‘day’, as though they are making a distinction, but it seems that their ‘period’ is the same as a day) is the single best indicator for traders.
- The studies were done over a 22-year period (January 1989 to September 2011).
- [my note: this may not be applicable for all markets, the S&P 500 is known to be more short-term mean reverting than say the DAX, so countertrend strategies work better for S&P compared to trending strategies for the DAX]
5-Day / 10-Day Highs and Low [Reversal]
- Buying new 5-day highs in the S&P 500 lost money when exiting one week later.
- Buying 10-day highs lost even more.
- If SPX was trading under its 200-day moving average, the losses would be even more.
- Buying new 5-day lows in the S&P 500 made money when exiting one week later.
- Buying 10-day lows made even more.
- If SPX was trading above its 200-day moving average, the gains would be even more.
- Valid for both S&P 500 index (SPX) and Nasdaq 100 index (NDX).
Higher Highs and Lower Lows [Reversal]
- Buying after the market made 3 or more consecutively higher (intraday) highs, would underperform the “average day” (i.e. overall average daily return of the index, which is the benchmark used for the tests) the next week.
- SPX itself lost money within a week after making a 3 or more day higher high.
- Buying after the market made 3 or more consecutively lower (intraday) lows, would outperform the “average day” after 1-day, 2-days, and 1-week.
- Valid for both SPX and NDX.
Higher Closes and Lower Closes [Reversal]
- After SPX has risen 2 days (or 3 days) in a row, the market has underperformed the benchmark over the next 1 day, 2 days, and 1 week.
- After SPX has declined 2 days (or 3 days) in a row, the market has outperformed the benchmark over the next 1 day, 2 days, and 1 week.
- The effect is stronger after 3 days compared to 2 days (i.e. after 3 days of rise it underperformed even more).
- Valid for both SPX and NDX.
Consecutive Days of Declining Issues Greater than Advancing Issues [Reversal]
- Consecutive days of declining issues greater than advancing issues for both SPX and NDX outperformed the benchmark over the next 1 week, and also outperformed the times when there were consecutive days of advancing issues greater than declining issues.
- Works for both 2 and 3 consecutive days.
- Multiple days of declining issues being greater than advancing issues is a sign that a rally is likely near.
- Valid for both SPX and NDX.
Advancing Issues Outnumbering Declining by 2-1 and 3-1 Margins [Reversal]
- When advancing issues outnumbered declining issues by 2-1, the market lost money over the next week for the SPX.
- When declining issues outnumbered advancing issues by 2-1 and 3-1 margins, the following week has seen gains.
Large Volume Days Alone Carries No Edge
- A single large volume day shows no evidence of any edge after 1 week.
- A large x-day volume shows no evidence of any edge regardless of whether the market was up or down that day.
- Considering the trend (i.e. 200-day moving average) did not impact the results.
Large Market Drops are Often Followed by Immediate Snap Backs
- Large price declines outperform large price gains.
- Large declines in the SPX outperformed the average day by a healthy margin after 1 day, 2 days, and a week when the drop was 1% or more. When the drop was 2% or more, the performance jumped even more significantly.
- Large declines in NDX outperformed large rises.
- Declines when index above the 200-day moving average resulted in greater outperformance
- 2% drops in SPX when above the 200-day moving average led to significant large outperformance. For NDX, it outperformed for 2 days and also 1 week.
New 52-week Highs and Lows [Reversal]
- HILO Index (daily) = Number of new 52-week highs for the day – Number of new 52-week lows for the day
- 1-, 5-, and 10-week lows of the HILO Index outperform 1-, 5-, and 10-week highs of the HILO Index over a 1-week and 2-week period.
Put/Call Ratio Gives Little to No Edge
- Put/call ratio = Total number of puts / Total number of calls, for all index and equity options traded on the Chicago Board of Options Exchange (CBOE). The put/call ratio is widely assumed to be a contrary indicator.
- The tests were done using the 1-month moving average put/call ratio as that is commonly relied on in trading circles.
- Buying when a 20-day high is made (closing data) showed a small edge over a 1-week period. However, 5-day high or 10-day high does not appear to give an edge.
- 5-day lows and 10-day lows both underperformed or equaled the benchmark return. 10-day lows lost money after a week in both SPX and NDX. Hence selling after a 10-day low can be profitable.
Look for Volatility Index (VIX) Closing 10% Above or 5% Below its 10-day Moving Average [Reversal]
- CBOE VIX measures the implied volatility of the S&P 500 options.
- Do not use static levels for buy and sell signals (e.g. sell at low 20s and buy at 30).
- In summer of 2002, VIX crossed 30 and went to 50 over the next few months as the SPX dropped further.
- In late 2003, VIX went into low 20s and went further into the teens as the market continued to rally.
- We compare the VIX versus its 10-day moving average. The 10-day moving average is dynamic and overcomes the problem with static levels.
- The outperformance when VIX closed 10% above its 10-period moving average, and the underperformance when the VIX closed 5% below its 10-period moving average, are very significant.
- When VIX closed 10% or more above its 10-period moving average, the market has outperformed its average gains over 1-day, 2-day, and 1-week. The greater the % above the 10-period moving average, the higher the probability of market reversal and the higher the magnitude.
- When VIX closed 5% or more below its 10-period moving average, the market has lost money over the next week.
- This is good for identifying overbought markets and the times when you should be aggressive in locking in gains and/or avoiding long purchases.
- None of the cumulative net weekly gains for the SPX over the 20-year test period occurred while the VIX closed 5% under its 10-period moving average.
- The results are similar using the VXN (Nasdaq equivalent).
Two-Period RSI [Best Indicator]
- The original 14-period RSI did not show any statistical edge.
- RSI has short-term predictive ability when the period was short and its levels reached extremes.
- Low RSI levels (below 5, especially below 2) have led to significant outperformance in SPX. Extreme low RSI levels below the 200-day moving average have seen healthy short-term gains (oversold dead-cat bounces)
- High RSI levels (above 95) have led to dead money as rallies have often stalled within a few days.
- Valid for both SPX and NDX.
Stocks With High Historical Volatility on Average Lost Money Over the Next One Year
- Momentum stocks (go-go story stocks) are high volatility stocks. The look good on the way up but devastate portfolio returns on the way down.
- Grouping stocks above $5 into 5 equal buckets based on their 100-day historical volatility, the lowest historical volatility stocks lost money 34% of the time (over the next 252 trading days), with an average loss of 19.8%. Over 55% of the stocks with the highest historical volatility lost money with the losers losing on average 44.4% per trade.
- A portfolio manager that simply avoids the stocks in the highest volatility bucket would greatly outperform the market indices since 1995.
- Trade once a week each Friday on the Close.
- SPX above 200-day moving average.
- Stock above 200-day moving average.
- Stock down 2 days in a row (yesterday and today).
- RSI must be under 15.
- Stock’s 100-day historical volatility is below 35.
- One the Close of each Friday, buy up to 10% per position with up to 10 total positions.
- If more than 10 stocks qualify, take the ones with the highest 100-day historical volatility (i.e. those that will move in the short-term).
- Exit the next Friday on the close and rotate into the next round of 10 S&P 500 stocks.
- CAGR of 10.48% from January 1, 2001 to September 30, 2011, versus SPX CAGR of 0.04%.
- Annual returns from -1.14% to 21.41%, versus SPX’s -38.49% to 26.38%.
- Max daily drawdown of -14.98% versus SPX’s -55.25%.
- 0.81 Sharpe Ratio versus -0.03 for SPX.
- 0.41 Correlation to SPX.
- Two small down years in 2002 and 2011, but still beating SPX.
- 70% less volatile than SPX, consistent results.
- Potential improvement
- Results of the studies can be improved dramatically using dynamic exits (e.g. wait for price to close on the other side of their 5-period moving average) instead of static exits (e.g. 1-week later).