SA Charts

[SA Home]        [SA Market Commentary]        [Models & Charts]


Click to read how to use trough detection model

This model uses 5 factors during a typical correction and compares them to a multi-decade historical record to determine probabilities we have seen the worst of the correction:
(1). Consecutive Down-Weeks (Fri-to-Fri closes) – DNWK
(2). Proprietary modified De Mark BUY-Setup counts – DEM++
(3). Extent of maximum draw-down in % points – DDN
(4). Duration of the correction in trading sessions – DUR
(5). Amount of established support failures (dead cat bounces or bull-traps) – FAIL

The first 2 factors impute short-term (higher frequency but more false positives) probabilities whilst the last 3 impute medium-term (low frequency, less false positives) probabilities. In our extensive research to create this model we have identified many useful metrics of SA40 corrections that can be used to derive the required probabilities, but many of them are highly correlated, making their inclusion in a multi-factor model less useful. These 5 metrics stand out in both their diversity and their loose correlations. By using multiple loosely correlated characteristics, we can determine optimum probabilities, since corrections are often characterized by extremes in different metrics depending on the reasons and varying personalities and psychological behaviors for the corrective market action. For example, a sharp, sudden collapse and a long drawn out price-bleed will display different extreme characteristics. By monitoring different characteristics we can match extremes in the broadest set of corrections possible.

With respect to attempting to identify statistically rare extremes in the market, we have found this approach more constructive than the use of traditional technical indicators since it is (1) using long-term history to (2) imply comparable probabilities using (3) multiple loosely correlated correction characteristics which are diverse and expressive of market psychology and (4) more intuitive than “a 14-day RSI below 30” for example.

The model is useful for short to medium-term traders looking for tools to asses buy-the-dip opportunities/risks as well as funds and asset managers seeking favorable points in time for regular deployment of funds into the stock market.

The current probabilities for each factor are shown on the chart below. A probability reading of 90% means that only 10% of the time in the historical record has the factor in question reached higher extremes than are currently being shown.We also show a PIVOT STOP (brown line called a TRENDEX) which is paired with improved DeMark BUY setup counts (DEM++) probability factor. Each day the J200 closes below the STOP the DEM++ probability of a trough increases. When the J200 closes above the STOP the DEM++ model considers the downtrend to be over and commences with SELL SETUP counts to determine probability of a top (you can see this on the companion SA40 TOP chart).

Whilst we also display the average probability of all 5 factors on the main chart, it is more instructive to gauge your current opportunity by observing the bottom pane for:
(1). Number of factors showing 90% probability or more (Diffusion)
(2). Average of the highest 3 probabilities.

When the Diffusion is 2 or more and/or the average top-3 probability is 90% or more you have high confidence buy-on-the-dip opportunities characterized by historical extremes. The bottom pane also shows the average probability of the 3 high-frequency factors and the 3 low frequency factors for those wishing to examine different time horizons for the probabilities. More aggressive traders might focus on the high-frequency probability average whilst those with a more conservative and sedate approach to the market might focus on the low-frequency average probability.

A similar approach is used in the companion SA40 TOP chart to impute probabilities of a market top. Since both Trough and Top probability models are always providing probability scores it makes sense to focus on the chart showing the HIGHEST probabilities to determine which one to focus on. If for example the probability of a trough is showing 25% but the companion chart in SA40 TOP is showing a probability of a market top of 70% then we ignore the trough probability and focus on the market top probability as obviously the odds we are in an uptrend are far higher than the odds we are in a downtrend.

[Research Note]

Click to read how to use peak-detection model

This model uses 5 factors during a typical rally and compares them to a multi-decade historical record to determine probabilities we have seen the best of the market rally:
(1). Consecutive Up-Weeks (Fri-to-Fri closes) – UPWK
(2). Proprietary modified De Mark SELL-Setup counts – DEM++
(3). Extent of maximum rally in % points – GAIN
(4). Duration of the rally in trading sessions – DUR
(5). Amount of new-highs (breakouts) – NEWHI

Market rallies are more sedate and drawn-out than corrections and thus this model is not exactly the inverse of the SA40 TROUGH model. The first 2 factors impute short-term (high frequency, but more false positives) market-peak probabilities whilst the last 3 impute medium-term (lower frequency but less false positives) market-peak probabilities. In our extensive research to create this model we have identified many useful metrics of JSE TOP40 rallies or up-trends that can be used to derive the required probabilities, but many of them are highly correlated, making their inclusion in a multi-factor model less useful. These 5 metrics stand out in both their diversity and their loose correlations. By using multiple loosely correlated characteristics, we can determine optimum probabilities, since corrections are often characterized by extremes in different metrics depending on the reasons and varying personalities and psychological behaviors for the melt-up/rally.

With respect to attempting to identify statistically rare extremes in the market rally, we have found this approach more constructive than the use of traditional technical indicators since it is (1) using long-term history to (2) imply comparable probabilities using (3) multiple loosely correlated rally characteristics which are diverse and expressive of market psychology and (4) more intuitive than “a 14-day RSI below 30” for example.

The model is useful for short to medium-term traders looking for tools to asses market risks as well as funds and asset managers seeking favorable points in time for acquiring protection through hedging operations.

The current probabilities for each factor are shown on the chart below. A probability reading of 90% means that only 10% of the time in the historical record has the factor in question reached higher extremes than are currently being shown. We also show a PIVOT STOP (brown line called TRENDEX) which is paired with improved DeMark SELL setup counts (DEM++) probability factor. Each day the SA40 closes above the STOP the DEM++ probability of a peak increases. When the SA40 closes below the STOP the DEM++ model considers the uptrend to be over and commences with BUY SETUP counts to determine probability of a trough (you can see this on the companion SA40 TROUGH chart).

Whilst we also display the average probability of all 5 factors on the main chart, it is probably more instructive to gauge your current opportunity/risk by observing the bottom pane for:
(1). Number of factors showing 90% probability or more (Diffusion)
(2). Average of the highest 3 probabilities.

When the Diffusion is 2 or more and/or the average top-3 probability is 90% or more you have high confidence market-peaks characterized by historical extremes. The bottom pane also shows the average probability of the 2 high-frequency factors and the 4 low frequency factors for those wishing to examine different time horizons for the probabilities. More aggressive traders might focus on the high-frequency probability average whilst those with a more conservative and sedate approach to the market might focus on the low-frequency average probability.

A similar approach is used in the companion SA40 TROUGH chart to impute probabilities of a market trough. Since both Trough and Top probability models are always providing probability scores it makes sense to choose the chart showing the HIGHEST probabilities to determine which one to focus on. If for example the probability of a peak is showing 25% but the companion chart in SA40 TROUGH is showing a probability of a market trough of 60% then we ignore the peak probability and focus on the market trough probability as obviously the odds we are in an downtrend are far higher than the odds we are in a uptrend.

[Research Note]

Click to read how to use trend direction signal lines

You can use the blue & orange lines in the bottom pane (net average probabilities) as more sedate timing signals than the TRENDEX STOP as they have less whipsaws (but more delayed entries as a result) The TRENDEX is a short-term signal and you can view the blue/orange lines as more medium-term timing signals. Signals above 80 and below -80 can be considered extremes (for the orange line only), with negative extremes more useful than positive ones. The blue signal line is by mathematical definition, slightly slower (more sedate) than the orange one, with less whipsaws but slightly more delayed entries and exits. Both signal lines are suitably early in warning of market tops but can be laggards at signalling market entries compared to some of our other market trough timing signals.We consider the blue line the most robust medium-term trend direction signal of the two over the long run. Diffusions give indication of historical extremes (number of 5 probability metrics showing > 90% probability), with readings on right axis of 4 or 5 particularly extreme.
[Research Note]