The STIX Delta Trade Entry timer (STIX-D) was developed by PowerStocks Labs to allow investors to guage their risks of JSE entry at any point in time. It is a modification of the venerable STIX Breadth Indicator that was published for U.S markets by the Polymetric Report for many decades before they were acquired. It is also a  very accurate trough detector, for both medium-sized and great troughs.

STIX is an acronym for Short Term Trading Index. STIX is a 21-day exponential moving average of daily advances divided by advances plus declines and represented as a percentage :

STIX = EMA(21) of A/(A+D)*100.

Low STIX readings are bearish and high readings are bullish. You will recall the A/(A+D)*100 from our very first trading timers, the BITS family which we publish in JBAR. BITS uses a different EMA duration and was an adaptation of work done by Zweig and Appel for generating signals to enter the market on strength. If you recall, BITS opens trades when the value of A/(A+D)*100 exceeds 60 (a very bullish condition) and holds them open until it declines to 45 or 41.

STIX is best used on stock market indices that have at least 200 active shares per day and lower than this becomes less effective. The JSE just makes this criterion. Apart from the Breadth Impulse (BITS) timing strategy developed by Appel, STIX was originally intended to show overbought and oversold levels of an index and was actually quite effective on the U.S markets. The original uses of STIX were first described in Stocks and Commodities magazine by Fay H. Dworkin "Defining Advance/Decline issues" ,Volume 8, July 1990 pp. 274-278 and by Tushar Chande "Breadth Stix and Other Tricks", Volume 12 May 1994 pp.211-214. Gregory L Morris also gives a review of its effectiveness in "The complete guide to Market Breadth Indicators."

STIX-D is a further adaptation invented by PowerStocks Labs. We use the same EMA as the BITS family of timers optimised for the JSE, but instead of waiting for the signal to reach 60 to issue a BUY signal, we issue a BUY signal every time the signal has INCREASED by a certain variable percentage D. The greater the value of D the better the results. To our minds, this is a better representation of a "Breadth Impulse" but because "Impulse" is associated with the original Appel method we decided to refer to D as a "Delta".

The theory is that when underlying market breadth, as represented by A/(A+D) improved by a percentage D then either a decline-reversal is taking place or the market is getting stronger. The argument is that both these conditions should qualify for JSE entry. One would expect low values of D to obviously generate lots of signals and have less accuracy than high values of D and our extensive back-testing indeed confirms this.

Two sample STIX Delta occurrences are shown below, when D is set to 50%. The chart depicts the period 22/12/2006 to 08/10/2007. This means that the system tracks the daily yellow STIX-D signal line and whenever it sees this line make a reading that is 50% more than a previous local minimum, it generates a BUY signal. The two examples are actually rather excellent entry points, but this is because a D of 50% is rather high and has very good confidence.

Below is another example, but this time D set to a much lower 25%, which relaxes the criteria for a signal somewhat. You can see the results are less accurate. Of the 10 signals below, only 5 or 50% were "accurate":

The next task is to find the optimum values for D. What we did was starting from D=20 (less than this yields very ineffective results) we stepped through the 12-year backtest and measured the effectiveness of the resulting entry points generated by the trading program with respect to a profit made 30 calender (20 trading days) after the entry. The actuarial table below refers to a summary of the results of this testing (we only show values for D in steps of 5 below to keep table short.)

"Delta" refers to the value for D. "Sigs" is how many BUY signals were generated. "Good" is how many signals resulted in a profit after 20 trading days. "Bad" is how many signals resulted in a loss after 20 days. "Win" is the win rate, the % of good signals. "Avg Win" is the size of the average winning trade and "Avg Loss" is size of average losing trade. "PTS UP" is cumulative % points gained in winning trades and "PTS DN" is sum of % points lost in losing trades. "NP" is net points gained and "GL" is the gain to loss ratio of PTS UP divided by PTS DN. The last column shows the "Expectancy" of the system, namely the size of the average trade gain including losing trades.

As expected, higher values of D lead to better results all round, but generate less and less signals. A value of D=80 only triggered 7 signals over 12 years meaning this occurrence is a very rare, but accurate event, resulting in only one "loss". The next few charts show a graphical representation of results for different values of D and their effect on Win rate, Gain/Loss ratio and Expectancy for a 20-trading-day holding period.

BEST D FOR WIN RATE : For optimum win rates, only play signals when D >= 48. For win rates over 75%, choose D 48-55, 60-65 and above 75:

BEST D FOR GAIN/LOSS RATIO : For optimum gain/loss ratios, only play signals when D >= 48. Superior ratios available from 51 to 55.

BEST D FOR TRADE EXPECTANCY : To keep average expectancy above 3% only play signals when D is greater than 43. For superior expectancy, only play signals when D is between 51 and 55 and D greater than 74.

It would appear a D of 48 is your optimum "all-rounder". A 12 year chart of TRIX-D(48) appears below to let you see for yourself what a nice set of signals it produced for investors and traders alike:

With D=48 there are 29 signals with 22 good and 7 bad, giving a 76% win rate for 30 day holding periods. Note that if we extend the holding period to 50 days, the win rate rises to 85%. Remember in all our computations above we have been assuming a 20-trading day holding period but the reality is that longer term investors will be looking at far longer horizons thus making the signals have much higher win rates.

We have taken a slightly different and innovative approach to the signal chart published in the ALSH TIMERS sheet in JBAR. A sample appears below:
The first thing you will notice is that the chart is a full page width in JBAR to show as much history as possible, going back to 2006. The chart is a REAL-TIME BACK-TESTER meaning it is showing all the signals the system would have generated based on TODAY'S reading of the STIX signal. It is not an indication of the systems' accuracy it is an indication of what to expect if you ENTER the JSE NOW.

There are some orange numbers you need to know in the chart:
DELTA : This shows the value of D on the day of the report (19.45% in this instance)
FREQ : How many BUY signals a system would have generated using a D of today (19.45)
PROB.: What percentage of the BUY signals would have resulted in a 20 trading day profit.
EXPT.: What the average expectancy of the system would be using D of today.
G/L : The gain/loss ratio of the system using a D based on today's reading.

There is also a 30-day Confidence rating, ranging from VERY LOW, LOW, LOW-MEDIUM, MEDIUM, HIGH that gives you advice on whether you should enter the JSE now or not. This is not based purely on the expected win-rate but the gain/loss ratio and the average points to be gained as well.

Lastly you will see a SIGNAL PRESENT or NO SIGNAL in the top right hand corner of the chart telling you if a system, using today's value of STIX for the D, has generated a signal or not. You must only enter the JSE when this is showing a SIGNAL PRESENT value. You will see the signal as a green vertical bar in the extreme right of the chart. This is there to ensure you are playing a signal that is true to the mechanical system we are using so that it can be related to all the actuarial charts. There are a number of reasons there will be no signal for a given D, and these are associated with extra rules we have programmed into the system to avoid false signals and turns etc. For example the system never opens a trade if STIX is falling (such as it is in the chart currently.)

So now you have a system that is not fixed to one set of parameters for trade entry, it is actually parameter-less and runs on-the-fly 12 year backtests to let you, the investor or trader decide if he/she likes the odds or not. You can look at the past 4 years signals visually on the chart to see if their accuracy makes you comfortable and you can see some trade stats related to win-rate, gain/loss ratios etc to see if they take your fancy. You can see what our guidance tells you about what we think of the trade and finally if you see a signal firing on the day you can follow the mechanical system into the indicated trade. Finally you can look at the yellow signal line and decide you are going to wait for it to rise a little more since you know this action will increase D and hence your odds of success.

One final word of WARNING : Just because STIX-D says a trade does not look good now DOES NOT MEAN its not a good time to enter the JSE! STIX-D is not the authoritative matter on this topic, even though it may be good at it! It just means that using the STIX STRATEGY with the current days' D reading has generally not yielded good results. It may very well be that some other system such as TroughFinder, Big Dipper or STORM5+ has already flagged a BUY signal but STIX is saying CONFIDENCE IS LOW. We choose most of our trough detection systems since they approach the problem from varying angles and quite often one system will pick up a lovely trough signal while another will be dead silent! What this does mean however is that if you have several systems showing a BUY and STIX also showing you a buy that you can have more confidence in your entry.

Please remember that STIX-D is a HYBRID system. It works of triggers related to a change in underlying advance/decline breadth. This could be as a result of a breadth thrust midway through a strong bull run (we call this a continuation signal) or it could be related to a trough reversal. That is why STIX can offer double the amount of signals of most of our other systems.
Make a Free Website with Yola.