Stock Market Forecast Update

*********************

EASY NOTE: I offer email newsletters documenting portfolio simulations that apply a concept with solid backtesting and intuitively sound principles.  Click here to learn more about the newsletters or sign up to receive them.  If they’re not outperforming the S&P 500 … they’re free!

*********************

I have updated my Stock Market Forecast page with the latest systems I’m testing.  You can read the explanation there in detail.  The quick summary is that I’m doing two primary models (4-week and 13-week), both using a linear regression model (a statistical way of finding a straight-line relationship between a set of variables and a calculated outcome) that involves four different technical analysis data points.

I graph the stock market forecast for each model over the coming 4-week or 13-week period.

Lastly, I will issue two weighted combination forecasts each week in my update post, each of which makes only one “official” forecast for the record book.  One represents a weighted average between where the two models think the S&P 500 will close this upcoming week.  The weights are based on how sure each model is – that is, using the “r-squared” for the regression analysis.  The other is a “Headline Adjusted” model, which tries to account for the fact that extreme and unforeseen events can throw off the models.  So, I remove data that seems affected by such effects and keep the more pure data.  But this model will also make only one forecast, one week in advance.

Performance of Last Week’s Forecast

Weekly Direction of the S&P 500

Correct:   4-week   /   13-week   /   Weight-Adjusted Combo   /   Headline-Adjusted

Incorrect:   None

Accuracy of the Weight-Adjusted Combination Models

Regular Weight-Adjusted Combination: 0.97 percent too pessimistic

Headline-Adjusted Combination: 0.62 percent too optimistic

Accuracy of Individual Models

4-week Model:  1.40 percent too pessimistic

Correct Prediction of S&P 500 Direction thru Last Week’s Close:  3 out of 4 predictions

Notes: All four times it made a forecast, it was too pessimistic (by 1.39 to 1.46 percent).  It got the direction from that point through the forecast date correct on 3 of 4 attempts.

13-week Model:  0.83 percent too pessimistic

Correct Prediction of S&P 500 Direction thru Last Week’s Close:  9 out of 13 predictions

Notes: The model started out about 1.5 percent too pessimistic and after getting worse in the wrong direction, it eventually improved to 0.8 percent too pessimistic, its most accurate guess being its last one.  It got the direction right on 9 of its 13 attempts, including 5 of its final 6 tries.

Estimated Effect of Headlines on Current Market Value

NOTE: This is based on a calculation I do after running the current week’s headline-adjusted forecasts.

4-week Model: Positive effect of  1.8 percent   (up 2.4 percent from last week)

13-week Model: Positive effect of 0.4 percent   (down 2.0 percent from last week)

Notes: The headline effect went up an average of about 0.2 percent from last week, so headlines are artificially elevating the market by about 1.1 percent (equally-weighted average of the two models).  Note: This may not be the same as the difference between the “weighted combo” and “headline-adjusted” forecasts because those forecasts are weighted according to the accuracy of each model.

Buy-Sell Simulation – Based on Headline-Adjusted Model

The simulated weekly trade using the Headline-Adjusted weighted average of the two primary models can be found on the “HA Buy-Sell Sim” tab of the spreadsheet

Compound Annualized Growth Rate (CAGR) since July 12, 2013:

  • Headline-Adjusted Model: (+) 27.43 percent
  • S&P 500 = (+) 3.21 percent during same period

Latest Weekly Profit/Loss vs S&P 500 buy-and-hold:  Gain of $108

Cumulative Profit/Loss vs S&P 500 buy-and-hold:  Gain of $471 (excluding costs of trading, initial $10,000 investment on July 12, 2013)

Wins and Losses vs S&P 500 buy-and-hold (Win = $50 gain or more, Loss = $50 loss or worse, Tie = Anything in between): 50 pct

Notes: The model correctly wanted to sell for the week, but the model was never allowed to go short per my rule about price being at signal level, thus it gained ground on the declining S&P 500 but not quite as much as it “should” have.  The system’s current streak stands at 1 win(any weeks classified as a tie aren’t considered in streak).

Stock Market Forecast Summary for Upcoming Week

Here’s the breakdown:

Forecast Change in S&P 500 This Week (September 30 – October 4, 2013) 4-Week Model 13-Week Model Weighted Average
Standard  (probably where the market “should” be) down 0.59 pct down 0.41 pct down 0.46 pct
Headline Adjusted   (closer to where market WILL be) up 1.21 pct unchanged up 0.32 pct (** my preferred prediction **)

The weighted average forecast for the two models says that on the close of October 4, 2013, the S&P 500 should be 1,684.03 – which translates to a 0.46 percent drop this week.  Adjusted for headlines, we see the market more likely will be 1,697.13 – which translates to a 0.32 percent rise this week.  Check out the Stock Market Forecast page for exact numbers and charts.

What Does Past Performance Tell Us About Upcoming Week?

In general, I am not suggesting that this type of analysis overrides the forecasts above – just another thing to consider.  My calculations suggest the models’ forecast deviation from the actual close show the following correlations with what the S&P 500 does the following week:

Forecast Model Correlation Type* Confidence Level Suggested S&P 500 Change This Week
4-Week Positive 81 percent + 1.40 percent
13-Week Positive 81 percent + 0.80 percent
Weighted Combo Positive 84 percent + 0.97 percent
Headline-Adjusted Weighted Combo Below 80 percent

*A “positive” correlation means that when the model is too optimistic, the S&P 500 performs better the following week than if the opposite were true, although it doesn’t necessarily mean the S&P 500 has dropped if the model was too pessimistic – just that it did worse than when the model was too optimistic.  “Negative” just means the opposite.

What If We Looked at Inputs from All Four Models as Potentially Relevant to the Forecast?

Using a regression analysis to combine some or all of my four forecast models into one forecast, my calculations suggest the ups and downs of one or more of the forecast models accounts for some portion of the variability in the actual S&P 500 close each week.

Forecast Model (Performance of its Final Forecast Only) Relevant in Explaining S&P 500 Ups and Downs?
4-Week Yes
13-Week No
Weighted Combo Yes
Headline-Adjusted Weighted Combo Yes

Suggested S&P 500 Close Next Week: 1,711.49

Suggested Change in S&P 500 Next Week: 1.2 percent rise

In general, I am not suggesting that this overrides the main forecasts above – just another thing to consider.

*********************

EASY NOTE: I offer email newsletters documenting portfolio simulations that apply a concept with solid backtesting and intuitively sound principles.  Click here to learn more about the newsletters or sign up to receive them.  If they’re not outperforming the S&P 500 … they’re free!

*********************

Evolution of the Forecast

Basic Concept: We’ve had several forecasts for next week’s close from the 13-week and 4-week models.  The evolution of the forecast could give us a sense for whether the forecasts are too high or too low.

4-week Model: The forecast has been fairly flat, so this forecast probably wouldn’t change if we had another round of forecasting.

13-week Model: The forecast has been flat over the last several weeks, so this forecast probably wouldn’t change if we had another round of forecasting.

Bottom Line of Forecast Evolution: The evolution of the forecasts suggests the forecasts probably wouldn’t change if we had another round of forecasting.

You can see the “evolution” data on “Forecast Archive” section of the Stock Market Forecast page by clicking on the tabs at the bottom of the chart.

Share