As of 09/25/2020
Indus: 27,174 +358.52 +1.3%
Trans: 11,270 +152.52 +1.4%
Utils: 808 +12.20 +1.5%
Nasdaq: 10,914 +241.29 +2.3%
S&P 500: 3,298 +51.87 +1.6%

YTD
4.8%
+3.4%
8.1%
+21.6%
+2.1%

28,650 or 26,000 by 10/15/2020
11,750 or 10,600 by 10/01/2020
845 or 775 by 10/01/2020
11,800 or 10,400 by 10/01/2020
3,600 or 3,200 by 10/01/2020

As of 09/25/2020
Indus: 27,174 +358.52 +1.3%
Trans: 11,270 +152.52 +1.4%
Utils: 808 +12.20 +1.5%
Nasdaq: 10,914 +241.29 +2.3%
S&P 500: 3,298 +51.87 +1.6%

YTD
4.8%
+3.4%
8.1%
+21.6%
+2.1%
 
28,650 or 26,000 by 10/15/2020
11,750 or 10,600 by 10/01/2020
845 or 775 by 10/01/2020
11,800 or 10,400 by 10/01/2020
3,600 or 3,200 by 10/01/2020
 
This article looks at the historical performance of the Dow Jones industrials over the last 10 years according to forecasted behavior and how the Dow is expected to move in the coming decade.
Most recent update: 1/2/2020.
The December 2011 issue of Active Trader magazine had an article titled, "Looking ahead to 2012" by Larry Williams. In it, he describes a method to forecast the future by using historical price behavior. In an earlier article, "Cycles and seasonals: A 2011 roadmap, January 2011," he provides a similar review.
Both articles are based on the work of Edgar Lawrence Smith in the 1930s. Smith said that the stock market followed a 10year cycle. Each year tended to repeat the behavior of the year a decade earlier. In other words, if you averaged all years ending in 1 (2001, 1991, 1981 and so on), that would give you a forecast for 2011. For 2012, you'd make a similar average, only use 2002, 1992, 1982, and so on.
That's what I did.
While the approach sounds easy, it does have some problems as you move from the monthly to weekly to daily scales. Monthly is easy. You take the closing price at the end of each month for all years ending in the target year (meaning 2001, 1991, 1981 and so on for years ending in 1) to get the forecast. Events like 9/11 where the markets were closed from 9/11/2001 to 9/16/2011 didn't matter much. I used the close closest to the end of the month. However, on the weekly and daily scales, those become a problem.
The daily scale is worse. January 3 might be a Tuesday in one year but it's a weekend in another. The average of those will be a mess because you could be leaving off big numbers. For example, back in 1928, the Dow was below 300. Today it's over 24,000. Leaving off the 24,000 reading will make the average of what remains much lower. Thus, the line tends to jump all over the place.
In that situation, I just used the prior day's close and copied that into the gap and averaged the values as normal. That smoothed out the curve.
Let me also say that my data only goes back to October 1928 (which I discard since it's not a full year). Williams appears to use data going back to 1900 and perhaps earlier. The peaks and valleys you will see on my charts may be earlier or later than his. He may show an up trend and my charts do not. It's because of the missing data. For example, my charts use 8 samples and his use 10. That may not sound like much but it's huge when you are dealing with numbers that range so widely.
In some cases, I removed the 20072009 bear market influence on the predictions. By that, I mean if I was looking for 2018's prediction, I'd skip any data from 2008 (which was in the middle of the bear market). That means the chart will show a gap (big rise or drop) because we've lopped off a sample. In that situation, I compute an adjustment factor which plots the new year starting at the same value.
I also do this for some of the charts shown below. I want to begin the prediction and the actual price series (if there is one) using the same value at the start of the year. That way, I can tell what the predicted ending year's value will be, as well as the yearly high and low values. The adjustment factor is this:
Adjustment Factor = (Prior Close)/(Predicted Close).
Each new prediction is multiplied by the adjustment factor once it's found. By that, I mean I only find one adjustment factor, the first one of the first year charted or if I remove the 2007 to 2009 bear market, I'll adjust the values to compensate for the missing years (and I remove the entire 2007, 2008, and 2009 years, not just the bear market from the prediction).
Simply put, this adjustment factor allows me to begin plotting the actual prices and the predicted values at or near the same value as the actual price (or as the last predicted close).
Let me also say that I often don't use the first value in the prediction when computing the adjustment. Why? Because January 2 is sometimes the first trading day of the year. Sometimes it's a weekend. When you average the numbers together, the first few predictions might be, say, 20,000 when later predictions are all in the 24,000 range. Thus, I wait for the daytoday prediction to vary less than 5% before using the predicted value in the formula. Often, on the daily scale, that means waiting one or two additional days. And that's why some of the charts don't show the predicted line starting exactly at the closing price of the first day's actual price value.
Once I have the adjustment factor for the daily scale, I also use the same adjustment factor for the weekly and monthly scales. That way, all of the charts begin at the same value. And that makes the predicted yearly high and low closer to that shown on the daily scale.
All of these adjustments means the predictions of my charts versus someone else's will likely be different. But if you follow these rules, using the same data as I am, you'll get the same result.
One other thing. If you compare the highest (or lowest) predicted closing price on the daily scale, it probably won't match the predicted close of the weekly or monthly scales. Why not? Because the weekly scale uses Friday's close and the monthly scale uses the last trading day of the month. Those can be far from the highest/lowest close found on the daily scale.
Date  Close  Comments 
12/31/1936  179.90  Note: First trading day is 1/4/1937 
1/2/1947  176.39  
1/2/1957  496.03  
12/30/1966  785.69  First trading day is 1/3/1967 
12/31/1976  1004.65  First trading day is 1/3/1977 
1/2/1987  1,927.31  
1/2/1997  6,442.49  
12/29/2006  12,463.15  First trading day is 1/3/2007 
2,934.45  Average of above numbers 
Let's discuss a few examples for the Dow Industrials.
Suppose we want to make a prediction for 2017. We start with 1/1/2017 but that's a holiday. So we skip that day and start with 1/2/2017.
The data I have goes back to late 1928, but we start with 1937, the first full year of data ending with a 7. The table on the right shows what I have.
When no trading occurred on the target date (1/2/year), then I used the prior close.
The average of those numbers is 2,934.45. That's far away from the current Dow's close of 24,386.
Let's compute the next day's prediction: 1/3/2017. I show the data in the table.
Date  Close  Comments 
12/31/1936  179.90  Note: First trading day is 1/4/1937. Use prior close 
1/3/1947  176.76  
1/3/1957  499.20  
1/3/1967  786.41  
1/3/1977  999.75  
1/2/1987  1,927.31  1/3/87 is weekend. Prior close is this one 
1/3/1997  6,544.09  
1/3/2007  12,474.52  
1/3/2017  Not used: We're trying to predict it  
2,948.49  Average of above numbers 
The average is 2,948.49
Because the two predictions are close, within 5% of each other, I use the first predicted close in the calculation for the adjustment factor. The adjustment factor allows us to plot the actual price and predicted price using the same scale. The first value of the two will be the same.
The closing price of the Dow industrials on the first trading day of 2017 is 19,881.76, so the adjustment factor would be AF = 19881.76/2934.45 or 6.77529.
If you multiply each predicted price for the remainder of the year by 6.77529, you'll get an adjusted price normalized to the current price. So the first plotted point would be 2,934.45 (the prediction for 1/2/2017) times 6.77529 or 19,881.76, which matches the first closing price of the Dow.
The next day's predicted close would be 6.77529 x 2,948.49 or 19976.87.
You'd continue this method for the remainder of the year and any succeeding year. Note that if you're plotting multiple years, you only calculate the adjustment factor for the first day of the first year. Then multiply that value by each predicted close for all remaining data (years).
This is the predicted path of the Dow industrials this year. It's a bumpy ride and it ends near where it started, if the forecast is correct. The Dow drops going into March and stages a recovery starting in July. Isn't there a saying to buy in May and go away? Try late May this year or early July.
This chart shows the longer term prediction as of 12/31/2018.
This chart shows weakness in 2019 followed by nice upward run to early 2021. A little turbulence sets in for about a year until mid 2022. Then it's up, up, and away until 2028 (notice a weak year happens a decade after 2018...).
Notice a difference between the predicted low price on the monthly scale (22,159) versus the close on the daily scale (prior chart, 21,526). The monthly scale (this chart) is more accurate because the daily scale can have missing data (like the week's close during 9/11).
This is the forecast (red line) and actual performance for 2019.
The red line is the forecast for 2018 and the black bars are the Dow industrials on the daily scale.
Earlier in 2008, probably around the August blog post, I received an email questioning the inclusion of 20072009 bear market data in the forecast. I thought the person made a good case, that the 2008 bear market wouldn't recur this year. So I removed the bear market data and posted the results.
Clearly that was a mistake. If you include the 2008 bear market data in the forecast, then the above chart is what you get.
Notice that if you invert the red line, it tracks quite closely the Dow industrials. The ending close isn't the same, but it's a very good prediction in my view.
Unfortunately, I've no idea ahead of time whether we'll need to invert the forecast or not.
The red line is the forecast for 2018 and the black bars are the Dow industrials on the daily scale. The actual close at year end versus the predicted close I show in the upper right of the chart.
This is the forecast for the Dow in 2016 (red), using the daily scale, accompanied by the actual Dow price action (black). Notice how close the predicted value was to the actual. Wow.
This is the Dow industrials on the daily scale in 2015 (black) and the predicted path (red) of the Dow.
Pictured above is the forecast for the Dow in 2014, daily scale.
This is the Dow industrials on the daily scale in 2013 and how it was predicted to do. Both the prediction and the Dow start the year from the same value.
The method predicted that the year would be a good one for the Dow. It was.
This is the Dow industrials on the daily scale in 2012
In May, the prediction headed lower, but the Dow turned up in June.
This is the Dow industrials on the daily scale in 2011.
The prediction did quite well in 2011. It may have peaked or hit a valley a few weeks out of synch, but it did predict a bumpy year.
This is the Dow industrials on the daily scale in 2010.
The prediction was off this year. The drop predicted to start in September and bottom in October never appeared.
This is the Dow industrials on the daily scale in 2009.
Although the prediction ended the year very close to the actual, it missed the large decline in March.
This is the Dow industrials on the daily scale in 2008.
This was another year that the index and the prediction missed each other. The Dow dropped even as the prediction said it would rise. If you invert the prediction, it would have been closer to reality.
 Thomas Bulkowski
See Also

Support this site! Clicking any of the books (below) takes you to Amazon.com. If you buy ANYTHING while there, they pay for the referral.
A closed mouth gathers no foot.