Indicators, What Are They Good For?

Indicators!  What are They Good For??

As traders, we all have personalities that drive what we like and what we should avoid!  This article is meant to illustrate how to find indicators that are useful for finding market conditions with certain trade characteristics.   It is assumed that the reader has some familiarity with indicators and typical conditions of interest.  This will NOT be a comprehensive article covering all indicators and their uses.

I'm not a details type of person, so I partner with technology to help me take care all the fussy details that go with trading.    Analyzing my own trades is sometimes painful and frustrating.     I've attempted to classify the reasons a trade doesn't work out.    It's really frustrating when I've got the direction correct, but didn't take enough risk and got stopped out.    Sometimes I ask too much and a trade that would have helped the bottom line ends up as a scratch.   Sometimes I get the direction wrong only to observe some obvious signs after the fact.   To that end, I've keep track of four primary metrics and one additional metric that is useful when attempting  indicator combinations.

​I find five metrics to be of use, in no particular order.

1.  Direction
   
2. Risk
3. Reward
4. Odds Density
5. Frequency of Occurence

For this evaluation of the mentioned five metrics, we'll contain the problem/solution space by putting constrains on the data.

• Two markets only, futures Crude Oil and Nasdaq using the front month non-adjusted prices.   The data period starts 2006-06-12  and  ends on 2017-04-28.
• We will only look at the typically active times in the market.  For CL (Crude) this will be between 7:50 and 11:00 CST.  For NQ (Nasdaq) this will be between 8:30 and 11:00 CST
• We will look at one time frame.   We will use a 1 minute bar/candle for indicator inputs and we will use 30 minutes as our time period in which the market may move.
• We used 'stock'/generic settings for almost all indicators.   In a few cases (noted by the name), I changed some of the parameters to see how the effects varied with relationship to the parameters.

​Here is the list of indicators used and the test condition for this article:
The same settings were used for both markets under evaluation.

Name	Condition
​Chaikin A/D Oscillator 13	greater than 80
Chaikin A/D Oscillator 13	less then negative 80.0
​Chaikin A/D Oscillator 7	greater than 80
Chaikin A/D Oscillator 7	less than negative 80.0
​Chande Momentum Osc. 13	less than negative 50.0
Chande Momentum Osc. 13	​ greater than 50.0
​Chande Momentum Osc. 7	less than negative 40.0
Chande Momentum Osc. 7	greater than 40.0
​Commodity Channel Index	greater than 100.0
Commodity Channel Index	less than negative 100.0
​Ultimate Osc	less than 30
Ultimate Osc	greater than 70.0
​aroon osc	up greater than  75.
aroon osc	​Down less than 25.0
​average directional index	greater than 35
average directional index	less than 18.0
​bollingerBands 13	​close greater than upper bband
​bollingerBands 13	close  less than lower bband
​bollingerBands 7	​close greater than upper bband
bollingerBands 7	close  less than lower bband
​hilbert transform trendMode	not trending true
hilbert transform trendMode	trending True
​macd 12 26 9	macd less than 0  And macd less than signal
​macd 12 26 9	macd greater than  0 And macd greater than signal
​ money flow index	less than 25.0
money flow index	greater than  75
​ relative strength index	less than 25.0
relative strength index	greater than 75.0
​ stochastic relative strength	fastd less than 15.0
stochastic relative strength	fastd greater than 85
​williams R 13	​greater than negative 15.0
williams R 13	​  less than negative 85
​williams R 7	greater than negative 15.0
williams R 7	​ less than negative 85

​On each chart, the 'baseline' will be listed so we can compare against each indicator.  The baseline  calculation is performed using all market data and no conditions.

Direction

Lots of traders would love to know what direction to trade in.  What can these indicators tell us about direction?

These charts show us that  Bollinger Band, Ultimate Oscillator and William %R are among the best of this set at detecting market direction while indicators like ADX(average direction index) and Hilbert transform have very little effect on being able to predict direction.   However, direction by itself is nowhere near the complete story.   We must consider some of the other factors to understand what an indicator may tell us.

Risk!

Risk, Risk Risk!    The previous charts showed us that some indicators are much better than others at detecting market direction.   What about risk?   Do any of the indicators affect the risk taken when entering a trade?    When we consider the effect an indicator has, we will consider the mean, the standard deviation, and the mode.  The mode will not be shown, because most indicator don't effect the mode, but we'll discuss the one indicator that does.

These charts tell us that most indicators increase the amount of risk.   The Average Directional Index is one of the view that both decrease the risk AND the standard deviation.   The Hilbert transform to detect trending or not also decreases both the risk and standard deviation, although not by a large amount.    Something very interesting is shown here.     Check out the Bollinger Bands both 7 and 13.   It's shown that they decrease the risk, but they also are at the top of the chart increasing the standard deviation.  The mode (most commonly occurring) (not shown) also increases for the Bollinger Bands.    So what's going on here and what's another way to look at this?    If the mean decreased but the standard deviation increases, this indicates that we should take a look at the distribution for the risk data.  To save space, I'll only look at the CL market, but the same effect is present in NQ also.

Note the difference in shape of the histograms.   The top chart is the baseline, capped at .50 cents risk.   All risks are assumed to be at lease .02 cents (thus the spike starting at .02).    The larger standard deviation of the Bollinger Bands gives a fatter/flatter distribution,  which is enough to move the mode several pennies (ticks for CL) larger.    Notice the ADX histogram can reduce risk but the shape of the histogram is very similar.   Risk by itself is somewhat boring without looking at the reward parts too.

Reward

In the same way that we can look at risk, we can also examine reward.   This means we'll be seeing both the mean and standard deviation.  Two of the indicators moved the mode and I'll note those at the end of these charts.

These charts indicate that many of the indicators decrease the average reward but increase the standard deviation.   It's worth noting that the average directional index both decreases the risk and increases the reward while decreasing the standard deviation of the reward.   The mode was increased by both the Chande Momentum Osc. 13  [(indData.cmo < -50.0)] and relative strength index  [(indData.rsi < 25.0)].
We can look at risk and reward individually, but we can also look at them as a ratio.

Odds Density

The previous charts showed the effects that these indicators have on risk and reward, but what about density?    Do these indicators have an effect on how likely it is that *something* will happen after you enter the market?    Shown in the charts is a 'No Odds' situation.    This describes a condition where the trader enters the market and both directions carry more risk than reward.    In other words, probably in the middle of volatile congestion.  This condition happens more on a shorter time frame since the market generally moves one way or the other given enough time.

Most indicators increase the possibility of no odds (bad) and decrease the odds density of have a ratio of 1.5 to 1 or higher for the tested conditions by varying amounts.   Again, this isn't the complete story.   We need to consider how often these
indicators happen, both on a daily basis and number of raw minutes. 

Frequency of Occurrence (Filtering)

The following charts show how often the conditions indicate happen.   Two charts are used, both in percentage.   Once shows the percentage of minutes while the other shows the number of unique days.     If the number of days is low, we might consider either changing the indicator parameters OR investigate to see if the indicator is telling us about a certain market condition or type.

These charts show that Ultimate Oscillator as well as RSI occurs the least, both in terms of minutes and day.   Most of the other indicators happen on greater than 95 percent of the days

Concluding thoughts

Trading is a multi-dimensional problem.   You can combine indicators to create
trades with a certain entry condition.    You can also study the sequences.   Let's say you enter a trade based on have a low ADX to minimize your risk as shown by the graphs.    The trade then moves into one of the Williams Percent R conditions.   You now have some edge on the likely direction (for the next 1/2 hour), but knowing that your risk has gone up, you can keep your stop a little looser.   Bollinger Bands are great at detecting direction, but that knowledge comes at a cost in terms of risk and reward.    You can use use the chart data to find a lower risk entry by using either a low ADX or not being too far to either side on a William Percent R.    The combinations are endless!   It's nice to see that these indicators do roughly the same thing on both markets, but one can use parameter tuning to be the most effective for a given market.     Thoughts, comments, questions???