Identifying Market Bottoms (SPX/SP500) - A Divergence/Convergence Approach On 120 Years Of Data

Technical Analysis approach based on MACD for spotting market bottoms in SPX/SP500/SPY.

KEY TAKEAWAYS

  • Data period: 1900-2023 (Jan 27).
  • MACD Divergence and Convergence principles used in the study.
  • This approach delivers 83% to 100% accuracy in identifying market bottoms/market reversals depending on the setup used.
  • The approach provides data on possible bearish continuations as well.

DEFINITIONS

  • Time Frame – Weekly chart.
  • Bear Market – A bear market is defined as price decline of more than 20% from the all time high. A new bear market is not accounted for until the price reaches and forms a new all time high first.
  • Divergence – Price forms lower lows, MACD MA’s  form higher lows.
  • Convergence – Price breaks the middle point of a divergence.
  • Convergence = market bottom identified / go long.
 
 A signal (convergence) is considered a successful one if the price reached the All Time High before making a new low.

THE GOAL

The question is simple: Does the MACD reversal setup hold any predictive qualities when applied to the higher time frames?

In other words, can we spot a market reversal in SPX/SPY/SP500 using a technical analysis approach such as the MACD divergence/convergence setup?

 

SPX/SP500/SPY STATISTICS

Under the above assumptions and definitions, SPX had a total of 20 bear markets in the past 120 years. I ran the analysis with the following criteria:

  • Was the divergence completed inside a bear market?
  • Completed divergence.
  • Completed convergence.
  • New All Time High reached.
  • New low formed.
  • Was there Recession in play?
Bear Market? Div Completed Convergence Trigger New ATH After Conv New Low Div Forms During or Slightly After Recession?
Y Oct 1896 Aug 1897 Y - Y
Y Jul 1918 - N Y Y
N Oct 1960 Jan 1961 Y - Y
N Aug 1973 Oct 1973 N Y N
Y May 1974 - N Y Y
Y Aug 1982 Aug 1982 Y - Y
Y Oct 2001 - N Y Y
Y Oct 2002 May 2003 Y N Y
Y Jul 2008 - N Y Y
Y Mar 2009 Jun 2009 Y N Y
Y Oct 2022 ? ? ? ?
  • Out of the 20 bear markets, MACD formed a divergence 10 times (excluding the one from the current bear market rally in October 2022).
  • 8 out of the 10 divergences were completed inside a Bear Market.
  • A convergence (signal) was formed 6 times out of the 10 divergence occurrences. The green rows + the orange one.

What is interesting to be noted here is that the only time the convergence signal failed was when it happened in a bull market and there was no recession at the time when the divergence was completed nor before that. This is the August 1973 signal.

It appears that two factors could influence the accuracy of the signal based on the data gather so far:

  • Is the signal coming in the midst of a recession or slightly after that?
  • Is the signal coming in a bear market or not? 

Since there are only two signals (October 1960 and August 1973) that occurred during a bull market we can’t really make solid conclusions. Nonetheless, the first one delivered (all time high reached after the signal). Which brings the recession filter to the top of my list. 

RETURNS (%) AND RISK/DRAWDOWN (%)

  • Distance to ATH – the return is measured as the distance between the convergence point to the all time high price level. That doesn’t mean the price didn’t continue further, once the ATH was reached.
  • For the failed signal (1973), i’m measuring the distance between the convergence point and the second swing low of the divergence.
  • Time to ATH – measures the time it took for the price to reach the target and in the case of the failed signal, the time it took to reach the swing low.
  • Drawdown – how much did the price fall (from the entry point = convergence point), after the signal occurred and before it reached the target?M
Convergence Trigger Distance to ATH(%) Time to ATH (weeks/months) Drawdown (%)
Aug 1897 37.2% 165 / 38.5 -4.4%
Jan 1961 4.5% 4 / 1 0%
Oct 1973 -10% (distance to low) 7 / 1.6 (time to low) -
Aug 1982 17.6% 9 / 2.1 -1.9%
May 2003 60.7% 214 / 50 -0.4%
Jun 2009 67% 199 / 46.4 -7.8%
Mean (excl. Oct 1973) 37.4% 118.2 / 27.6 -2.4%

BROKEN DIVERGENCE = CONTINUATION LOWER?

Next I assess the situations where a divergence was completed but the convergence never got triggered (no signal). Instead, the price moved lower, breaking below the second low of the divergence.

What i’m trying to figure out here is whether broken divergences have any predictive value. Does the price continue lower and how far does it go when this happens?

Figures are derived by measuring the decline from the breakout point to the lowest price reached within the given cycle.

Div Completed Decline After Div Break (%)
Jul 1918 -8.2%
Aug 1973 -38.7%
May 1974 -29%
Oct 2001 -18.7%
Jul 2008 -44.4%
Mean 27.8%

CONCLUSION

The convergence signal tends to perform really well in identifying market bottoms in SPX/SP500.

However there are a few considerations and assumptions that should be mentioned:

  • The number of signals is too small, even though the analysis is performed on over 120 years of data. This is understandable given that we are working on weekly chart.
  • Not every market bottom ends with this pattern. Which means if you are waiting for this signal, you might actually miss the opportunity. That said, it is best to perform additional analysis on top of this reversal pattern.
  • The data shows that we can use the opposite signal (breakout in the direction of the divergence, rather than convergence) as a signal for further market decline.

Happy Trading!

Yordan K