The Forecasting Systems Letter
Jeffrey Mishlove

---

The following chart, produced by BioComp Profit's "Chart It" function should, by now, be familiar to the readers of this Letter.  It shows that the daily price changes in the Value Line Arithmetic Index, times minus one, provides a good forecast of S&P 500 contract price changes.

I am intruigued by something that I discovered in this indicator that has, thus far, not occurred with regard to the other leading indicators reported in this Letter.  The chart above is based upon using the signal to determine trades that are placed at the close of the following trading day.  The chart below, however, uses the normal signal, not the reverse, as a leading indicator for trades that are placed at the open of the Globex session -- 23 hours and fifteen minutes earlier.

And, the following chart goes one small step (i.e., 45 minutes earlier) further, by using the Value Line Arithmetic Index as a signal for trades placed at the close of the same day that the signal is derived:

I find it interesting that this signal is the strongest of all and that, furthermore, the other leading indicators reported earlier in this Letter, apparently, do not exhibit a similar pattern.

Here are two more interesting leading indicators -- both calculated in the usual way.  The daily change of the signal is reversed by multiplying it by -1.  Then the signal is converted to either +1 or -1, depending upon whether it is above or below 0.  Then that signal is used to determine whether to buy or sell at the close of the following trading day:

I would have thought that the use of a majority vote to combine the signal from different leading indicators would have produced a more successful result than, say, the best of the leading indicators in the vote.  This was not the case however -- as can be seen from the following chart:

In the above chart, ten different leading indicators were transformed into binary +1 or -1 signals.  These were then added together to produce a signal shown in yellow that was normalized by "Chart It" between +1 and -1.  This result, of course, is less profitable than those reported on October 24, 2002. 

The next chart is the same as that shown above, except that the starting date is January 1, 2002.  Again the result is considerably less than that attained by the strongest individual indicators:

It should be remembered that this analysis pertains only to the majority vote of all ten leading indicators.  I have not yet tested all possible subcombinations of those indicators.  However, I think it is likely that certain pairs of indicators may work well together.

After the majority-vote experiment, shown above, produced very modest results, I decided to simply use the same ten, leading indicators as inputs for a neural network system.  As one would expect, the modeling function of BioComp Profit was able to create -- as it almost always does -- excellent "in-sample" results.  The chart shown below combines both "in-sample" and "out of sample" (starting 1/1/02) periods

Actually, it can be seen that the slope of this equity curve starts to taper off in July 2001, about five months before the start of the "out of sample" period.  This, already, suggests to me that it may well be advisable to look for leading indicators using shorter time frames.  Nevertheless, the "out of sample" results, shown below, are still about as profitable as the results attained by the majority-vote method, shown in the article above.

For purposes of comparison, regarding the above two articles, the chart below shows the trading results that would have been attained -- since January 1, 2002 -- using simply the trading signal provided by the Sine function as applied to the S&P 500 adjusted contract price:

Unlike many of the other leading indicators, there was no need to multiply the Sine by -1.  Also, I found that the binary +/- function did not change the equity performance at all.  As usual, this leading indicator is set to trade at the close of the trading day following the signal.  It is also interesting to notice that the performance of this indicator did not really kick in until about June 1.

BioComp Profit Neural Network S&P 500 Futures Contract "MegaSystem" Forecast:
Tuesday's Close:  Down from Monday's close
Wednesday's Close: Down from Tuesday's close

Nirvana OmniTrader Composite Technical Forecast for the S&P 500 Futures Contract:
Short-term, aggresssive strategy: Down.

NeuroShell Volatility Breakout System, Daily, for S&P Futures Contract: 
Out of the market.

KWIK*POP Daily Trading Signal (As Confirmed by Hourly Chart)
Out of the market

View the Results of Previous Forecasts
Forecasting Systems Letter Archive
Return to ForecastingSystems.com